THE INTERSECTION OF ARTIFICIAL INTELLIGENCE IN HUMAN RESOURCE

MANAGEMENT: AN EXPLORATION OF TALENT COMMUNICATION

by

ALEXIS MEGAN VOTTO, M.S.

DISSERTATION
Presented to the Graduate Faculty of
The University of Texas at San Antonio
in Partial Fulfillment
of the Requirements
for the Degree of

DOCTOR OF PHILOSOPHY IN INFORMATION TECHNOLOGY WITH A

CONCENTRATION IN ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING

COMMITTEE MEMBERS:
Rohit Valecha, Ph.D., Chair
H.Raghav Rao, Ph.D., Co-Chair
Paul Rad, Ph.D.
Kathryn Keeton, Ph.D.

THE UNIVERSITY OF TEXAS AT SAN ANTONIO

Alvarez College of Business
Department of Information Systems and Cyber Security
August 2023
 DEDICATION

This dissertation is dedicated to the many hands which have guided and pushed me through
unforeseen challenges…knowing I could make it. Namely, I dedicate this to my husband, Joe.
You’ve been my foundation for when my own footing has failed me, never failing to pick me up
and keep me in the fight. I am not a poet, but if I was, I’d write you the greatest love song in the
world….let this just be a Tribute. I also dedicate this to the many mentors who have dedicated
many hours in my development. Had it not been for your words of encouragement and honesty, I
would not be the Airman I am today. I salute you and thank God I’ve had such wonderful
influences who have consistently embraced my quirks. More importantly, y’all have imparted the
trust of brotherhood/sisterhood which I value to my very soul. Lastly, I dedicate this to my late
mother, Sherry. I miss you so much and know you’ve watched over me during this endeavor.
 ACKNOWLEDGEMENTS

Writing this dissertation was one of the best experiences I could have had because of my

outstanding committee members. The dedication and time spent helping me grow as an

academic and officer within the United States Air Force is unfathomably appreciated. I charged

into this program with a time limit. Understanding this, each of my committee members and the

entire department stepped up to the challenge and did not waiver their expectations. To Dr.

Rohit Valecha and Dr. H. Raghav Rao, my Chair and Co-Chair, respectively. From the start, I

knew we would be a great team. People with Buffalo, NY lineage tend to find each other in the

craziest places, Texas being one of them (Go BILLS!). Words cannot express how much of a

positive influence these two men have on my academic development at UTSA. Although the

work was hard, Dr. Valecha’s optimism and patience allowed me to see the fun in research and

made it unquestionably enjoyable. I have learned how to better value patience within the

scientific process and I credit this to Dr. Valecha’s mentorship. Moreover, Dr. Rao’s expertise

and insight guided our research endeavors to ensure the work we dedicated many hours to was

impactful. Through his challenging seminars and genuine feedback, I have grown better as

scholar. My committee members, Dr. Paul Rad and Dr. Kathryn Keeton have also been essential

to my development within this program. Before I met Dr. Rad, Python would have been another

snake in the jungle, and “AI” would have just been Skynet from The Terminator…or HAL 9000.

He was my first introduction to coding in Python and general Artificial Intelligence. I thank him

sincerely for helping me see the forest through the trees and connecting me with my other

committee members. Furthermore, my many conversations with Dr. Keeton gave me a sense of

applicability and relevance. Her reassurance and experience with military veterans encouraged

iii
me that the research we worked on together would make some waves and possibly help people. I

thank her deeply for her unwavering support throughout this process.

Dr. Nicole Beebe is another faculty member I’d like to highlight and acknowledge. I

know there was a bit of a gamble accepting me into the folds, but I am grateful for her support.

As an Air Force veteran herself, I know she understands the responsibility that comes with

wearing the uniform and I appreciate her putting her faith in my success. I hope to proudly

represent Nicole, our fantastic department, and UTSA as I rush back to the greatest job in the

world. I will showcase the education and hopefully make things better for our Airmen. Lastly,

I’ll echo an acknowledgment from another one of her student’s dissertation; Nicole has always

been a positive force of nature, setting a gold standing in my book for women in science and

technology. Maybe one day when I grow up, I’ll be just as successful as her. I thank her for

supporting me throughout this crazy endeavor and for allowing me to join the Ph.D. ranks.

I’d be remorse not to acknowledge my fantastic cohort. We joined this program at a very

challenging time in our history given the global pandemic. Social distancing and high infection

rates forced us to interact virtually which made socializing difficult. We didn’t “meet” in person

until Dr. Bachura’s teaching seminar in the Summer of 2021, which was the end of our first year.

I cherish the friendships we have nurtured together through a global pandemic, challenging

seminars, and research endeavors. That said, you all have made this experience unforgettable,

and I am truly indebted to you all for helping me navigate this program. I look forward to seeing

you all graduate. You can trust I’ll be in the crowd of many other people who believe in you,

ready to hoist a beer into the air. Keep up the great work and continue to charge forward!

Another acknowledgement I’d like to take a paragraph to acknowledge Lee Ring and my

CrossFit Lobo family. Lee, thank you for continuously being a persistent, respectful, intuitive,

iv
challenging, and knowledgeable influence on my coaching career. You’re a wonderful mentor

and I appreciate your brutal honesty. Words cannot express how much I appreciated our offline

banter on my research and the refreshing insights you could provide as a prior veteran yourself.

To my Lobo Family, you all have provided me with a community I can rely on for advice,

counsel, and the occasional competitive edge. I cherish the special memories we’ve built

together, whether suffering through one of Lee’s workouts or enduring one of my “train-with-

the-trainer” sessions. Your friendship gave me a great sense of community and belonging while

enduring the academic rigor at UTSA. I look forward to the future memories to come!

Last and most certainly not least, I’d like to personally thank all the research assistants,

colleagues, and teams which have helped me collect, clean, and process the data utilized in this

manuscript. Without your attention to detail and willingness to help, I would not be here. Thank

you. Thank you. THANK YOU.

v
“This Master’s Thesis/Recital Document or Doctoral Dissertation was produced in accordance
with guidelines which permit the inclusion as part of the Master’s Thesis/Recital Document or
Doctoral Dissertation the text of an original paper, or papers, submitted for publication. The
Master’s Thesis/Recital Document or Doctoral Dissertation must still conform to all other
requirements explained in the “Guide for the Preparation of a Master’s Thesis/Recital
Document 6 or Doctoral Dissertation at The University of Texas at San Antonio.” It must
include a comprehensive abstract, a full introduction and literature review, and a final overall
conclusion. Additional material (procedural and design data as well as descriptions of
equipment) must be provided in sufficient detail to allow a clear and precise judgment to be
made of the importance and originality of the research reported.

It is acceptable for this Master’s Thesis/Recital Document or Doctoral Dissertation to include as

chapters authentic copies of papers already published, provided these meet type size, margin,
and legibility requirements. In such cases, connecting texts, which provide logical bridges
between different manuscripts, are mandatory. Where the student is not the sole author of a
manuscript, the student is required to make an explicit statement in the introductory material to
that manuscript describing the student’s contribution to the work and acknowledging the
contribution of the other author(s). The approvals of the Supervising Committee which precede
all other material in the Master’s Thesis/Recital Document or Doctoral Dissertation attest to the
accuracy of this statement.”

August 2023

vi
 THE INTERSECTION OF ARTIFICIAL INTELLIGENCE IN HUMAN RESOURCE

MANAGEMENT: AN EXPLORATION OF TALENT COMMUNICATION

Alexis M. Votto, M.S.

The University of Texas at San Antonio, 2023

Supervising Professor: Rohit Valecha, Ph.D.

Within this research, we sought to explore and understand where Artificial Intelligence

(AI) intersects with human resource information systems. We further investigated the AI

phenomenon by leveraging machine learning tools to understand the following :1) where AI

exists within human resource management (HRM) via a systematic literature review *; 2) how to

propose a research framework to investigate data science competencies within private and

federal sector job postings, and 3) how military veterans with data science skills communicate

their experience concerning employability. The following sections discuss these papers further.

Paper 1: Artificial intelligence in tactical human resource management: A systematic

literature review (2021)

The first essay is a systematic literature review investigating where AI exists within

published HRM literature. Within this literature review, we leveraged a 2-phased methodology to

navigate 315,053 articles from various multidisciplinary publication sources. The findings of this

review indicate research opportunities to grow within HRM's competency management systems

(pay and benefits). Furthermore, it highlighted a gap within AI research exploring qualitative

components of recruitment information systems, which spurred the motivation for the rest of this

dissertation.

*
Published: International Journal of Information Management Data Insights, 1(2), 2021

vii
 Paper 2: JC-Compass: A Framework for Conducting Competency-Based Job Posting

Research and Analysis (In Progress) †

The second essay proposes a research framework to conduct competency-based job

description research. This paper leverages a design science approach to propose and explore a

new competency-based job description research method. The results indicate that the federal

sector potentially places a decisive influence on problem-solving terms rather than artificial

intelligence terms. We also discovered that the private sector strongly influenced statistics and

ethics terminology, whereas the federal focused on problem solving and ethics.

Paper 3: Veteran Talent within Data Science: An Exploratory Resume Analysis on

the Employability of Active-Duty Veterans

The third and final paper explores how military veterans seeking jobs within the data

science community communicate their skills and how their employability is affected by their

choice of words within resumes. Understanding that drawdowns are posturing some military

veterans to leave the service and pursue other endeavors, we sought to evaluate how Gulf War II

("post 9/11") veterans advertise their skills relative to how many times they have been

unemployed throughout their resume. In conducting this research, we aspire to provide insight

into how veterans seeking data science jobs could better posture themselves to be more

marketable and provide employers insight into current trends and expectations.

†
Submitted and presented at 2022 Pre-ICIS SIGDSA Symposium, award recipient of “Best Paper”

viii
 TABLE OF CONTENTS

Acknowledgements ........................................................................................................................ iii

Abstract ........................................................................................................................................ vii

List of Tables ................................................................................................................................ xi

List of Figures .............................................................................................................................. xiv

Essay I: Artificial Intelligence in Tactical Human Resource Management: A Systematic

Literature Review. ...........................................................................................................................1

Chapter 1: Abstract ..............................................................................................................2

Chapter 2: Introduction ........................................................................................................3

Chapter 3: Background and Related Work of T-HRIS and AI ...........................................5

Chapter 4: Methodology ...................................................................................................18

Chapter 5: Findings ...........................................................................................................23

Chapter 6: Discussion .......................................................................................................42

Chapter 7: Conclusion........................................................................................................44

Appendix ...........................................................................................................................47

References ..........................................................................................................................49

Essay II: JC-Compass: A Framework for Conducting Competency-Based Job Posting Research

and Analysis ..................................................................................................................................57

Chapter 1: Abstract ............................................................................................................58

Chapter 2: Introduction ......................................................................................................59

Chapter 3: Background ......................................................................................................61

Chapter 4:Introducing Competency-Based HRM as a Kernel Theory for Research .........67

ix
 Chapter 5: General Overview of Text Analysis Methods For Competency Research ......72

Chapter 6: The Job Competency Research Framework Development ..............................76

Chapter 7: The JC-Compass Framework for Job Description Research ...........................83

Chapter 8: Empirical Demonstration of JC-Compass ........................................................85

Chapter 9: Conclusion......................................................................................................113

Appendices .......................................................................................................................115

References ........................................................................................................................124

Essay III: Veteran Talent Within Data Science: An Exploratory Resume Analysis on the

Employability of Active-Duty Veterans ......................................................................................134

Chapter 1: Abstract ..........................................................................................................135

Chapter 2: Introduction ....................................................................................................136

Chapter 3: Literature Review and Theoretical Background ............................................140

Chapter 4: Methodology ..................................................................................................149

Chapter 5: Results ............................................................................................................157

Chapter 6: Post-Hoc Analysis ..........................................................................................165

Chapter 7: Discussion ......................................................................................................173

Chapter 8: Conclusion......................................................................................................176

Appendices .......................................................................................................................177

References ........................................................................................................................180

Vita

x
 LIST OF TABLES

Essay 1:Artificial Intelligence in Tactical Human Resource Management: A Systematic

Literature Review

Table 1 Tasks within Tactical HRIS “Managerial” Categories ............................................8

Table 2 Tasks within Tactical HRIS “Technical” Categories.............................................10

Table 3 Previous AI and HRM Literature Overviews ........................................................13

Table 4 AI Methods ............................................................................................................17

Table 5 Table 5 AI and HRM Key Word Searches ............................................................20

Table 6 AI Representation within SLR ...............................................................................38

Table 7 Analysis Representation.........................................................................................39

Table 7 Empirical Analysis Breakdown .............................................................................40

Essay 2: JC-Compass: A Framework for Conducting Competency-Based Job Posting

Research and Analysis

Table 1 Previous Job Posting NLP Literature Summary ....................................................64

Table 2 Previous Job Posting NLP Research Framework Summary ..................................66

Table 3 Various Definitions of Competency ......................................................................69

Table 4 Data Pre-Processing Steps .....................................................................................80

Table 5 Token Comparison .................................................................................................93

Table 6 Data Similarity Comparison ..................................................................................93

Table 7 Most_Similar Queries ............................................................................................94

Table 8a Results of Inter-Coder Reliability (Private) ...........................................................99

Table 8b Results of Inter-Coder Reliability (Federal) ..........................................................99

Table 9a Results of Preliminary Validation (Private) ...........................................................99

xi
Table 9b Results of Preliminary Validation (Federal) ..........................................................99

Table 10a Results of Batch 4 (Validator 1) ..........................................................................100

Table 10b Results of Batch 4 (Validator 2) .........................................................................101

Table 11a Spearman Correlation Charts (Private) ................................................................102

Table 11b Spearman Correlation Chart (Federal) ................................................................103

Table 12 Private and Federal TF-IDF Ranking ..................................................................104

Table 13 ANOVA Means Comparison ...............................................................................106

Table 14 T-Test TF-IDF Analysis ......................................................................................107

Table 15 Private and Federal Job Titles ..............................................................................109

Essay 3: Veteran Talent Within Data Science: An Exploratory Resume Analysis on the

Employability of Active-Duty Veterans

Table 1 Previous Veteran Employability Research ..........................................................141

Table 2 Continuous Variables ...........................................................................................151

Table 3 Categorical Variables ...........................................................................................153

Table 4 Descriptive Statistics Continuous Variables ........................................................154

Table 5 Descriptive Statistics: Categorical Variables .......................................................154

Table 6 Main Model Results .............................................................................................155

Table 7 Alternate Measures Analysis ...............................................................................158

Table 8 Alternate Dictionary Analysis..............................................................................159

Table 9 Categorical Variables ...........................................................................................160

Table 10 Hypotheses Results ..............................................................................................162

Table 11 Top 25 Data Science Technologies from O*Net .................................................164

Table 12 Top Software Related to Resumes .......................................................................165

xii
Table 13 Top Programming Languages Related to Resumes .............................................166

Table 14 Data Science Software Model Results .................................................................168

Table 15 Data Science Programming Language Model Results.........................................170

xiii
 LIST OF FIGURES

Essay 1: Artificial Intelligence in Tactical Human Resource Management: A Systematic

Literature Review

Figure 1 T-HRIS Framework .................................................................................................5

Figure 2 SLR Methodology..................................................................................................22

Essay 2: Transparent Artificial Intelligence and Human Resource Management: A

Systematic Literature Review

Figure 1 Summarized Steps of Word2Vec ...........................................................................74

Figure 2 The JC-Compass Framework .................................................................................83

Figure 3 Adopted Data Science Competencies ....................................................................88

Figure 4a Most_Similar Comparison .....................................................................................95

Figure 4b Most_Similar Comparison .....................................................................................96

Figure 5 Both Sector TF-IDF Comparisons .......................................................................105

Figure 6a Private Sector Job Category TF-IDF Comparison ...............................................110

Figure 6b Federal Sector Job Category TF-IDF Comparison ..............................................110

Figure 7 Summarized Framework ......................................................................................113

Essay 3: Veteran Talent Within Data Science: An Exploratory Resume Analysis on the

Employability of Active-Duty Veterans

Figure 1 2022 Average Unemployment Population Percentage ........................................137

Figure 2 Employability Construct: Modified from Fugate et al. (2004) and Ngoma and

Ntale (2016) .........................................................................................................144

Figure 3 Employability Research Model ...........................................................................148

Figure 4 Data Science Software vs Unemployment ..........................................................166

xiv
Figure 5 Data Science Programming Languages vs Unemployment ................................167

xv
ESSAY I. ARTIFICIAL INTELLIGENCE IN TACTICAL HUMAN RESOURCE

MANAGEMENT: A SYSTEMATIC LITERATURE REVIEW

1
 CHAPTER 1: ABSTRACT

Digitization within Human Resource Management (HRM) has resulted in Artificial

Intelligence (AI) becoming increasingly prevalent in Human Resource Management Systems

(HRMS) and HR Information Systems (HRIS). The tactical procedures of recruitment, employee

performance evaluation and satisfaction, compensation and benefit analysis, best practice

analysis, discipline management, and employee training and development systems have seen a

growth in the incorporation of AI. To better understand this evolution, we seek to explore

publication sources and literature that feature the application of AI within HRM. By utilizing a

systematic literature review methodology, this paper identifies which tactical HRIS (T-HRIS)

components are featured in literature and how each T-HRIS component is represented. This

paper gives insight to which component of tactical HRM/HRIS receives attention and identifies

gaps in research to give direction to future research agendas.

2
 CHAPTER 2: INTRODUCTION

Human Resource Management (HRM) modernization has experienced a grand evolution,

as digitization infiltrates the tedious processes which exist within its respective operations. From

earlier inventions like the computer and the internet, HRM has found a way to navigate these

advancements to electronically increase productivity, cost effectiveness, and market competition

(Hmoud and Várallyai, 2020). Like a trebuchet, advanced technology launched the rapid

evolution of Human Resource Information Systems (HRIS) as newer capabilities like Artificial

Intelligence began to infiltrate tactical practices within HR operations, otherwise known as

tactical HRIS (T-HRIS). The amount of organizational, personnel, and task-orientated data HR is

inherently responsible has led to the incorporation of AI in many tactical HR processes, as it

enhances sustainable business models (Di Vaio et al., 2020). However, this evolution and growth

in capabilities comes with a responsibility of understanding the current state of AI within tactical

HR processes, requiring both HR professionals and academics to dive into existing literature

which highlights AI-enhanced HR capabilities and areas of growth within the HR discipline.

Literature reviews leading up to this paper have provided a foundational understanding of

where Artificial Intelligence exists within HRM and T-HRIS (Di Vaoi et al., 2020; arg et al.,

2021a; Qamar et al., 2021; Vrontis et al., 2021). However, these extensive reviews have failed to

consider how AI applications are utilized in a managerial and technical standpoint, providing

little insight to which components of HRIS are underrepresented with AI capabilities. Identifying

this deficiency, we seek to explore this consideration to provide the academic community and

professional sector insight to where AI is potentially lacking and where it is thriving. The aim of

this research is to explore AI within the HRM and tactical HRIS discipline. Thus, we conduct a

systematic literature review (SLR) to provide a baseline to understand the status of T-HRIS

3
components within literature and how it is represented. This SLR, has the objective of

identifying and understanding the components of tactical HRIS that are represented most in

literature. Within this manuscript, we attempt to bridge this understanding by answering the

following research questions: RQ1) What are the tactical HRIS components which exist in

published literature? RQ2) How have the components of tactical HRIS been represented in

literature?

As a contribution, we seek to inform both the academic community and professional

sector with insights to where AI is potentially lacking and where it is thriving. This work

scientifically reviews existing literature, identifying what has been accomplished by organizing

T-HRIS components on a technical and managerial spectrum. Given the everevolving state of

technology and newer applications coming to fruition within HRM, the implications of this

research are imperative for academics and industry professionals to historically understand the

direction research has taken with regards to AI and T-HRIS. This historical understanding will

provide insight to potential deficiencies and boons which may exist with regards to AI

applications.

The arrangement of this paper is as follows. It first provides a background and framework

to tactical HRIS and HRM, the evolution of AI within HR, and the AI methods acknowledged in

this paper. Secondly, it provides an explanation of the SLR methodology used. Next, it provides

the results that include insight to the which components of T-HRIS exist in literature, how they

are represented, which publication sources described the research, and future research

considerations. Finally, the paper concludes with a discussion of future research opportunities.

4
 CHAPTER 3: BACKGROUND AND RELATED WORK OF T-HRIS AND AI

Within this section, we define what tactical HRIS is comprised of, propose a framework

which outlines the structure of tactical HRIS, identify previous literature reviews which have

been conducted in this field, and define common AI methodologies pertaining to this literature

review.

Figure 1. T-HRIS Framework

Tactical Perspective vs. Strategic

Tactical HRIS (T-HRIS) refers to the Human Resource personnel and technology

responsible for performing specific tasks inherent to the profession of Human Resource

Management to achieve organizational goals and objectives. T-HRIS equips HR leaders with the

technological infrastructure to navigate decisions concerning company best practices, employee

performance and satisfaction, recruitment information systems, compensation and benefit

information systems, employee training and development systems, and employee discipline

management. To understand further, this discipline of HRIS comprises of two separate

5
intellectual assets: managerial HRIS (otherwise known as “soft”, “collaborative” and “people-

centric”) and technical HRIS (otherwise known as “hard”, “calculative” and “data-centric”)

(Collings et al., 2018; Mayfield et al., 2003; Stewart, 2007). Technical HRIS reflects logic,

reasoning, data, and understanding. In contrast, the managerial side, is dedicated to relationship

building, workplace synthesis and creativity, and the care of the employees who work for the

organization (Cregan et al., 2021; Laker and Powell, 2011). To expand upon the structure of T-

HRIS, Figure 1 provides a framework outlining the fundamental components which make up T-

HRIS (Collings et al., 2018; Kumar, 2012).

“Collaborative” Managerial HRIS

Managerial human resource management refers to the unique people-centered

organizational strengths that contribute to making decisions relating to employee skills,

expertise, culture, and commitment (Mayfield et al., 2003; Stewart, 2007). The term

“managerial” is about soft skills, which are actions and interactions with others and how we

communicate ideas. Connecting these “managerial-skills” with HRIS implies technological

capabilities that enable HRM professionals to better connect with employees and effectively

make decisions within their respective companies. At the tactical level, enhancing the

information system components of managerial HRM practices improves HRM capabilities by

streamlining time-consuming tasks and giving time back to employees and customers alike.

Furthermore, managerial HRIS emphasizes the importance of humanizing employee motivation

practices to enhance workplace relationships between the organization and its respective

employees (Cregan et al., 2021). Two components distinguish the collaborative nature of

managerial HRIS. As seen in Figure 1, the first branch of managerial HRIS is the Human

Relations component.

6
 Human Relations

The human relations element of managerial HRIS relates to technologies that enhance an

organization’s ability to generate and maintain professional and effective interpersonal and

intrapersonal relationships. Regarding interpersonal relationships, this is an organization’s ability

to connect and exchange information and ideas between two or more individuals by way of any

channel (in person, online, or written) and build relationships (Collinson, 1996; Laker and

Powell, 2011). These interactions tend to be in developing relationships and developing an

extensive professional understanding on one’s organization and those they regularly

communicate with (Collinson, 1996). Similarly, intrapersonal relationships refer to an

organization’s ability to effectively self-reflect on successes and failures and grow. These

interactions tend to be very self-reflective and perceptive to the environment (Laker and Powell,

2011). The human relations component of T-HRIS provides two components to an

organization’s ability to manage its employees effectively. The first component is Employee

Performance and Satisfaction (EPS), which refers to how an organization and its managers

connect with its employees, understanding employee diagnostics, and retains talent. This pillar

provides insight into how an employee is performing by technologically facilitating performance

evaluations and feedback sessions for managers to vector employee behavior and provide them a

human connection to discuss successes and concerns within the work environment.

The second component is Discipline Management Systems (DMS) which focuses on

employee behavioral rehabilitation and termination processes (Tariq et al., 2016). Regarding

DMS, navigating disciplinary issues within an organization requires an in-depth understanding of

company policy. However, it also requires strong inter/intrapersonal skills to navigate delicate

conversations and potential terminations carefully. Furthermore, connecting with employees and

7
understanding their family demographics to ensure they are correctly being compensated and are

receiving the correct benefits is important to the CBA pillar of T-HRIS. Lastly, providing a

technological infrastructure that effectively and professionally manages the EPS processes

provides both managers and employees more time to communicate and grow within the

organization. We provide Table 1 to further define and explore specific task descriptions of these

“managerial” categories within T-HRIS to give insight and outline expectations of

responsibilities.

Table 1. Tasks within Tactical HRIS “Managerial” Categories

Name Task Descriptions Reference

Employee Provide performance feedback to employees by tracking employee Tong et al (2021)
Performance behavior at work
and Satisfaction Assess Employee Productivity Tong et al (2021)
Automate performance evaluations Tong et al (2021)
Generate Personalized Recommendations for Job Improvement Tong et al (2021)
Gauging Workplace Morale Amer-Yahia et al. (2020);
Garg et al. (2021); Rathi
(2018)
Identifying employees who are at risk for leaving Rathi (2018)
Identifying when an employee applies him/herself physically, Hughes (2019)
cognitively, and emotionally toward their work

Employee Fielding harassment claims Eubanks (2018)

Discipline Regulation of the employment relationship through active Jones and Saundry (2012)
Management intervention in disputes between employers and managers
Systems
Remain neutral and ensure employees are treated fairly Bourhis et al. (2019); Jones
and Saundry (2012)
Design policy and procedures for disciplinary actions Jones and Saundry (2012)
Enforce disciplinary rules consistently Jones and Saundry (2012)
Provide legal guidance to ensure managerial decisions do not lead Jones and Saundry (2012)
to expensive litigation
Provide broad view of organizational implications of disciplinary Jones and Saundry (2012)
actions

8
 “Calculative” Technical HRIS

Contrary to managerial HRIS practices, “technical” HRIS practices historically have a

reputation for reflecting the data-driven and technical skills and capabilities within an

organization (Eubanks, 2018; Laker and Powell, 2011). Managerial HRIS centers itself around

HumanRelation technologies that help facilitate inter/intrapersonal connections within the

organization. Technical HRIS revolves around technologies and information systems that

facilitate data analysis, technical understanding, and efficient workflow. This side of the HRIS

coin contains two branches that capture the responsibilities within an organization: 1) Resource-

Driven capabilities and 2) Policies and Standards of Practice (SOPs).

Whereas managerial HRIS is considered “collaborative” approaches to navigating

challenges within an organization, technical HRIS is seen as “calculative” and direct, using data

to influence decisions and navigate the challenge of maintaining the competitive edge. Specific

categories within technical T-HRIS include customer engagement and workflow Best Practices

within the organization, Recruiting Information Systems for talent management, Employee

Training and Development systems, and Compensation and Benefits Information Systems which

manage an employee’s benefits and pay. To further expand on the responsibilities of each

category, Table 2 summarizes task descriptions of these “technical” categories within T-HRIS.

Resource Driven

The phrase “resource” alludes to a variety of different ingredients that help propel an

organization into success and notoriety. Some examples include allocated budgets, time,

technological equipment, and infrastructure. In HRM lens, acquiring and retaining premier talent

creates a demand for streamlined HRIS processes to assist in hiring decisions and prevent a

9
company from falling short due to the inability to screen, evaluate, interview, and onboard

enough qualified candidates (Ahmed, 2018; Hmoud and Laszlo, 2019).

Furthermore, the pressure of reducing discrimination within the recruiting process and

employee compensation/benefits is imperative to maintaining a notable reputation within the

industry the organization is a part of (Rathi, 2018). These modern challenges have created a

demand signal to bolster these technological capabilities to streamline hiring decision-making

processes when acquiring new talent to grow the company. From the technical HRIS perspective,

acquiring the employee and retaining them comes first. The managerial practices come later after

the asset has been obtained, trained, and put to work.

Table 2. Tasks within Tactical HRIS “Technical” Categories

Name Task Descriptions Reference

Best Practices Reduce Administrative Burden Eubanks (2018)
(Customer
Track and Analyzes interviews (feedback) with Kumar (2012)
Engagement/Workflow)
supervisors and employees
Track and Analyze input and output of employees Amer-Yahia et al. (2020);
Kumar (2012)
Track and Analyze positive action guidelines Kumar (2012)
Recruiting Information Reduce Discrimination within Recruiting Process Rathi (2018)
systems
Receive and organize applications for HRM team Kumar (2012)
Screen, evaluate, select, onboard prospective Kumar (2012)
candidates
Employee Training and Identifying and personalizing employee professional Obeidat (2012)
Development Systems development options
Track employee enrollment of adequate or required Obeidat (2012)
training courses related to job
Track employee enrolment in courses to develop skill Obeidat (2012)
and abilities to carry out new jobs
Compensation and Facilitate easy calculations of salaries and wages of Eubanks (2018)
Benefits Information employees
Systems Ensuring Gender Parity Eubanks (2018)
Provide self-services access to payroll system for Muhammad et al. (2021)
employees
Assist management for salary planning Muhammad et al. (2021)
Eases distribution of benefits to employee from Muhammad et al. (2021)
management

10
 Policies and Standards of Practice (SOPs)

Company workflow (a measure of employee input and relative output), modernized

policies, and logical SOPs create two sub-branches within the second branch of technical HRIS.

The first branch sets its attention to Best Practices that exist within an organization. Specifically,

this data revolves around employee output relative to the input, customer engagement and

satisfaction statistics, policies to reduce administrative burdens, tracking and analyzing employee

performance evaluations, and feedback interviews (Eubanks, 2018; Kumar, 2012; Rathi, 2018).

The second branch of technical HRIS sets its focus on analyzing and interpreting

employee training and development data. This branch signifies an organization’s ability to invest

in human capital and bolster assets that already exist within to strengthen retention and further

develop their skills. Providing employees with training and education to perform new or hone

existing skills can be seen as a return on investment, which this component of HRIS evaluates.

Furthermore, this branch statistically identifies deficiencies and knowledge gaps within an

organization, prompting training managers to engage more effectively and secure training for

those who need it most.

The Evolution of HRIS

The historical review (Bhuiyan et al., 2014) provides an in-depth insight into the

evolution of HRIS. The drastic change in information technology during the 1990s

revolutionized HR professionals’ roles. Many of the hands-on and analog practices within HRM

experienced a great deal of technological evolution (Bhuiyan et al., 2014). Punch cards that log

an employee’s time became digital interfaces that allow employees to use their biometric

information to “clock in” to work. Engaging with one’s manager or supervisor to request paid-

time-off became a simple email rather than a conversation and logging a hand-written calendar.

11
Furthermore, digital job advertisements and applications have become more prevalent in the job

market for both the prospective employee and employer.

With the rise of big data and computing capabilities, HRM and HRIS constantly need to

cope with the amount of data they receive more effectively. The next frontier of digitization

evolution has required incorporating AI within these HRIS functions. HR personnel screening of

prospective applicants and their respective resumes can be done by AI applications that utilize

fuzzy and neuro-fuzzy-based agent approaches. AI can also create a short list of qualified

candidates based on what the HR department is looking for in prospective hires and candidates

(Doctor et al., 2009a; Doctor et al., 2009b).

Reviews of AI within HRM and T-HRIS

There has been recent academic interest in how organizations implement AI within HRM

business practices (Abdeldayem and Aldulaimi, 2020; Collins, Dennehy, Conboy, & Mikalef,

2021; Di Vaio, Palladino, Hassan, & Escobar, 2020; Garg, Kiwelekar, Netak, & Ghodake, 2021;

Vrontis, Christofi, Tarba, Makrides, & Trichina, 2021). We refer to three systematic literature

reviews that have previously investigated AI within HRM (Di Vaio, Palladino, Hassan, &

Escobar, 2020; Garg, Sinha, Kar, & Mani, 2021a; Qamar, Agrawal, Samad, & Chiappetta

Jabbour, 2021; Vrontis, Christofi, Tarba, Makrides, & Trichina, 2021) and highlight how this

work differs from them (see Table 3).

The previous literature reviews that have been conducted regarding AI and HRM

practices, although similar in nature, offer very different methodologies and perspectives to

consider. For instance, the bibliometric analysis conducted by Di Vaio et al. (2020) reviewed 73

academic articles that met their specified criteria. Specifically, Di Vaio et al. (2020) focused their

SLR toward understanding the state of the art of AI and small business models (which included

12
HRM practices) and providing a sense of direction for future research. Whereas, the SLR

conducted by Vrontis et al. (2021) analyzed 45 articles and mapped out research of AI within

HRM by identifying the broad themes involved (humanrobot collaboration, decision-making,

learning opportunities, recruiting, training, job performance, and job replacement). The SLR

conducted by Qamar et al. (2021) provided insight into the state-of-the-art applications of AI

within the HRM domain and sets the stage for a future research agenda. They uniquely

developed a taxonomical overview of the AI applications within HRM upon reviewing 59

articles. Lastly, (Garg, Sinha, Kar, & Mani, 2021a) provided a semi-systematic review of 105

articles and identified a strong use of ML applications in recruiting and performance

management functions within the HRM spectrum and highlighted the need for HR experts and

ML specialists to work together when incorporating the newer AI methodologies within HR

practices.

Table 3. Previous AI and HRM Literature Overviews

Authors Purpose of Literature Years Number of
Included Primary
Studies
Di Vaoi et Comprehensive review of relationship between AI and sustainable 1990-2019 73
al. business models; identify research gaps between knowledge
management systems and AI; implications of AI within sustainable
development goals
Vrontis et Holistic SLR on AI within HRM practices; understand impact of AI unspecified 45
al. on HRM strategies; understand impact of AI on HRM activities
(recruitment and job performance)
Qamar et SLR of AI and HRM to capture current state-of-the-art and prepare "as of July 59
al. for new research agendas 2020"

Garg et al. Semi-systematic literature review; understand current state of ML 2002-2018 105
integration within HRM; showcase relationship between HR
experts and ML specialists
This Study Explore Tactical HRIS literature and come to understand which 2014-2020 33
components are exist in literature and how they are further
represented.

13
 Complementing these three SLRs, this paper studies how AI exists within tactical HRIS

(T-HRIS) and HRM practices. We propose a framework that expands the difference between

technical T-HRIS (resource and data-driven components) and managerial T-HRIS (human-

centric components). Our analysis surveys the last six years of this development (2014-2020),

picking up where Bhuiyan et al.’s (2014) survey on the evolution of HRIS left off. Furthermore,

it uniquely focuses on AI’s application and relationships within tactical components of HRIS and

HRM.

Definitions of Common AI Methods within HRM and T-HRIS

AI uses many analytical methods ( Maettig & Foot, 2020). In this paper, we focus on four

methods that frequent AI and HRM/HRIS literature. Table 4 provides definitions of the four AI

categories.

Machine Learning

“Machine learning” refers to a machine’s ability to learn and perform a process given a

goal and defined steps (tasks) to train off to reach said goal. A generic example of machine

learning within the context of THRIS would be training a machine to recommend an employee

for promotion. The goal in this example would be the promotion recommendation (the task), and

the steps to get there would be the qualifications and criteria weighted against candidates in this

decision (the performance metric). Ideally, the training data would be a list of previously

recommended employees for promotion and their respective records, contributing to the

decision. The machine would then study those tens, hundreds, or thousands of records and

identify patterns to recommend given a set of new data (the stimulus/experience) (Mitchell,

2006).

14
 Machine learning also utilizes deep learning techniques, which utilize neural networks

(NN) to analyze further and accomplish these complicated recommendation tasks. NN can

analyze copious amounts of data quickly and assist humans in recommending business decisions

based on analysis completed via NNs (Khan et al., 2020). As an example, such NNs can find

correlations between real-life events and sentiment changes via thousands of tweets (from

Twitter) and identify topics which influenced the sentiment of the public toward events

(Ridhwan and Hargreaves, 2021).

Natural Language Processing

Natural Language Processing (NLP) refers to a machine’s ability to effectively

communicate with humans in their native tongue. Such capability has enabled machines to better

understand both speech and text and generate relevant responses to the stimulus it receives from

the human.

Such AI practices have become more prevalent when interacting with customers and

assisting employees as chatbots and language analysis algorithms have started introducing

themselves and streamlining various T-HRIS processes like employee onboarding, recruitment,

training, and leave requests (Garg et al., 2021) Majumder and Mondal, 2021). A chatbot can

interact with the individual and answer their questions autonomously (Eubanks, 2018; Majumder

and Mondal, 2021; Shum et al., 2018).

Machine Vision

Machine vision uses cameras and other sensory devices to provide machines with the

ability to process visual data to better understand the world around them and make autonomous

decisions based on the received data. Examples of such capabilities are evident in the

development and research of self-driving vehicles, medical body-scan technologies in search of

15
cancerous abnormalities at the cellular level, facial recognition systems, and the US Post Office’s

ability to sort letters containing handwritten addresses (Mitchell, 2006; Rudin, 2019).

Within the context of T-HRIS, machine vision capabilities have started to integrate

themselves within recruitment processes and prescreening interviews. Specifically, the

development of a machine “Looking at People”(LAP)is underway. This investigates AI

algorithms that review video data to assist in automating the recruitment process by providing

HR professions a first impression, personality analysis, and recommending the pursuit of the

candidate (Escalante et al., 2017).

Recommendation Engines

Recommendation engines are AI tools that have increasingly infiltrated everyday life in

activities ranging from online shopping to find the right playlist of music or podcast to engage.

These engines allow companies like Amazon, Facebook, and Netflix to offer personalized

experiences to their customers (Xiao and Benbasat, 2007). These unique AI engines highlight

content or products that are more likely to appeal to the customers or employees of the company

based on behavioral trending data collected on them. The engines also analyze demographic

information on the individuals to help connect people via social media websites (“you may also

know” functions).

T-HRIS has also seen a growth in these capabilities via candidate recommendations,

promotion recommendations, and employee training and development. Recommendation engines

not only streamline what once hours of individualized analysis were, but they have empowered

HRM professionals with the ability to process copious amounts of data and identify trends that

are outside of human capabilities to the computer. These recommendation engines are in action

in a variety of publicly accessible applications. For instance, LinkedIn, Instagram, and Facebook

16
are well known for connecting colleagues with one another. LinkedIn serves as a more

“professional” social media outlet (think virtual curriculum vitae), Instagram and Facebook cater

to the more personal aspects and hobby-based activities in one’s life (Papacharissi, 2009;

Stapleton et al., 2017).

In summary, this section has provided insight and definitions to tactical HRIS

components and showcased a framework which highlights their placement within the structure.

We have further explored previous literature reviews which have been accomplished regarding

AI applications within HRM and highlighted their purposes, primary citations, and date-range

which they cover. Lastly, we have defined AI methodologies with HRM and applicable uses of

AI within HRM practices.

Table 4. AI Methods
Name
Machine Learning
Natural Language Processing
AI Descriptions
The goal is defined and the steps to reach said goal are learned by
a machine (training). Uses algorithms and/or deep learning
techniques (neural networks) to learn autonomously from data
provided and assists by making decisions and providing insights
(Khan et al, 2020; Kumar 2019; Ridhwan and Hargreaves, 2021)
Automatic manipulation of natural language such as speech and
text. Aims to understand input from human speech as well as
generating responses in human languages (Kumar 2019)

Machine Vision Machine analyses visual information using a camera, analog-to-

digital conversation, and digital signal processing. Help
computers understand content and context of the world
surrounding them. (Kumar 2019)

Recommendation Engines A tool which provides a personalized recommendation to the user

by identifying the right product or content relative to the
interactions via any digital channel (Xiao and Benbasat, 2007.)

17
 CHAPTER 4: METHODOLOGY

This chapter provides a description of the systematic literature review (SLR) process

utilized within this study ((Collins, Dennehy, Conboy, & Mikalef, 2021; Di Vaio, Palladino,

Hassan, & Escobar, 2020; Vrontis, Christofi, Tarba, Makrides, & Trichina, 2021). Using the

SLR methodology we: 1) generate copious amounts of literature to analyze; 2) seek to answer

specific research questions; 3) seek to extract relevant pieces of academic literature relating

systematically to AI, HRM, and tactical HRIS.

Review Process

Following Collins et al. (2020) and Okoli (2015), we conduct this review in 2 phases:

The first phase of this process filters literature, while the second phase extensively focuses on the

content of each article. The goal of this SLR is to understand the representation of T-HRIS in

published literature and how the representation of specific components differs. RQ1 seeks to gain

a complete understanding of which components of T-HRIS exist in published literature. RQ2

seeks to understand how the components of T-HRIS have been represented in literature by

identifying the analysis of the paper (empirical analysis, qualitative analysis, conceptual analysis,

or literature review). Through this SLR, we seek to provide the research community with a

review that investigates AI within HRM and T-HRIS. Considering AI and HRM are

multidisciplinary (Vrontis et al., 2021), we decided to utilize the following databases:

1) Business Source Complete (EBSCO) (Nolan and Garavan, 2016; Vrontis et al., 2021)

2) AIS eLibrary (AIS) (Collins et al., 2021;)

3) Web of Science (WOS) (Collins, Dennehy, Conboy, & Mikalef, 2021; Di Vaio,

Palladino, Hassan, & Escobar, 2020; Pisani, 2009)

4) ABI-INFORM (ABI) (Baskerville & Myers, 2009; Nolan & Garavan, 2016).

18
 Upon identifying our databases, we established our inclusion and exclusion criteria for

Phase 1 (Journal Demographic Filtration) and Phase 2 (Content Filtration). For Phase 1, there are

two filtration steps: 1) the initial search and 2) the journal verification. Our inclusion and

exclusion criteria consisted of 9 factors (5 for inclusion, 4 for exclusion).

We identified the following as necessary components to include a piece of literature for

the initial search (step 1, Phase 1). The article must be: 1) peer-reviewed; 2) written in English;

3) published between 20142020; 4) an academic article or conference proceeding (Collins et al.,

2021; Dhamija and Bag, 2020; Di Vaio et al., 2020; Nolan and Garavan, 2016; Qamar et al.,

2021; Scandura and Williams, 2000). The exclusion of literature occurred if it: 1) was not written

in English, 2) was not peer-reviewed or was non-academic, 3) was an editorial or simulation

piece.

For the journal verification (step 2, Phase 1), we identified 90 academic journals

(Appendix A from a wide range of disciplines that have historically published academic articles

and conference proceedings relating to HRM and AI (Collins, Dennehy, Conboy, & Mikalef,

2021; Di Vaio, Palladino, Hassan, & Escobar, 2020; Pisani, 2009; Qamar, Agrawal, Samad, &

Chiappetta Jabbour, 2021; Vrontis, Christofi, Tarba, Makrides, & Trichina, 2021). These

publication sources serve as inclusion criteria #5 and exclusion criteria #4. We included the

article if it belonged to the 90 journals or excluded it if it did not belong to any of those journals.

Once the journal verification state is complete, Phase 2 of the methodology beings: The

Content Filtration phase. The inclusion and exclusion criteria of this phase consist of two steps.

The first step is a title, abstract, and keywords (if provided) review, where we reviewed the title,

abstract and associated keywords of the article to see if it matched our pre-established definitions

of T-HRIS (Table 1 and Table 2) and AI Methods (Table 4). We also deleted duplicate articles

19
within step 1 of Phase 2. The second step of Phase 2 consists of reviewing the article in its

entirety to verify it met the requirement to be an article whose focus is on AI and T-HRIS

components of HRM.

Table 5. AI and HRM Key Word Searches

Topic
Artificial Intelligence
Human Resource Management
String
("AI" OR “artificial intelligence” OR "Natural Language Processing"
OR "NLP" OR "Chatbot" OR "Machine Vision" OR "Machine
Learning" OR "ML" OR “Recommendation Engine“ OR “Deep
Learning” OR “Neural Networks”)
(“Human Resource Management” OR “HRM” OR “Human Resource
Information Systems” OR “HRIS” OR “Human Resources” OR “HR”
OR “Human Resource Management Systems” OR “HRMS”)

Full String
("AI" OR “artificial intelligence” OR "Natural Language Processing"
OR "NLP" OR "Chatbot" OR "Machine Vision" OR "Machine
Learning" OR "ML" OR “Recommendation Engine“ OR “Deep
Learning” OR “Neural Networks”) AND (“Human Resource
Management” OR “HRM” OR “Human Resource Information
Systems” OR “HRIS” OR “Human Resources” OR “HR” OR “Human
Resource Management Systems” OR “HRMS”)

Once the inclusion and exclusions criteria became concrete for Phase 1 and Phase 2, we

then sought to standardize our search string. One general search string strategy is to base the

string on specified research questions and a list of synonyms (Collins et al., 2021; Kitchenham,

2012). We pulled from our research questions and pre-defined AI methods (see Table 4). Our

search string consists of two pieces. The first piece is our key phrases regarding AI, while our

second piece comprised keywords relating to HRM, HRIS, and HRMS. We used a Boolean

practice when developing our search string. The "OR" operator exists between different words.

The "AND" operator’s purpose is to connect keywords. The use of parentheses

compartmentalized the AI-specific keywords from the HRspecific terms. Table 5 outlines the

search string used on all databases chosen.

20
 Methodology in Action

Figure 2 demonstrates the selection process of our literature. Plugging the standardized

string into each database, cumulatively yielded 315,053 articles. Isolating the articles which met

the date, language, and peer-review requirements, achieved a total of 36,856 articles.

Subsequently we filtered based on the inclusion and exclusion criteria previously discussed.

Following this step, we further sorted through these articles based on the academic journals to

which they belonged, yielding a tertiary selection of 697 articles from the four publication

sources (EBSCO: 29; AIS: 246; WOS: 79; ABI: 343).

The tertiary selections from Phase 1 feed Phase 2 of our methodology. From this number

(697), we look at the content of each article selected. For step 1 of phase 2, we reviewed each

article’s title, abstract, and keywords based on definitions established in section 2 (Table 1, Table

2, and Table 4). Phase 2 was our initial screening to validate which of the 697 articles truly

connected AI with T-HRIS within HRM. We also took this time to delete duplicate articles from

the databases to secure a list of unique articles. Upon completing step 1 of Phase 2, we yielded a

secondary selection of 74 unique articles. These 74 articles fed the last and final step of our

method, which required us to read through the entirety of each article and validate that they were

genuinely related to AI practices within T-HRIS components of HRM. Of these 74 articles, 33

met the criteria to be the primary studies selected based on our SLR.

21
Figure 2. SLR Methodology

22
 CHAPTER 5: FINDINGS

The findings and analysis of literature within AI and HRIS are presented in this chapter.

Managerial T-HRIS Results

Managerial T-HRIS provides an avenue to focus on the more humancentric

responsibilities HR professionals are responsible for managing and nurturing. Figure 3

showcases three unique branches which fall within this managerial T-HRIS literature umbrella.

The first branch of this consideration is employee performance and satisfaction, and the second

is discipline management systems. Understanding employee performance, satisfaction, and

discipline within an organization are imperative to securing retention and potentially

understanding where weaknesses exist. Given that the nature of delivering feedback is sensitive

explore AI-applications to streamline these managerial responsibilities provides HR

professionals with more time to understand and interpret the analysis provided by the machine

and carefully deliver the feedback and findings to the employee in question. Furthermore, the

exploration of AI within compensation and benefits systems provides HR professionals with a

tool to ensure their employees are engaged and receiving fair and correct entitlements based on

their unique circumstances (Hughes et al., 2019).

AI Literature within Employee Performance and Satisfaction

Pratt et al. (2021) investigate how technological-based communication tools can

autonomously engage with employees and streamline processes via simulating human-to-human

interactions. Their proposed model provides insight into an employee’s satisfaction and

performance concerning cultural and motivational factors, communication techniques, and the

work descriptions of each employee. The presented model demonstrates essential factors to

consider in AI-based communication tools and how they interact with one another. Where

23
faceto-face techniques excelled indirectly in motivating employees and increasing their

satisfaction, the AI-enhanced communication tool experienced more of a challenge when

motivating and directly influencing an employee’s satisfaction. Thus, the tactical act of

communicating with employees and making that connection presented a challenge for AI-

enhanced communication tools within (Pratt et al., 2021) research.

Similarly, (Tong et al., 2021) research explores AI’s theoretical application in employee

performance feedback. It empirically evaluates the quality of feedback (which may increase

employee productivity) and the employee perception of the AI from a field experiment. Given

AI’s superior standardized and analytic nature, evaluating employee input and output, providing

areas of improvement, and predicting future performance has become a streamlined process for

HR professionals and T-HRIS applications. HR teams would manually review and scrub an

employee’s record and identify positive/negative trends. An AI-enhanced T-HRIS application

can do the analytics for them and provide recommendations directly to the employee. Despite the

brutally honest system, AI feedback can negatively affect the employees by disclosing their

weaknesses.

Consequently, direct honesty discourages them from furthering (otherwise known as the

"adverse disclosure effect"). This effect creates a unique paradox when considering improving an

employee’s motivation, satisfaction, and performance: Do not communicate the weaknesses

directly and allow them to achieve complacency or disclose the employee’s areas of

improvement, potentially discouraging them further to seek improvement and achieve the same

complacency. Ultimately, (Tong et al., 2021) experiment and research conclude that these

considerations coexist, and the mitigation of adverse disclosure effects is dependent on an

employee’s tenure within a firm.

24
 Both (Pratt et al., 2021) and (Tong et al., 2021) highlight one of the most significant

weaknesses AI has within the managerial T-HRIS applications, lacking emotional intelligence.

Within employee performance and satisfaction, an element of emotional intelligence requires

effective and meaningful communication with employees. Managers and HR professionals need

to have the emotional wherewithal to cope with difficult emotional hurdles that may present

themselves. However, they also need to have the ability to communicate effectively while

respecting their employees’ boundaries. Prentice et al. (2020) explore this notion by empirically

studying the effects emotional intelligence and AI has on employee retention, satisfaction, and

performance within the hotel industry. The findings within their study conclude that emotional

intelligence is a valid predictor of employee attitudes and behaviors, and that AI does well to

focus on the technical and functional efficiency an employee has on organizational performance.

Despite having a significant impact on employee performance, AI holds little significance about

employee retention, further expanding upon the importance of establishing interpersonal

connections with employees.

Finally, (Garg et al., 2021) provide insight to employee satisfaction within the logistics

and freight forwarding organizations by developing a novel approach to analyzing employee

feedback and satisfaction using AI algorithms. This further empowers an employee to answer a

climate survey on their experiences, analyzes the respondent’s input, and provides the

organization actionable insight which allows organization stakeholders more insight to how to

improve employee engagement, retention, and efficiency. Through this analysis, (Garg et al.,

2021) provide both the professional and academic community insight on how the use of AI has a

profound effect on the organization and its respective employees. These capabilities benefit the

25
employee’s voice and provides an avenue to communicate concerns and boons directly to the

organization.

AI Literature from SLR regarding Employee Performance and Satisfaction

The SLR we undertook yielded 4 total articles which featured Employee Performance

and Satisfaction. These 4 articles were consistent of 2 qualitative paper and 2 literature reviews.

The qualitative analyses conducted gives insight to individualization of HRM practices and

AImediated social exchanges and how AI impacts HRM practices and the attitudes and

behaviors of employees. Their collective findings via interviews suggest generational differences

in the adoption of AI and that AI-enabled bots and digital personal assistants are utilized in

analytical and routine tasks involving employees (Kaminska and Borzillo, 2018; Malik et al.,

2020). Within Malik et al.’s (2020) analysis, is suggested that these AI-enabled programs bolster

HR cost-effectiveness and enhance the overall employee experiences, thereby increasing

employee commitment and satisfaction within an organization. While, Kaminska and Borzillo

(2018) suggest fundamental differences between Genration X and Y employees and the adoption

of enterprise social networking systems driven by AI. The 2 literature reviews discovered within

this SLR provide insight to how AI technologies have infiltrated organizational settings by

influencing personnel development. Specifically, Kotera et al. (2019) provide a meta-analysis on

natural language processing AI technologies and how it affects the feelings and thinking of

employees within the work environment. While Kock et al. (2020) systematically review AI

capabilities within employee performance reviews and how cognitive factors relating to rating

quality and personalityrelated factors differ within HRM practices.

26
 AI Literature within Discipline Management Systems

The research conducted by Dressel and Farid (2018) explores a new light of discipline

management by evaluating the predictive analysis of recidivism within the criminal justice

system. With the exponential growth of data and the development of AI-empowered machines to

cope with this growth, commercial software runs the risk of developing inaccurate predictions

that could affect human beings’ lives. Specifically, AI-algorithmic developments have fueled

tools like the Correctional Offender Management Profiling for Alternative Sanctions

(COMPAS), which predicts a defendant’s risk of committing a misdemeanor or felony within

two years of assessment from 137 different features about the individual and their respective

criminal history. (Escalante et al., 2017)’s analysis compared the overall accuracy and bias in

human assessments (individuals with no criminal justice background) with the algorithmic

assessment of COMPAS. Through (Dressel and Farid, 2018) ’s empirical analysis, the prediction

of the COMPAS algorithm was no more accurate or fair than the predictions of humans with

little to no criminal justice experience. The results were near indistinguishable, thus showcasing

the dangers of blindly implementing AI practices in high-risk decisions where human lives are

on the line.

Formal grievance processes have also experienced an AI facelift, providing employees

with a discrete avenue to communicate and report abhorrent behaviors within an organization,

minimizing the fear of retribution and feelings of shame. Olson (2018) highlights Spot’s online

chatbot, which engages with the harassment victim on their experiences. This chatbot utilizes

NLP techniques to provide a "cognitive" interview to the employee and gain insight discretely on

details about their harassment claim. As Spot has grown within the company, it has developed

analytic capabilities to gauge organizational diagnostics regarding the health of the

27
organization’s corporate cultures. This development has provided insight into where reports are

deriving from and identifying patterns that HR professionals cannot humanly conceive through

pattern recognition. Although it may not solve the problems directly, (Escalante et al., 2017)

concludes Spot gives insight into more significant issues that employees are not willing to speak

about aloud.

AI Literature from SLR regarding Discipline Management Systems

The qualitative analysis conducted by Bhattacharyya and Nair (2019) discusses how

future of work applications (robotics, artificial intelligence, internet of things) have multiple

facets via semistructured open-ended interviews with 26 respondents. Their analysis gives

insight to how organizations will have data dependencies and that employee will be expected to

synthesize data for sense making and decision processes which could affect work performance

and disciplinary actions (Bhattacharyya and Nair, 2019).

Technical T-HRIS Results

AI applications within Technical T-HRIS have enabled once timeconsuming and task-

saturated processes to be automated and streamlined for both the employee and customer. These

tasks are data-driven to ensure the organization is optimized and can maintain or obtain a

competitive edge. The first branch of Technical T-HRIS we will review concerning AI is the

“Best Practices”branch.Four prominent pieces of literature explore how AI has influenced

workflow and customer engagement within the Best Practices branch of T-HRIS. The second

branch explored will be the Recruitment Information Systems component which contains four

separate pieces of literature which review the effects and consequences of implementing AI

within these processes. The third branch is the Employee and Training Development Systems

branch. This branch contains three pieces of literature that explore the effects of AI within

28
employee education systems. Compensation and Benefits Analysis is the last branch, which

focuses primarily on securing employee benefits and financial considerations relative to

organization resources and employee qualifications. The exploration AI within compensation

and benefits systems provides HR professionals with a tool to ensure their employees receive the

correct entitlements based on their unique circumstances while effectively budgeting company

expenses.

AI Literature within Organizational Best Practices

Cain et al. (2019) provide a compilation and systematic review of current studies and

journals which examine the impact of AI and robotics have within the hospitality and tourism

industry. Their review proposes emphasizing the importance of exploring research within the

human-robot interaction as these technologies grow to help streamline processes to give time

back to employees. Furthermore, they insist that research should focus on understanding how

customers interact with automated AIs capable of performing an employee’s tasks quickly.

Rahmani and Kamberaj (2021) review how introducing AIenhanced chatbots that interact

with customers and employees provides organizations with machines capable of learning and

automatically adapting to the environment based on data received and processed. Their analysis

provided insight into how the automation of administrative tasks within employee onboarding,

process improvement, and recruitment provide humans with the necessary information, time, and

psychological energy to make well-informed decisions for their organization. Their paper

assesses how an organization can use chatbots to streamline administration and explores what

technology they can run. It discusses how AI-enhanced chatbots can simulate comparably

“human” conversations using NLP and machine learning processes to enhance the customer

experience.

29
 Further expanding on the use of Chatbots, Majumder and Mondal (2021) provides an in-

depth overview of the use of chatbots within HRM and the streamlined processes. The influence

chatbots have within decision-making processes is prevalent, but they also facilitate a better

understanding of AI and ML innovations among the employees within an organization. As an

example, Majumder and Mondal (2021) highlight the innovative application of an AI chatbot

within employee training modules, making them more interactive and engaging directly with the

employee, compared to the mandatory videos these same employees once had to watch.

Lastly, Chakraborty and Kar (2021) provide an example on how AI can be used to mine

data to gain insight on the wellbeing of employees during the unprecedented times following the

COVID-19 pandemic. Specifically, they conducted a mixed methodology analysis which utilized

AI technologies to mine from Twitter in addition to conducting interviews to identify challenges

COVID-19 created for employees across the work spectrum. Specifically, they addressed how

professionals across industries are affected by the pandemic, how that impact translates into

academia, and the nature of the impact on the social welfare of faculty members. Ultimately,

their analysis concluded there are systemic challenges revolving around infrastructure and digital

readiness, workforce demand and supply, and job losses.

These innovations spotlight the benefits of having a machine engage directly with the

employee to answer questions they may have regarding pay, benefits, or any other HR-related

topic. Not only does this free up time for the HR professional, but it also provides a near-human

experience for the employee to directly engage with an HR entity to get the information they

were looking for more quickly.

30
 AI Literature from SLR regarding Organizational Best Practices

Through our SLR, organizational best practices saw an influx of research within AI

capabilities as it was the most represented T-HRIS component with 11 empirical articles, 2

conceptual research articles, and 1 qualitative research article. Regarding the empirical analyses,

predictive AI methodologies to enhance organization performance resonated greatly in

organizational best practices (Bani-Hani and Khasawneh, 2019; Chang and Jung., 2017; Cheng-

Kui et al., 2020; Fehrenbacher, 2017; Jabr and Zheng, 2014; Saha et al., 2016). Furthermore,

these empirical analyses provide insight to how AI technologies can replicate and influence HR

decision-processes within organizations, highlighting both the challenges and benefits that come

with the responsibility of leveraging these autonomous systems (Bani-Hani and Khasawneh,

2019; Castillo et al., 2018; Chang and Jung., 2017; Cheng-Kui et al., 2020; Fehrenbacher, 2017;

Jabr and Zheng, 2014; Kretzer and Maedche, 2018; Lankton et al., 2015; Lee and Ahn, 2020;

Rybinski and Tsay, 2018).

The conceptual research articles identified within this SLR regarding best practices

provide insight to how to cope and navigate computing systems which take on human-like

abilities to make decisions within its environment while providing transparency and a sense of

intuitive intelligence (Huang and Rust, 2018; Scheutz and Venkastesh, 2020). These conceptual

articles provide insight to how an organization and its employees can better interface with AI

technologies to enhance workplace procedures, services, and interactions (Huang and Rust,

2018; Scheutz and Venkastesh, 2020).

Finally, one qualitative research article regarding best practices was identified within our

SLR. Yorks et al. (2020) interview 7 doctors and nurses from a variety of medical fields to gain

insight to AI adoption within the workplace and how it has transformed processes. Their

31
interview findings provided insight to how certain professional demographics (healthcare

workers) are reacting to the technological evolution within their respective workplaces.

AI Literature within Recruitment Information Systems

Johnson et al. (2020) explore implementing electronic HRM and AI to help recruit highly

qualified employees, increase individual retention rates, and decrease the amount of time it takes

to onboard/replace new employees within the hospitality and tourism industry. Their research

proposes that AI facilitates two cognitive elements that promote decision-making within an

organization.

The first introduced is a cognitive insight which implies the algorithms and ML

techniques AI utilizes assists in interpreting the copious amount of data received and discovers

patterns not previously identified by the organization. Such advancements have enabled

organizations to engage in activities more effectively like predicting potential candidates for

hiring and potential internal hires within the organization. The second cognitive element Noone

and Coulter (2012) address is cognitive engagement, which implies using a chatbot to simulate a

human-like and near-natural social interaction with the employee and customer through NLP

technologies. These intelligent agents help with customer decisions and employee questions

within the “tech-support” realm giving time back to HR professionals at the tactical level (Noone

and Coulter, 2012).

Given the recent global pandemic, Koch et al. (2021) utilize advanced web mining

robotics to evaluate how companies in Germany adjusted to the new processes caused by

COVID-19. Given the increased emphasis on public safety, HR professionals heavily relied on

technological capabilities to continue the necessary hiring and recruitment processes.

Specifically, these authors explored how the German public sector job market reacted to the

32
pandemic and how this pandemic affected IT professionals with regards to electronic recruitment

systems (job listings, interview processes, telework capabilities). By utilizing intelligent robots to

scrape data necessary for their research, Koch et al. (2021) were capable of conjuring results

which indicate the importance of advertising work from home availability, increased planning

uncertainty (longer job postings), and that job advertisements with a healthcare background grew

less strongly than they expected.

To expand on the utilization of intelligent agents and chatbots, Rahmani and Kamberaj

(2021) further develop the research by assessing how a company can effectively use chatbots, the

technology on which chatbots run, and evaluating how well a chatbot can simulate a

conversation as human as possible. Their academic exploration showcases a variety of different

types of chatbots and their respective architectures. This in-depth exploration provides HRM

professionals with insight into what chatbots exist. However, it also provides insight into how

they could be utilized in a professional environment to alleviate administrative burden. Upon

interviewing a director from an IT company and an online survey of 210 respondents, (Stewart,

2007) could determine the following: 1) The implementation of AI in automated tasks has

proven to be beneficial for employees and customers. 2) AI should not be responsible for the

entire procedure of hiring new employees. 3) Empathy and emotional bias should play a role in

the employee recruitment process, where AI can help with eliminating bias the human-factor is

still important.

AI Literature from SLR regarding Recruitment Information Systems

Within the SLR we conducted, 8 empirical, 3 conceptual, and 1 qualitative articles were

discovered regarding Recruitment Information Systems. Specifically, the empirical articles

provided thorough analyses on how data-mining techniques on unemployment rates can

33
influence recruitment systems (Li et al., 2014) and how well AI systems can enhance the

candidate selection process within company recruitment initiatives (Karatop et al., 2015;

Martinez-Gil et al., 2020; Pessach et al., 2020; Sajjadiani et al., 2019; van Esch et al., 2019). The

3 conceptual papers expanded on how AI can impact how a company brands and markets itself

to potential recruits (Dabirian et al., 2017). Additionally, these conceptual papers provide insight

to when recruiters should incorporate AI within practices (Black and van Esch, 2020) and how to

use it to develop models which can predict company turnover via case-based reasoning (Wang et

al., 2017). Finally, the qualitative research article discovered within this SLR regarding

Recruitment Information Systems provides insight to how customizable, technology-based

services connect organizational learning and recruitment via interviewing 12 soft-ware firms

located within Norway. These interviews highlighted a tension associated with the need to create

stable individual knowledge systems for employees and dependencies on external software

capabilities to accumulate knowledge (Jøranli, 2018).

AI Literature within Employee Training and Development Systems

Maintaining an employee’s professional development and growth within an organization

is crucial in securing various organizational goals. For one, ensuring employee participation in

annual security compliance and company-specific module training is a responsibility that are

levied upon HR personnel and departments writ large. Mandatory training compliance aside,

grooming employees by presenting educational opportunities eligible for postures for internal

hires and retention benefits. Understanding that the number of employees within an organization

varies, T-HRIS processes are ripe opportunities for AI to grow and streamline educational

processes. For example, a machine’s ability to recommend specific training to an employee

based on an analysis of their aptitude, interests, and success potential is neigh. These capabilities

34
empower HR professionals to explore an employee’s potential further and not be burdened with

the hours of analysis it would take to run this analysis on each employee.

To further expand, Noone and Coulter (2012) evaluate Zaxby’s, a popular fast-food

restaurant originating from Athens, GA, and how they utilize autonomous robotic applications to

reduce service times and food waste substantially. They further explore how these same AI

applications can increase labor training efficiencies and opportunities for enhanced process

management and decision-making through demand prediction and production management AI

applications. Where information had been passed manually from person to person, quick service

restaurants, like Zaxby’s, have successfully applied AI-enhanced tools which store production

knowledge to enhance continuity of best practices to train future employees upon arrival.

Maettig and Foot (2020) further explore improving technical THRIS concerning

employee training by reviewing industrial applications using intelligent augmentation and

human-in-the-loop strategies. Like Rahmani and Kamberaj, 2021, Maettig and Foot (2020)

acknowledge that digital training assistants are great for storing best practices from older

employees to train newer hires. They further acknowledge the value in having digital assistants

analyze performance trends; however, they address that AI-enhanced digital assistants lack the

detailed experience knowledge that older employees may have. Particularly, the implicit

knowledge of experienced employees cannot be easily replicated on digital platforms.

Similarly, Xu and Xiao (2020) introduce the concept of using AIenhanced virtual reality

simulators to enhance mandatory employee training, increasing employee participation by 79%.

Through their empirical analysis, these authors highlighted the application of virtual reality

technology within enterprise HRM, which can improve efficiency, reduce costs, and enhance the

competitiveness of organizations across their respective industries.

35
 Thus, these AI-enhanced training assistants are best used as augmentation devices to

streamline the development of employees. Methods of improvement include tracking,

recommending, and analyzing training, allowing humans to play a prominent role in filling in the

gaps, and adding a tailored and personalized approach to the onboarding process.

AI Literature from SLR regarding Employee Training and Development Systems

Regarding Employee Training and Development Systems, our SLR provided two articles

which expanded upon the incorporation of AI. The first article is empirical and conceptual in

nature. Through interviews, Maity (2019) discovered 33.33% of respondents believed intuitive e-

learning interfaces would be beneficial within their workplace. Maity (2019) further concludes

that 92.6% of HR training professionals believe that AI software within the digital learning

environment should interact with employees to further their engagement and development within

the company. The second article is conceptual in nature, as Lima (2020) explores the smart

organizations and how companies can start to build smarter learning platforms to increase

performance levels and help innovative organizations develop talented, creative, and diverse

employees.

AI Literature within Employee Compensation and Benefits Systems

The advancements of AI within resource-driven T-HRIS are also within the scope of

employee compensation and benefits. In particular, Robert et al. (2020) explore how

organizations have rapidly deployed AI systems to manage their employees and provide three

solutions to address AI unfairness within decision-making processes. Establishing the

operationalization of an organization’s management practices is crucial before delegating

specific tasks to AI technologies. Regarding compensation and benefits, this pushes

organizations to move beyond vague statements of what is considered fair to more specified

36
metrics to secure equal practices and train the AI. Should an AI be used to determine an

employee’s compensation or benefits package, the organization should determine whether

employees work similar hours in the same position and receive equal pay regardless of specific

demographics relating to the employee (sex, age, race).

Ahmed (2018) furthers this discussion by highlighting specific HR functions which AI

can permeate and streamline. Data analytics relating to pay equity is crucial in securing fair and

ethical compensation and benefits practices within an HR department. Not only does an HR

department need to balance the needs of the organization’s employees, but it also must secure

budgeting expenses appropriately. Allowing an AI to undertake the burden of tediously number-

crunching and inferring the unfathomable amount of information collected on its employees

enables HR professionals with the time to review the analysis and validate fair practices.

(Ahmed, 2018) provides insight into how crucial interpretability and transparency of the AI

mechanism is, as it provides HR professionals with an insight into the justification the AI

produces (Bourhis et al., 2019). Thus, this provides HR departments a more trustworthy process

(Hmoud and Várallyai, 2020). The notion of using black-boxes within high-stakes decision-

making, especially within HR departments, is unquestionably discouraged, as the use of “black-

box”modelingtechniques allows for potential unforeseen biases to develop, resulting potentially

in unethical practices within an organization (Ahmed, 2018; Amer-Yahia et al., 2020; Rudin,

2019).

AI Literature from SLR regarding Employee Compensation and Benefits Systems

Within this SLR, there were no articles regarding Employee Compensation and Benefits

Systems discovered within the 33 articles that met our selection criteria.

37
 How have the components of AI and T-HRIS been represented in the literature?

This research aimed to identify how the components of T-HRIS are represented within

the literature. Upon reviewing the methodologies of these articles, we identified that 19 were

empirical analysis, 5 were qualitative analysis, 7 were conceptual analysis, and 2 were literature

reviews. Based on this SLR, empirical studies dominated the results of our SLR, as our technical

T-HRIS components dominated the total count number of articles.

Through our extensive review and using the definitions outlined in Table 2, we identified

which AI methodologies and applications are more represented within HRM and T-HRIS

literature (Table 6). Regarding the AI-type (y-axis), the variable titled “General AI Reference”

represents an article that referenced AI in a general sense and did not specify an AI method

within their paper. Using the definitions our analysis concluded a high representation of machine

learning techniques and general AI reference. We further identified a substantially lower

representation of machine vision methodologies within T-HRIS and HRM applications.

Table 6. AI Representation within SLR

Recommendation Engines

Natural Language Processing

Machine Vision

Machine Learning

General AI Reference

0 2 4 6 8 10 12 14

38
 Component Representation within each Analysis-Type

Of the 19 empirical analyses, 5 qualitative analyses, 7 conceptual analyses, and 2

literature reviews identified, Best Practices and Recruitment Information Systems and Employee

Training and Development were the most represented component of T-HRIS regarding AI

application research. The managerial components, Employee Performance and Satisfaction and

Discipline Management, did not have representation with regards to empirical analysis.

Furthermore, Compensation and Benefit Information Systems component of technical T-HRIS

was only referenced passively and in general terms of AI applications in 8 articles within this

SLR; there were no articles identified within this SLR which primarily focused on Compensation

and Benefit Information Systems and AI applications (see Figure 3). Given that technical T-

HRIS components are resource-driven and based on policies and SOPs, we have identified

through our SLR that empirical studies are currently more represented by technical T-HRIS

components rather than managerial THRIS.

Table 7. Analysis Representation

12
10
8
6
4
2
0
Best Practices Recruitment Employee Training Compensation and Employee Discipline
(Technical) Information and Development Benefit Information Performance and Management
Systems (Technical) Systems Satisfaction Systems
(Technical) (Technical) (Managerial) (Managerial)

Empirical Analyses Qualitative Analyses Conceptual Analyses

Literature Reviews General AI Reference

39
 Specifically, the Policies and Standards of Practice components of technical T-HRIS are

most represented by empirical research. Within the empirical analyses from our SLR, we further

identified common AI themes which these articles highlighted (see Figure 4). This provides

insight to the lack of empirical studies and analyses within managerial T-HRIS components. For

the components which were represented (Best Practices, Recruitment Information Systems, and

Employee Training and Development), our SLR identified the underrepresentation of machine

vision applications within HRM and T-HRIS, while the other methodologies (Recommendation

Engines, Natural Language Processing, Machine Learning, and AI Writ Large) are better

represented within research.

Table 8. Empirical Analysis Breakdown

9
8
7
6
5
4
3
2
1
0
General AI Machine Learning Machine Vision Natural Language Recommendation Not Reprsented
Reference Processing Engines

Best Practices (Technical) Recrutiment Information Systems (Technical)

Employee Trianing and Development (Technical) Compensation and Benefits Systems (Technical)
Employee Performance and Satisfaction (Managerial) Discipline Management Systems (Managerial)

Given the context of this SLR, we have identified a potential avenue to expand upon

empirical academic research within T-HRIS. Within our rigorous SLR, no managerial T-HRIS

components were represented within the group of empirical articles, giving reason to believe

future research could be done in this field to expand the T-HRIS research agenda.

40
 Of the four qualitative studies, our SLR identified, four T-HRIS components were

represented (Best Practices, Recruitment Information Systems, Employee Performance and

Satisfaction, and Discipline Management Systems). Each component contained one piece of

literature which focused on an A1 methodology. Comparatively, the qualitative analysis article is

the second least represented article type within this SLR. Nevertheless, it was the only article

type to contain equal representation for managerial and technical T-HRIS.

Of the seven conceptual analyses, our SLR identified, there was representation in both

technical T-HRIS and managerial T-HRIS components. Like the empirical representation, the

conceptual analysis favors technical T-HRIS components considering six of the seven articles

related to Best Practices (three articles), Recruitment Information Systems (two articles), and

Employee Training and Development (one article). The only component of managerial T-HRIS

represented was Employee Performance and Satisfaction, leaving room for future research

opportunities.

Finally, of the two literature reviews, our SLR identified, only one component out of all

six possible (four technical, two managerial) was applicable. The Employee Performance and

Satisfaction component was represented in both reviews, leaving room for future literature

reviews analyzing technical T-HRIS and AI applications.

41
 CHAPTER 6: DISCUSSION

Based on the results analysis of our research, the application of AI within tactical HRIS

practices needs further exploration within academia in a variety of areas. First, we consider the

contributions this SLR has provided to the existing literature and surrounding communities.

Secondly, we discuss future research opportunities this SLR has shined a light upon to further

develop this field and agenda. Lastly, we discuss the limitations of our research and how future

scholars can mitigate the limitations we faced.

Through our research, we seek to shine a light on the importance of AI and HRM

practices through the exploration of existing literature relating to tactical HRIS components. The

prominent growth of digitalization within the work force and the evolution of AI within HRM

has helped improve areas like employee experiences and performance (Garg et al., 2021). Thus,

our contribution highlights the importance of understanding where AI is best represented in

tactical HRIS components and where academics writ large are focusing their attention. We

sought to provide inductive insights to the current state of AI within HRM based on a literature-

driven and systematic methodology that is repeatable for future research opportunities to come

for both the academic community and industry professionals.

With regards to research avenues to explore regarding AI applications within tactical

HRIS components, we discovered several gaps which may help expand the agenda of AI within

HRM. First, there is a glaring gap in published literature (relative to our search) between

managerial and technical HRIS practices. Where task-oriented and datadriving HRIS

components see more representation in literature, the more intuitive and emotionally intelligent

practices lack attention. Comparatively, this provides an opportunity to dive into why this gap in

42
research exists and further exploration as to why organizations veer away from incorporating AI

into practices that require intuitive, genuine, humancentric perspectives.

Secondly, we showcase Employee Compensation and Benefits programs research lacks

heavily in academic literature, given the parameters we set. Although we cast a large net of 90

journals to consider during our analysis, no articles of the 33 identified celebrated direct AI

applications within Employee Compensation and Benefits programs. This discovery invites

future scholars to empirically, qualitatively, and conceptually explore how AI applications

directly affect benefits and compensation programs within an organization and make that tactical

HRIS component the star of the article.

Moving forward, the implications of this research can equally guide the academic

community and industry professionals on understanding where AI exists within T-HRIS. This

offers an opportunity for synergy between the academics and industry professionals, as this work

provides insight to potential research areas regarding T-HRIS components and where AI is

lacking or is overrepresented. Furthermore, these findings provide an opportunity for industry

professionals to educate themselves on potential technologies which may streamline monotonous

processes. Moreover, the effects of this research provide the academic community with

awareness to the state of which AI is explored and the lack of research within managerial T-

HRIS components when compared to the technical spectrum.

43
 CHAPTER 7: CONCLUSION

The purpose of this research was to review the integration of AI within HRM and HRIS

components; admittedly, there were limitations we encountered during our research that merit

exploring. First, we used a broad range of keywords like “artificial intelligence” and “human

resource management” within our search string to provide us with a wide coverage of topics in

HR and AI. This may have inhibited our ability to gather insights from other authors which may

have emerged outside of the selected keywords used in our string. Furthermore, we limited our

search to a specified number of research journals, which may have impacted the number we were

able to generate and analyze within our search. Lastly, we were limited in the access to

databases. To enhance this search, incorporating other reputable databases like SCOPUS may

provide more robust results to incorporate in a researcher’s literature search.

Additionally, this paper presents an SLR analyzing AI-enhanced technical (data-driven)

and managerial (human-centric) T-HRIS. The reviewed literature presents a research gap

between technical and managerial T-HRIS, which is evident upon reviewing 33 articles across

four different databases, six T-HRIS components, four different AI methodologies, and four

article types. This analysis discovered that the more data-driven and task-orientated T-HRIS

applications are saturated in literature, and the managerial practices are less represented. The

review of existing research provides a foundation and direction for future research to address the

gap between technical and managerial T-HRIS AI applications.

Going into this analysis, we set a goal to identify which T-HRIS components existed in

academic literature and how they were represented and sought to answer the research questions

derived from this goal. Based on our SLR, we could sift through 350,053 articles based on

specified criteria and narrow the list down to 33 articles that met the requirements and

44
parameters we sought to investigate. Upon completing our meticulous analysis, we have decided

to address some opportunities for future research based on the SLR conducted: First, technical T-

HRIS (resource and task-driven) components are overrepresented. There is potential for future

exploration on the managerial T-HRIS components, which are more human-centric. Second,

there are significantly more empirical studies when analyzing the relationship between AI and T-

HRIS, leaving room to expand the research agenda by providing more qualitative analysis,

conceptual analysis, and literature review research. In addition, out of all the literature analyzed

within this SLR, to include Phase 1 and Phase 2 of our methodology, this research solely takes

technical and managerial HRM and HRIS practices into consideration and individually analyzes

them. This review also paves the way for future exploration regarding technical and managerial

HRM and HRIS and its relationship with AI. Finally, based on our SLR, the Information

Systems and Management publication community claims 15 of the 33 total articles, which gives

us reason to believe these sources may be more receptive to research regarding AI and HRM

practices. This highlights potential blind spots for other publication venues seeking analysis on

AI and T-HRIS components.

Through this research, we seek to enhance research developments within the HR, IS, and

AI communities by bringing to light gaps within existing literature through a systematic review.

Although we faced limitations regarding our search string, journal analysis, and database usage,

we reinforce the future research agenda by providing new avenues to explore. We further

reinforce the motion that there is room for future academics and professionals to explore AI

applications in managerial HRIS components. Similarly, we showcase the research opportunity

for Employee Compensation and Benefits systems research and AI applications. To conclude, we

celebrate the academic and literature advancements of AI within HRIS and HRM and are excited

45
highlight the research potential which exists within tactical HRIS components. This review

provides the necessary groundwork to further grow this research agenda and gain a deeper

understanding of why the identified gaps exist with in literature and where professional and

academic stakeholders can focus their attention to further grow the field.

46
 APPENDIX
Appendix A: Academic Journals
Information Systems and Management (18):
Decision Support Systems
European Journal of Information Systems
Expert Systems with Applications
Government Information Quarterly
Information and Management
Information and Organization
Information Society
Information Systems Frontiers
Information Systems Journal
Information Systems Research
Information Technology and People
International Journal of Electronic Commerce
International Journal of Information Management
Journal of Association of Information Systems
Journal of Information Technology
Journal of Management Information Systems
Journal of Strategic Information Systems
Management Information Systems (MIS) Quarterly
Human Resource Management Journals (9):
Human Resource Management
Asia Pacific Journal of Human Resources
Employee Relations
Human Resource Management Journal
Human Resource Management Review
International Journal of Human Resource Management
International Journal of Manpower
New Technology, Work and Employment
Personnel Review
Human Resource Development Journals (12):
Advances in Developing Human Resources
Career Development International
Education and Training
European Journal of Training and Development
Human Resource Development International
Human Resource Development Quarterly
International Journal of Training and Development
Journal of Education and Work
Journal of Management Development
47
General Management Journals (40):
Academy of Management Annals
Academy of Management Journal
Academy of Management Review
Administrative Science Quarterly
Asia Pacific Journal of Management
Business Horizons
Business Strategy and the Environment
Decision Sciences
European Journal of Operational Research
Global Strategy Journal
Human Relations
Industrial Relations
International Business Review
International Journal of Management Reviews
International Marketing Review
Journal of Applied Behavioral Science
Journal of Applied Psychology
Journal of Business Research
Journal of International Business Studies
Journal of International Management
Journal of International Marketing
Journal of Knowledge Management
Journal of Management
Journal of Management Education
Journal of Management Studies
Journal of Occupational and Organizational
Psychology
Journal of Organizational Behavior
Journal of Service Research
Journal of Vocational Behavior
Journal of World Business
Management International Review
Management Science
Organization Science
Organization Studies
Organizational Behavior and Human Decision
Processes
Personnel Psychology
Psychological Bulletin
Strategic Change
Journal of Vocational Education and Training
Journal of Workplace Learning
Human Resource Development Journals Continued
(12):
Learning Organization
Computer Science and Engineering Journals (4):
Computers and Industrial Engineering
Computers in Human Behavior
International Journal of Human-Computer Studies
Journal of Computer Mediated Communication
48
Strategic Management Journal
Work, Employment and Society
Small Business Journal (7):
Entrepreneurship Theory and Practice
International Entrepreneurship and Management
Journal
International Journal of Entrepreneurial
Behavior and Research
International Journal of Entrepreneurship and
Innovation
International Small Business Journal
Journal of Small Business and Enterprise
Development
Journal of Small Business Management
 REFERENCES

Ahmed, O. (2018). Artificial intelligence in HR. International Journal of Research and Analytical
Reviews, 5(4), 971-978.

Amer-Yahia, S., Basu Roy, S., Chen, L., Morishima, A., Abello Monedero, J., Bourhis, P., ... &
Yoshida, K. (2020). Making ai machines work for humans in fow. ACM SIGMOD Record, 49(2), 30-
35.

Badri, H., Bashiri, M., & Hejazi, T. H. (2013). Integrated strategic and tactical planning in a supply
chain network design with a heuristic solution method. Computers & Operations Research, 40(4),
1143–1154. https://doi.org/10.1016/j.cor.2012.11.005

Bani-Hani, D., & Khasawneh, M. (2019). A Recursive General Regression Neural Network (R-
GRNN) Oracle for classification problems. EXPERT SYSTEMS WITH APPLICATIONS, 135, 273–
286. https://doi.org/10.1016/j.eswa.2019.06.018

Bhattacharyya, S. S., & Nair, S. (2019). Explicating the future of work: Perspectives from India. The
Journal of Management Development, 38(3), 175–194.http://dx.doi.org.libweb.lib.utsa.edu/10.1108/
JMD-01-2019-0032

Black, J. S., & van Esch, P. (2020). AI-enabled recruiting: What is it and how should a manager use
it? Business Horizons, 63(2), 215–226. https://doi.org/10.1016/j.bushor.2019.12.001

Bourhis, P., Demartini, G., Elbassuoni, S., Hoareau, E., & Rao, H. R. (2019). Ethical Challenges in
the Future of Work. IEEE Data Eng. Bull., 42(4), 55-64.

Bhuiyan, F., Chowdhury, M. M., & Ferdous, F. (2014). Historical evolution of human resource
information system (HRIS): An interface between HR and computer technology. Human Resource
Management Research, 4(4), 75-80.

Cain, L. N., Thomas, J. H., & Alonso, M. (2019). From sci-fi to sci-fact: The state of robotics and AI
in the hospitality industry. Journal of Hospitality and Tourism Technology, 10(4), 624–650. http://
dx.doi.org.libweb.lib.utsa.edu/10.1108/JHTT-07-2018-0066

Castillo, A., Vander Meer, D., & Castellanos, A. (2018). ExUP recommendations: Inferring user’s
product metadata preferences from single-criterion rating systems. Decision Support Systems, 108,
69–78. https://doi.org/10.1016/j.dss.2018.02.006

Chakraborty, A., & Kar, A. K. (2021). How did COVID-19 impact working professionals – a
typology of impacts focused on education sector. The International Journal of Information and
Learning Technology, 38(3), 273–282. https://doi.org/10.1108/IJILT-06-2020-0125

49
Chang, W., & Jung, C. (2017). A hybrid approach for personalized service staff recommendation.
Information Systems Frontiers, 19(1), 149–163. ProQuest One Academic; SciTech Premium
Collection. https://doi.org/10.1007/s10796-015-9597-7

Cheng-Kui, H., Wang Tawei, & Huang Tzu-Yen. (2020). Initial Evidence on the Impact of Big Data
Implementation on Firm Performance. Information Systems Frontiers, 22(2), 475–487. ProQuest One
Academic; SciTech Premium Collection. https://doi.org/10.1007/s10796-018-9872-5

Collings, D. G., Wood, G. T., & Szamosi, L. T. (Eds.). (2018). Human Resource Management: A
Critical Approach (2nd ed.). Routledge. https://doi.org/10.4324/9781315299556

Collinson, V. (1996). Becoming an Exemplary Teacher: Integrating Professional, Interpersonal, and

Intrapersonal Knowledge. https://eric.ed.gov/?id=ED401227

Cregan, C., Kulik, C. T., Johnston, S., & Bartram, T. (2021). The influence of calculative (“hard”)
and collaborative (“soft”) HRM on the layoff-performance relationship in high performance
workplaces. Human Resource Management Journal, 31(1), 202–224. https://doi.org/10.1111/1748-
8583.12291

Dabirian, A., Kietzmann, J., & Diba, H. (2017). A great place to work!? Understanding crowdsourced
employer branding. Business Horizons, 60(2), 197–205. https://doi.org/10.1016/j.bushor.2016.11.005

Dahlbom, P., Siikanen, N., Sajasalo, P., & Jarvenpää, M. (2020). Big data and HR analytics in the
digital era. Baltic Journal of Management, 15(1), 120–138. https://doi.org/10.1108/BJM-11-2018-
0393

De Kock, F. S., Lievens, F., & Born, M. Ph. (2020). The profile of the ‘Good Judge’ in HRM: A
systematic review and agenda for future research. Human Resource Management Review, 30(2),
100667. https://doi.org/10.1016/j.hrmr.2018.09.003

Dhamija, P., & Bag, S. (2020). Role of artificial intelligence in operations environment: A review and
bibliometric analysis. The TQM Journal, 32(4), 869–896. https://doi.org/10.1108/TQM-10-2019-
0243

Doctor, F., Hagras, H., Roberts, D., & Callaghan, V. (2009a). A fuzzy based agent for group decision
support of applicants ranking within recruitment systems. 2009 IEEE Symposium on Intelligent
Agents, 8–15. https://doi.org/10.1109/IA.2009.4927494

Doctor, F., Hagras, H., Roberts, D., & Callaghan, V. (2009b). A neuro-fuzzy based agent for group
decision support in applicant ranking within human resources systems. 2009 IEEE International
Conference on Fuzzy Systems, 744–750. https://doi.org/10.1109/FUZZY.2009.5277379

Dressel, J., & Farid, H. (2018). The accuracy, fairness, and limits of predicting recidivism. Science
Advances, 4(1), eaao5580. https://doi.org/10.1126/sciadv.aao5580

50
Escalante, H. J., Guyon, I., Escalera, S., Jacques, J., Madadi, M., Baró, X., Ayache, S., Viegas, E.,
Güçlütürk, Y., Güçlü, U., van Gerven, M. A. J., & van Lier, R. (2017). Design of an explainable
machine learning challenge for video interviews. 2017 International Joint Conference on Neural
Networks (IJCNN), 3688–3695. https://doi.org/10.1109/IJCNN.2017.7966320

Eubanks, B. (2018). Artificial Intelligence for HR: Use AI to Support and Develop a Successful
Workforce. Kogan Page Publishers.

Garg, S., Sinha, S., Kar, A. K., & Mani, M. (2021). A review of machine learning applications in
human resource management. International Journal of Productivity and Performance Management,
ahead-of-print(ahead-of-print). https://doi.org/10.1108/IJPPM-08-2020-0427

Garg, R., Kiwelekar, A. W., Netak, L. D., & Ghodake, A. (2021). i-Pulse: A NLP based novel
approach for employee engagement in logistics organization. International Journal of Information
Management Data Insights, 1(1), 100011. https://doi.org/10.1016/j.jjimei.2021.100011

Fehrenbacher, D. D. (2017). Affect Infusion and Detection through Faces in Computer-mediated

Knowledge-sharing Decisions. Journal of the Association for Information Systems, 18(10), 703–726.

Foot, H. (2020). Approach to improving training of human workers in industrial applications through
the use of Intelligence Augmentation and Human-in-the-Loop. 2020 15th International Conference on
Computer Science Education (ICCSE), 283–288. https://doi.org/10.1109/ICCSE49874.2020.9201867

Hmoud, B. I., & Várallyai, L. (2020). Artificial Intelligence in Human Resources Information
Systems: Investigating its Trust and Adoption Determinants. International Journal of Engineering and
Management Sciences, 5(1), 749–765.

Hmoud, B., & Laszlo, V. (2019). Will artificial intelligence take over human resources recruitment
and selection?. Network Intelligence Studies, 7(13), 21-30.

Huang, M.-H., & Rust, R. T. (2018). Artificial Intelligence in Service. Journal of Service Research,
21(2), 155–172. https://doi.org/10.1177/1094670517752459

Hughes, C., Robert, L., Frady, K., Arroyos, A., Hughes, C., Robert, L., Frady, K., & Arroyos, A.
(2019). Artificial Intelligence, Employee Engagement, Fairness, and Job Outcomes. In Managing
Technology and Middle- and Low-skilled Employees (pp. 61–68). Emerald Publishing Limited.
https://doi.org/10.1108/978-1-78973-077-720191005

Jabr, W., & Zheng, Z. (Eric). (2014). Know Yourself and Know Your Enemy: An Analysis of Firm
Recommendations and Consumer Reviews in a Competitive Environment. MIS Quarterly, 38(3),
635-A10.

Johnson, R. D., Stone, D. L., & Lukaszewski, K. M. (2020). The benefits of eHRM and AI for talent
acquisition. Journal of Tourism Futures, 7(1), 40–52. https://doi.org/10.1108/JTF-02-2020-0013

51
Jones, C., & Saundry, R. (2012). The practice of discipline: Evaluating the roles and relationship
between managers and HR professionals. Human Resource Management Journal, 22(3), 252–266.
https://doi.org/10.1111/j.1748-8583.2011.00175.x

Jøranli, I. (2018). Managing organisational knowledge through recruitment: Searching and selecting
embodied competencies. Journal of Knowledge Management, 22(1), 183–200. Entrepreneurship
Database; ProQuest One Academic; SciTech Premium Collection. https://doi.org/10.1108/JKM-12-
2016-0541

Kaminska, R., & Borzillo, S. (2018). Challenges to the learning organization in the context of
generational diversity and social networks. Learning Organization, 25(2), 92–101. https://doi
.org/10.1108/TLO-03-2017-0033

Karatop, B., Kubat, C., & Uygun, Ö. (2015). Talent management in manufacturing system using
fuzzy logic approach. Computers & Industrial Engineering, 86, 127–136. https://doi.org/10.1016/
j.cie.2014.09.015

Karim, A. J. (2011). The significance of management information systems for enhancing strategic
and tactical planning. JISTEM - Journal of Information Systems and Technology Management, 8,
459–470. https://doi.org/10.4301/S1807-17752011000200011

Khan, W. A., Chung, S. H., Awan, M. U., & Wen, X. (2020). Machine learning facilitated business
intelligence (Part I). Industrial Management & Data Systems, 120(1), 164–195. https://doi.org/10.1
108/watIMDS-07-2019-0361

Kitchenham, B. A. (2012, September). Systematic review in software engineering: where we are and
where we should be going. In Proceedings of the 2nd international workshop on Evidential
assessment of software technologies. 1-2.

Koch, J., Plattfaut, R., & Kregel, I. (2021). Looking for Talent in Times of Crisis – The Impact of the
Covid-19 Pandemic on Public Sector Job Openings. International Journal of Information
Management Data Insights, 1(2), 100014. https://doi.org/10.1016/j.jjimei.2021.100014

Kotera, Y., Sheffield, D., & Van Gordon, W. (2019). The applications of neuro-linguistic
programming in organizational settings: A systematic review of psychological outcomes. Human
Resource Development Quarterly, 30(1), 101–116. https://doi.org/10.1002/hrdq.21334

Kretzer, M., & Maedche, A. (2018). Designing Social Nudges for Enterprise Recommendation
Agents: An Investigation in the Business Intelligence Systems Context. Journal of the Association for
Information Systems, 19(12), 1145–1186. http://dx.doi.org.libweb.lib.utsa.edu/10.17705/1jais.00523

Kumar, R. (2012). Human resource information system: An innovative strategy for human resource
management. Gian Jyoti E-Journal, 1(2), 1-12.

52
Laker, D. R., & Powell, J. L. (2011). The differences between hard and soft skills and their relative
impact on training transfer. Human Resource Development Quarterly, 22(1), 111–122. https://doi
.org/10.1002/hrdq.20063

Lankton, N. K., McKnight, D. H., & Tripp, J. (2015). Technology, Humanness, and Trust: Rethinking
Trust in Technology. Journal of the Association for Information Systems, 16(10), 880–918.

Lee, D., & Ahn, C. (2020). Industrial human resource management optimization based on skills and
characteristics. Computers & Industrial Engineering, 144, N.PAG-N.PAG. https://doi.org/10.1016
/j.cie.2020.106463

Li, Z., Xu, W., Zhang, L., & Lau, R. (2014). An ontology-based Web mining method for
unemployment rate prediction. DECISION SUPPORT SYSTEMS, 66, 114–122. https://doi.org/
10.1016/j.dss.2014.06.007

Lima, M. (2020). Smarter organizations: Insights from a smart city hybrid framework. International
Entrepreneurship and Management Journal, 16(4), 1281–1300. Entrepreneurship Database; ProQuest
One Academic. https://doi.org/10.1007/s11365-020-00690-x

Liu, Y., Pant, G., & Sheng, O. R. L. (2020). Predicting Labor Market Competition: Leveraging
Interfirm Network and Employee Skills. Information Systems Research, 31(4), 1443–1466. https://
doi.org/10.1287/isre.2020.0954

Maity, S. (2019). Identifying opportunities for artificial intelligence in the evolution of training and
development practices. Journal of Management Development, 38(8), 651–663. https://doi.org/10.1
108/JMD-03-2019-0069

Majumder, S., & Mondal, A. (2021). Are chatbots really useful for human resource management?
International Journal of Speech Technology. https://doi.org/10.1007/s10772-021-09834-y

Malik, A., Budhwar, P., Patel, C., & Srikanth, N. R. (2020). May the bots be with you! Delivering
HR cost-effectiveness and individualised employee experiences in an MNE. The International
Journal of Human Resource Management, 0(0), 1–31. https://doi.org/10.1080/09585192.202
0.1859582

Martinez-Gil, J., Paoletti, A. L., & Pichler, M. (2020). A Novel Approach for Learning How to
Automatically Match Job Offers and Candidate Profiles. Information Systems Frontiers, 22(6),
1265–1274. ProQuest One Academic; SciTech Premium Collection. https://doi.org/10.1007/s10796-
019-09929-7

Mayfield, M., Mayfield, J., & Lunce, S. (2003). Human resource information systems: A review and
model development. Advances in Competitiveness Research, 11(1), 139–151.

McCarthy, J., Minsky, M., Rochester, N., & Shannon, C. (2006). A Proposal for the Dartmouth
Summer Research Project on Artificial Intelligence: August 31, 1955—ProQuest. AI Magazine,
27(4), 12–14.

53
Mitchell, T. M. (2006). The discipline of machine learning (Vol. 9). Pittsburgh: Carnegie Mellon
University, School of Computer Science, Machine Learning Department.

Molnar, C. (2020). Interpretable machine learning. Lulu. com.

Moor, J. (2006). The Dartmouth College Artificial Intelligence Conference: The Next Fifty Years. AI
Magazine, 27(4), 87–87. https://doi.org/10.1609/aimag.v27i4.1911

Muhammad, A. U., Shah, Z. A., & Azhar, K. A. (2021). The Increasing Role of Hris in Facilitating
Hr Functions in Pakistan’s Banking Sector. International Journal of Information, Business and
Management, 13(1), 24–34.

Noone, B. M., & Coulter, R. C. (2012). Applying Modern Robotics Technologies to Demand
Prediction and Production Management in the Quick-Service Restaurant Sector. Cornell Hospitality
Quarterly, 53(2), 122–133. https://doi.org/10.1177/1938965511434112

Obeidat, B. Y. (2012). The Relationship between Human Resource Information System (HRIS)
Functions and Human Resource Management (HRM) Functionalities. Journal of Management
Research, 4(4). https://doi.org/10.5296/jmr.v4i4.2262

Okoli, C. (2015). A Guide to Conducting a Standalone Systematic Literature Review.

Communications of the Association for Information Systems, 37(1). https://doi.org/10.
17705/1CAIS.03743

Olson, P. (2018, March 2). This Chatbot Is Helping People Track Harassment At Work. Forbes.
https://www.forbes.com/sites/parmyolson/2018/03/02/chatbot-spot-sexual-harassment-ai/

Papacharissi, Z. (2009). The virtual geographies of social networks: A comparative analysis of

Facebook, LinkedIn and ASmallWorld. New Media & Society, 11(1–2), 199–220. https://doi.org/
10.1177/1461444808099577

Pessach, D., Singer, G., Avrahami, D., Chalutz Ben-Gal, H., Shmueli, E., & Ben-Gal, I. (2020).
Employees recruitment: A prescriptive analytics approach via machine learning and mathematical
programming. Decision Support Systems, 134, 113290. https://doi.org/10.1016/j.dss.2020.113290

Pratt, M., Mohcine, B., Taskin, N., & Cakula, S. (2021). Use of AI for Improving Employee
Motivation and Satisfaction (pp. 289–299). https://doi.org/10.1007/978-3-030-68201-9_30

Prentice, C., Lopes, S. D., & Wang, X. (2020). Emotional intelligence or artificial intelligence– an
employee perspective. Journal of Hospitality Marketing & Management, 29(4), 377–403. https://doi
.org/10.1080/19368623.2019.1647124

Rahmani, D., & Kamberaj, H. (2021). Implementation and Usage of Artificial Intelligence Powered
Chatbots in Human Resources Management Systems.

54
Rathi, D. R. (2018). Artificial intelligence and the future of hr practices. International Journal of
Applied Research, 4(6), 113-116.

Ridhwan, K. M., & Hargreaves, C. A. (2021). Leveraging Twitter Data to Understand Public
Sentiment for the COVID‐19 Outbreak in Singapore. International Journal of Information
Management Data Insights, 100021.

Robert, L. P., Pierce, C., Marquis, L., Kim, S., & Alahmad, R. (2020). Designing fair AI for
managing employees in organizations: A review, critique, and design agenda. Human–Computer
Interaction, 35(5–6), 545–575. https://doi.org/10.1080/07370024.2020.1735391

Rudin, C. (2019). Stop explaining black box machine learning models for high stakes decisions and
use interpretable models instead. Nature Machine Intelligence, 1(5), 206–215.
https://doi.org/10.1038/s42256-019-0048-x

Rybinski, K., & Tsay, V. (2018). The Application of Machine Learning in Faculty Assessment: A
Case Study of Narxoz University. HUMAN RESOURCE MANAGEMENT, 122/123(3/4), 145–170.

Saha, P., Bose, I., & Mahanti, A. (2016). A knowledge based scheme for risk assessment in loan
processing by banks. DECISION SUPPORT SYSTEMS, 84, 78–88. https://doi.org/10.1016/j.dss.
2016.02.002

Scandura, T. A., & Williams, E. A. (2000). Research Methodology in Management: Current

Practices, Trends, and Implications for Future Research. The Academy of Management Journal,
43(6), 1248–1264. https://doi.org/10.2307/1556348

Schuetz, S., & Venkatesh, V. (2020). Research Perspectives: The Rise of Human Machines: How
Cognitive Computing Systems Challenge Assumptions of User-System Interaction. Journal of the
Association for Information Systems, 21(2), 460–482. http://dx.doi.org.libweb.lib.utsa.edu/10.177
05/1jais.00608

Shum, H., He, X., & Li, D. (2018). From Eliza to XiaoIce: Challenges and opportunities with social
chatbots. Frontiers of Information Technology & Electronic Engineering, 19(1), 10–26. https://doi.
org/10.1631/FITEE.1700826

Stapleton, P., Luiz, G., & Chatwin, H. (2017). Generation Validation: The Role of Social Comparison
in Use of Instagram Among Emerging Adults. Cyberpsychology, Behavior, and Social Networking,
20(3), 142–149. https://doi.org/10.1089/cyber.2016.0444

Stewart, T. A. (2007). The wealth of knowledge: Intellectual capital and the twenty-first century
organization. Currency.

Tariq, O., Sang, J., & Gulzar, K. (2016). Design and Implementation of Human Resource Information
Systems Based on MVC a Case Study Vocational Education in Iraq. International Journal of U- and
e- Service, Science and Technology, 9(11), 15–26. https://doi.org/10.14257/ijunesst.2016.9.11.02

55
Tong, S., Jia, N., Luo, X., & Fang, Z. (2021). The Janus Face of Artificial Intelligence Feedback:
Deployment Versus Disclosure Effects on Employee Performance. Strategic Management Journal.

University of Maryland. (2021). University Libraries Research Guides: Steps of a Systematic Review.
University of Maryland Libraries. https://lib.guides.umd.edu/SR/steps

Using machine learning to translate applicant work history into predictors of performance and
turnover. - PsycNET. (n.d.). Retrieved August 25, 2021, from https://psycnet.apa.org/doiLan
ding?doi=10.1037%2Fapl0000405

van Esch, P., Black, J., & Ferolie, J. (2019). Marketing AI recruitment: The next phase in job
application and selection. COMPUTERS IN HUMAN BEHAVIOR, 90, 215–222. https://doi.org/
10.1016/j.chb.2018.09.009

van Esch, P., & Black, J. S. (2019). Factors that influence new generation candidates to engage with
and complete digital, AI-enabled recruiting. Business Horizons, 62(6), 729–739. https://doi.org/10.1
016/j.bushor.2019.07.004

Wang, X., Wang, L., Zhang, L., Xu, X., Zhang, W., & Xu, Y. (2017). Developing an employee
turnover risk evaluation model using case-based reasoning. Information Systems Frontiers, 19(3),
569–576. ProQuest One Academic; SciTech Premium Collection. https://doi.org/10.1007/s10796-
015-9615-9

Watson, H. (2017). Preparing for the cognitive generation of decision support. MIS Quarterly
Executive, 16, 153–169.

Wohlin, C., Runeson, P., Höst, M., Ohlsson, M. C., Regnell, B., & Wesslén, A. (2012).
Experimentation in software engineering. Springer Science & Business Media.

Xiao, B., & Benbasat, I. (2007). E-Commerce Product Recommendation Agents: Use,
Characteristics, and Impact. MIS Quarterly, 31(1), 137–209. https://doi.org/10.2307/25148784

Xu, D., & Xiao, X. (2020). Influence of the Development of VR Technology on Enterprise Human
Resource Management in the Era of Artificial Intelligence. IEEE Access, 1–1. https://doi.org
/10.1109/ACCESS.2020.3020622

Yorks, L., Rotatori, D., Sung, S., & Justice, S. (2020). Workplace Reflection in the Age of AI:
Materiality, Technology, and Machines. Advances in Developing Human Resources, 22(3), 308–319.
https://doi.org/10.1177/1523422320927299

Zehir, C., Karaboğa, T., & Başar, D. (2020). The Transformation of Human Resource Management
and Its Impact on Overall Business Performance: Big Data Analytics and AI Technologies in
Strategic HRM. In U. Hacioglu (Ed.), Digital Business Strategies in Blockchain Ecosystems:
Transformational Design and Future of Global Business. Springer International Publishing, 265-279.
https://doi.org/10.1007/978-3-030-29739-8_12

56
ESSAY II. JC-COMPASS: A FRAMEWORK FOR CONDUCTING COMPETENCY-

BASED JOB POSTING RESEARCH AND ANALYSIS

57
 CHAPTER 1: ABSTRACT

As Artificial Intelligence (AI) and Machine Learning (ML) capabilities expand across the

data science career field, Human Resource (HR) professionals face the challenge of acquiring top

talent to fill necessary vacancies within their respective organization. Thus, developing strategies

to orient and communicate the needs of the organization have become paramount in maintaining

a competitive edge in recruiting employees for job vacancies. Accordingly, there has been

considerable emphasis on understanding the specific skills and experience demanded of data

scientists within job postings. However, most methods have yielded lists of specific skills

demanded of data scientists and very little insight on systematic approaches to thematically

analyzing job postings.

Thus, we build and introduce a framework to guide such research called Job Competency

Compass: Initiate, Establish, Analyze, and Socialize (JC-Compass). The JC-Compass framework

provides a focused and systematic approach with detailed guidelines for industry and academic

entities interested in understanding and researching thematic trends within job postings by

building dictionaries of word similarity . The four steps to conducting job description research

are outlined: 1) initiate competency research, 2) establish competencies, 3) analyze

competencies, and 4) socialize the findings. To illustrate how the JC-Compass can be utilized, an

example empirical study leveraging the TF-IDF method is reported. This example analysis

illustrates how JC-Compass can guide industry professionals and scholars in analyzing thematic

word similarity within job description when evolving recruitment strategies.

58
 CHAPTER 2: INTRODUCTION

With the evolution of Artificial Intelligence (AI) and Machine Learning (ML) saturating

the workforce, companies face the challenge of hiring and internally securing the right talent to

manage, develop, and operate these assets ethically (Enholm et al., 2022). This evolution has

consequently affected how human resource professionals manage the necessary skill demands

that come with the evolution of technology and work environments. The same human resource

professionals are responsible for effectively communicating the company's needs and tailoring

this demand to attract prospective applicants and train current employees. Thus, competency-

based approaches have become a popular technique to oversee the portfolio of skills the

organization owns, as there is an added pressure to ensure those competencies capture the

emerging skills demanded of the 21st Century (Boyartiz, 2008).

Big data, AI, and ML have accelerated business intelligence capabilities for the United

States federal government and private sector. As a result, research endeavors exploring job

targeting accuracy and skills needed to manage challenges within jobs have grown considerably

in recent years. Social capabilities (soft skills), team experience (both leader and followership),

and technical understanding have become key points of interest for academic and industry

researchers investigating job postings (Anton et al., 2020; Huang, 2021). However, one major

research gap that IS scholars have currently left unexplored is the investigation of leveraging

natural language processing (NLP) techniques to gain thematic understanding of competency

trends within job postings.

Today, there has been little work which has reported on large-scale identification,

collection and analysis of competency themes which exist within job postings. Within current job

posting research, many authors and investigators explore specific skills needed for the job; albeit

59
we understand skill identification is imperative to grasp current workforce demands given a

specific career field. Modeling our research after Kim et al.’s (2005) design science approach,

we propose a job posting research framework which steps away from analyzing specific skills

and encourages NLP capabilities to investigate thematic trends based within job postings. Such a

framework could enrich academic and industry professional’s perspectives by providing a

systematic approach to explore job posting themes as well as enhance acquisition strategies.

The arrangement of this paper is organized into the following sections. First, we discuss

the background on job posting research which leverages NLP capabilities. This section compares

job posting research with other tactical human resource management (T-HRM) research that has

specifically used NLP capabilities. Furthermore, this section highlights relevant and recent job

posting research and existing frameworks. Next, details are provided concerning how to

operationalize a job posting research project. Within this section, we highlight methods in

identifying appropriate data sources (job posting sites), as well as techniques for collecting

content from these sources. We further provide insight to potential analytical directions once the

job postings have been collected. Following this section, we conduct a post-hoc analysis to

further explore and identify how the groups may differ from each other by year. Finally, we

discuss the contributions and implications of this work. To summarize, this paper provides a

comprehensive roadmap for industry professionals and researchers to successfully conduct

competency-based job posting research

60
 CHAPTER 3: BACKGROUND

While human resource analytics is a rich topic within business literature, there has been a

drastic growth in interest surrounding the importance of business intelligence and streamlining

decision processes via the incorporation of machine learning and AI. That said, HRM is not a

catch-all term. There are many spectrums, perspectives, and unique responsibilities of human

resource management that exist. For instance, there is a strategic lens of HRM which considers

the overall effects of activities which enable an organization to achieve its goals. This lens

considers how HRM -as a whole- affects performance-related outcomes at the organizational

level (Boon et al, 2018; Perlman, 2016). Unethical practices (biases), market competition, and

reduction of resources are challenges that exist at this level which affect an organization’s health,

reputation, and performance (Perlman, 2016). Beneath the strategic level, there is a tactical lens

which refers to the managerial and technical actions and responsibilities which fall under the

HRM scope of responsibility (Votto et al., 2021). Managerial components are centered around

employee performance, expertise, culture, and discipline. This subsection of T-HRM consists of

employee performance, satisfaction, and discipline management. The technical component refers

to data-driven capabilities and responsibilities within an organization. Best practices (customer

engagement/workflow), recruitment, employee training and development, and compensation

management fall under the technical component.

Given the rich amount of data T-HRM encompasses, it is no stranger to machine learning

and AI methodologies to help cope with analyzing trends. Specifically, NLP capabilities have

become a useful tool when analyzing text, speech, and semantics. In essence, NLP capabilities

seek to understand input from human speech and/or text and generate responses understandable

to humans (Kumar, 2019). Such technologies include chatbots which that interact with humans

61
and autonomously answer questions related to recruitment qualifications, training opportunities,

leave requests, and onboarding procedures (Garg et al., 2021; Majumder and Mondal, 2021).

Other T-HRM components have leveraged NLP capabilities to analyze copious amounts of data

and extract impactful information from it. Such practices have been seen screening prospective

resumes (Chuang et al., 2009; Cole et al., 2003), reviewing employee climate surveys (Layug

2018), and detecting fake job postings (Amaar et al., 2022). The ability for NLP to help HRM

professionals develop informed decisions creates opportunities for business researchers to

contribute to an extremely impactful body of work. Thus, the exploration and development of

new capabilities for NLP could support organizations and businesses in generating stronger

strategies to achieve established goals.

Specifically, recent research within the recruitment T-HRM component has investigated

the ability to leverage job postings from across the internet to make inferences on trending skill

demands of employers. There are online communities which foster a space for businesses to post

job requisitions such as LinkedIn, Indeed, GlassDoor, Google, Monster, and many others. Within

these prominent websites are millions of job postings, usually tied to keywords related to the

vacant position. This enables prospective employees to type keywords into a search engine and

receive job postings catered to their search criteria. From these sites, employers and applicants

are connected. Information, application submissions, company hyperlinks, and other valuable

information are shared between the two parties. Given the amount of text information embedded

within each job posting, NLP has become an invaluable tool for analyzing thousands of job

postings to generate an understanding of trends and patterns. Although business literature is rich

with job posting research, there lacks a structured framework to guide job posting research. That

said, the following subsection dives deeper into recent literature surrounding the “recruitment”

62
T-HRM domain. In particular, job posting research and the use of NLP to conduct analyses. We

also discuss published frameworks involving job posting research.

Previous Job Description NLP Analysis and Framework Literature

With the recent explosion of academic and industry interest in NLP capabilities within

recruitment interests, HRM literature has yielded a variety of analyses that have built a

foundation for our research. Highlighting these impacts, we seek to convey how the contribution

of our research is unique and differs from existing initiatives. To comprehensively review

previous literature, we refer to eight peer-reviewed articles involving job description analysis and

NLP capabilities which are summarized in Table 1. We also review three published frameworks

which discuss job posting research efforts which is summarized in Table 2. We extracted these

articles from Google Scholar, leveraging the “most relevant” search option. Regarding Table 1,

we leveraged the following strings to generate a list of articles to review: “Job Posting Research

NLP”, “Job Description Research NLP”, “Job Posting Analysis NLP” and “Job Description

Analysis NLP”. We manually reviewed the first page of results (most relevant) and extracted

analyses which were directly related to job postings and NLP capabilities. Regarding Table 2, we

leveraged the same technique with different search strings. The search strings we utilized for

Table 2 were “Job Posting NLP Research Framework” and “Job Description NLP Research

Framework”.

Regarding Table 1, each article is unique and provides a thorough analysis of job

postings, we identify a similar structure in how these analyses are conducted, such that a

potential framework could be derived and built from them to enhance future research efforts. For

instance, each article identifies a source to collect their job postings from, be it Indeed or general

web crawling efforts. Furthermore, each article determines a collection methodology such as

63
scraping websites or manual extraction. Lastly, each article identifies an analysis technique such

as comparing statistical models to machine learning efforts or conducting content analyses.

The articles summarized in Table 2 showcase existing frameworks within the job posting

domain. Two of the four articles promote frameworks for extracting information out of resumes

to match up against job descriptions. The remaining articles discuss systematic approaches to

reviewing job descriptions via temporal, spatial, and curriculum considerations. By analyzing

these published frameworks, we discovered an opportunity to present a research model which

could systematically lead future research endeavors in analyzing job descriptions with NLP

capabilities. Rather than specialize in resume extraction and specific considerations against job

postings, we identified a prospect to add a new framework to the existing literature which

provides a systematic approach to conducting content analysis. Specifically, the proposed

framework promotes the use of Word2Vec modeling to build dictionaries based on established

job competencies within the organization. Our proposed framework differs from existing

literature as it promotes a thematic exploration of job postings based on competencies the

organization deems relevant to the career field. The following subsection dives deeper into the

theories which guide our competency-based job description research framework.

Table 1. Previous Job Posting NLP Literature Summary

Previous Data Sample Size Research Analytic Methods Findings

Studies Sources
Goldfarb US Job 4,269,779 Job Effectiveness of Correlation analysis Machine
et al. 2023 Postings Postings General Purpose between patent-based Learning and
(Burning Technologies GPT measures and job related data
Glass (GPT) posting measures science
Technologie technologies
s) growing in
popularity to be
a GPT

64
 Table 1. Previous Job Posting NLP Literature Summary Continued

Kortum et Computer 291 NLP Job Leverage text Evaluated data via Generated
al. 2022 Vision/NLP Postings; 591 mining similarity test and collection of
Job Computer techniques to extracted skills needed skills related to
Descriptions Vision Job compare from both categories of AI and ML,
(Indeed) Postings Computer Vision jobs showcased NLP
to NLP job and Computer
descriptions Vision Jobs
skills differ
significantly
Varelas et Online Job Not Disclosed Leverage NLP to Evaluated Support Proves simpler
al. 2022 Portals from classify jobs Vector Model, Random methods can be
Greek according to Forest, K-Nearest equal or more
Portals ISCO Neighbors, Stochastic powerful than
Occupation Gradient Descent, and deep learning
Codes for Labor Neural Network techniques if
Market classifier performance they are
Intelligence via accuracy and F1- combined
efforts scores properly
Vo et al. Public Job 2,907 Job Leverage NLP to Benchmarked Developed and
2022 Postings Descriptions; recommend Computer validated a tool
(GlassDoor); 1,886 Course personalized Science/Information which
Course Descriptions course/job Technology universities can
Descriptions recommendation Recommendation use to help
(university s for Computer System (CSIT-CRS) students find
websites) Science/Informat against other jobs
ion Technology established/validated
Students models & user survey
Zhu et al. Postdoctoral 823 Leveraged web Utilized NLP toolkit Demonstrated
2021 Job Postings Engineering scraping tools comprehensively ability of NLP
(Indeed) and Computer and NLP to review Knowledge techniques to
Science extract and Skill Assessments, generate insights
Postdoc Job analyze word frequency, and to postdoctoral
Postings Postdoctoral job domain disciplines experiences
postings within the computer within the
science and engineering computer science
career field – conducted and engineering
Kruskal Wallis test to career field
demonstrate
significance of
differences
Lunn et US 3,824 Job Demonstrate Python Libraries: Programming
al. 2020 Computer Postings (770- NLP capabilities WordCloud Bigram and testing are
Science Job New York; of extrapolating Analysis, Pandas (data important skills
Postings 774- San job data from manipulation), for employment
(Indeed) Francisco; online resources Matplotlib (generate within the
745- San Jose; diagrams), NumPy Computer
752- Seattle; (mathematics extension Science
783- of Matplotlib). community.
Washington Validated by randomly Specifically,
DC) inspecting 50 postings Python and
manually for accuracy machine learning
knowledge is
preferred.

65
 Table 1. Previous Job Posting NLP Literature Summary Continued
Meyer US 198 Job Conduct content Leverage R- Showcased
2019 Healthcare Postings analysis which programming NLP to uptick in demand
Data categorizes core code job postings, of data scientists
Science Job information from develop skill matrices, within healthcare
Postings job postings and visualize content organizations
(Indeed) within healthcare analysis and need for
industry professional
development,
training and
education to
grow career field
Sibarani UK Job 872 Job Analyze job Ontology-Based Automated
et al. 2017 Postings Advertisement postings based Information Extraction efforts proved
(Indeed) s on the context of and Paris/Keele method feasible in
skills and for cluster analysis extracting job
competencies compared against postings when
needed to fill a manual (human) compared to
role equivalent for accuracy of
extracting skills from human
job posts equivalent

Table 2. Previous Job Posting NLP Research Framework Summary

Previous Purpose of Validation Methods Implications

Studies Framework
Barducci et al. Extract Statistical evaluation Model demonstrated suitable
2022 information of BERT model. performance on job-person
from resumes Compared optimized fitting within Italian labor
based on job model to unoptimized market
requirements model. Experimented
on real/current data
Heidarysafa et Provide Using the proposed Pave path for temporal and
al. 2021 systematic model, evaluated spatial considerations when
process to 244K job postings analyzing job postings
promote within a 12-month
knowledge period
discovery
within the data
science job
market
Shaikh et al. Classify Compared 3 Provide technique to
(2021) candidates Statistical models automatically and accurately
based on (KNN, SVM, match applicant to lob
information Decision Tree) listings
from their against proposed NLP
resumes model
Yu et al. 2021 Guide research Application Scenario Provide a quantifiable
to connect job leveraging job education-occupation
requirements posting data and alignment metric to
and university course syllabi empower stakeholders to
curricula make more informed
decisions around recruitment

66
 CHAPTER 4: INTRODUCING COMPETENCY-BASED HUMAN RESOURCE

MANAGEMENT AS A KERNEL THEORY FOR

JOB COMPETENCY RESEARCH

As discussed in the previous section, and evidenced by Table 1, prior studies

investigating job descriptions have focused interests on specific skill trends and mechanisms to

efficiently extract information. Furthermore, previous research frameworks have specialized in

providing resources to match candidates with listed postings, as further discussed in Table 2.

Although previous research initiatives have been skill-based and extensive, limitations

surrounding the generalizability of the results have been identified (Lunn et al. 2020; Meyer,

2019). In other words, the specificity of the targeted job demographic has encouraged future

research efforts to explore different career fields and themes to enhance job description analysis

techniques (Lunn et al., 2020). Prior research surrounding frameworks have indicated a paucity

of representation in competency-based job description empirical research (Chouhan and

Srivastava, 2014; Laber and O’Connor, 2000). Considering these deficiencies, cross-disciplinary

approaches for analyzing job descriptions are needed (Yu et al., 2021).

To address this call, we consider a design science paradigm leveraged by Kim et al.

(2005) to develop a cross-disciplinary information technology (IT) artifact. Within design

science, artifacts take many forms. For instance, an IT artifact can be a construct, model,

technological rule, or method applied in the development and use of information systems and

science (Gregor and Hevner, 2013; Hevner et al., 2004). Hevner et al. (2004) define methods as

a mechanism which provides guidance on how to solve or explore problems within a solution

space. To further expand, methods can range from formal equations and algorithms to textual

descriptions of “best practices” (Abbasi et al., 2012; Hevner et al., 2004). Studies prior have

67
leveraged the design science paradigm to develop frameworks to guide research endeavors

(Abbassi et al., 2012; Benjamin et al., 2019; Kim et al., 2005). Accordingly, we were inspired to

develop an IS research framework to enhance thematic job description research efforts.

In the absence of sufficient design guidelines, studies have placed a strong emphasis on

the need for design theories to govern the development process of an IT artifact (Abbasi et al.,

2012; Abbasi and Chen, 2008; Markus et al., 2002). Thus, to achieve developing a cross-

disciplinary framework, we call upon competency-based human resource management (C-

HRM) as a kernel theory to guide the development of our proposed job competency research

framework. In the following subsections, we present an overview of C-HRM and competency

modeling methods which can be used to address the research gaps mentioned previously. This

results in a proposal for an enhanced job description analysis technique. In subsequent sections, a

job description analysis framework based on C-HRM is presented and later evaluated with an

empirical example.

Competency-Based Human Resource Management

Competency-based HRM has become an integral practice within organizations over the

last thirty years, as it has enabled businesses to better structure the needs from its human capital.

Competency models have historically been used to help organizations match employees to jobs

(Carretta, 1992), design improvement to job redesign efforts, recruitment strategies, and career

management systems (United Nations Industrial Development Organization, 2002; Vathanopas,

2007).

To understand competency-based human resource management, it is important to define

what competency is. To accomplish this, we review definitions of competencies throughout

HRM literature. These definitions are summarized within Table 3.

68
 Table 3. Various Definitions of Competency

Previous Studies Industry Definition of Competency

Ratnawat (2015) HRM Characteristics of individuals

comprising of their knowledge,
skills, attitudes, and behavior based
on their beliefs and values, resulting
into superior job performance
Kandula (2013) HRM Quality of being physically,
psychologically, and intellectually
well-qualified to perform a job/role
Gangani et al. (2008) HRM Descriptive tool which identifies
skills, knowledge, and
characteristics of an employee to
effectively perform in a role to meet
business goals/objectives
Boyatzis (2007) HRM Characteristic that is directly related
to an individual’s superior
performance within a job
UNIDCO (2002) Industry Androgynous Skills, knowledge, and attributes
which allow employees to perform
an activity within a specified
function
Hornby and Thomas (1989) HRM Ability to perform functions
associated with management within
the work environment
Rogriguez et al. (2002) HRM A measurable pattern of knowledge,
skills, and behaviors needed of an
employee to do a job
Klemp (1980) Education Characteristic of an individual that
underlies effective work
performance

Taking these definitions into consideration, we derive that competencies refer to themes

that are directly related to an employee’s work performance. To expand, competencies act as an

umbrella which house specific skills, knowledge, and characteristics related to the employee’s

work responsibility, such that enhanced skills lead to better competency comprehension which

then leads to better work performance (Ratnawat, 2015). Within literature, competencies have

been used for a wide range of HRM function such as employee selection, retention initiatives,

and organizational planning (Ratnawat, 2015; Rodriguez, 2002). Thus, leveraging a competency

framework requires a strategic orientation to ensure organizational goals are regarded

69
 Human Resource Competency Models

Throughout HRM literature, approaches to building competencies within an organization

is often associated with frameworks and guidelines to help organizations develop systematic best

practices. Within this research, we first observe Kandula’s (2013) macro competency modeling

technique. Specifically, this approach outlines the need to initiate strategic oversight and

socialization of determined competencies. This specific approach declares any proposed

competency framework should have a strategic tie to the organization. In other words, when

developing macro competencies, it is imperative to take the organization’s vision, mission, and

goals into consideration when determining competencies for its employees. Furthermore, once

the competencies are developed, socializing the results, and gaining insight to future directions

of the established competencies is crucial to successful development. Through this perspective, it

is determined that competency frameworks that have no linkage or weak linkage to company

vision , mission, and goals will contribute very little to the desired results.

Additionally, we consider the “One-Size-Fits-All” competency model discussed by

Mansfied (1996) and Chouhan and Srivastava (2014). There are two additionally approaches

discussed throughout HRM literature (“Single-Job Approach” and “Multiple-Job Approach”);

however, this is considered the fastest and broadest approach to establishing competencies within

an organization (Chouhan and Srivastava, 2014). Industry and HRM professionals seeking a fast

and comprehensive competency modeling technique gravitate toward this approach, as it

encourages defining a set of competencies for a broad range of related jobs (e.g., all data science

jobs) (Chouhan and Srivastava, 2014; Mansfield, 1996). Where the former approach engages

employees directly, this second approach charges a research team to develop competencies using

concepts from books, job descriptions, articles, and existing frameworks within leadership,

70
business, organizational development, and human resource outlets. Where the “Single-Job”

technique is tailored to a specific job within an organization and the “Multiple-Job” focus more

on general competencies (regardless of the job), the “One-Size-Fits-All” approach allows

There are two primary considerations we take away from the “One-Size-Fits-All”

approach. The first is identifying the population for whom the model will apply (e.g., data

science jobs). Secondly, developing an HR application to assist in establishing and analyzing the

proposed competencies is also a common practice within this competency model (Mansfield,

1996). An example of such an HR application would be a set of analysis guidelines to assist in

developing competencies.

To summarize these perspectives, we synthesize the perspectives from Kandula (2013),

Chouhan and Srivastava (2014), and Mansfield (1996) to create our research model. Kandula’s

(2013) perspective is represented at the beginning and end of the model, “Initiate” and

“Socialize”, respectively. The “initiate” component encourages the study of company goals,

mission, and vision, whereas the “socialize” component encourages reflection and

communication of results. Furthermore, the research insights of Chouhan and Srivastava (2014)

and Mansfield (1996) are represented within the second and third section in the figure. The

“establish” component encourages the identification of job population and review of literature,

while the analyze phase reinforces the need to establish analysis guidelines. Subsequent sections

dive deeper into specified methods extrapolated from each of these theories to further develop

our proposed research model.

71
 CHAPTER 5: GENERAL OVERVIEW OF TEXT ANALYSIS METHODS FOR

COMPETENCY RESEARCH

An important first step prior to developing our competency-based job description

research model was understanding recent text analysis methods used within HRM research

(Valecha et al., 2013). Understanding analysis guideline development is an important component

to C-HRM, it is important to review text analysis techniques which could potentially tie into

research frameworks. Where previous sections of this paper have provided insight to published

NLP literature and driving C-HRM theory leading the development of this proposed research

framework, this section specifically focuses on contextualizing the relevancy of word

embedding and text frequency analysis methods to competency-based job description research

initiatives.

Word2Vec Embedding Analyses

Upon reviewing literature surrounding word embedding models (Appendix A), it is

evident Word2Vec was published before the others observed. Subsequent models identified

Word2Vec limitations and further optimized its capabilities to accomplish new tasks such as text

sentiment classification (Shen and Liu, 2021). For instance, Shen and Liu (2021) highlight

significant differences between Word2Vec and the Bidirectional Encoder Representation from

Transformers (BERT) model. Where Word2Vec generates word embeddings that are context-

independent, BERT generates embeddings which allow a word to have multiple vectors because

it takes context into consideration when Word2Vec does not (Shen and Liu, 2021).

However, these newer models developed limitations of their own, despite their enriching

abilities. Although the first of its kind and its limitations, Word2Vec is still considered an

efficient word embedding strategy to investigate similarity relations between specified domain

72
terminologies (competencies) (Goldberg and Levy, 2014; Quan et al. 2018). It continues to grow

in popularity and assist in semantic research endeavors given its minimal memory requirement

and word similarity analysis capabilities (Kai et al., 2019; Gao et al., 2022).

In a historical context, Word2Vec was developed by Tomas Mikolov and his colleagues

to provide a state-of-the-art word embeddings software to assist in understanding the

distribution and composition of words within a corpus (Goldberg and Levy, 2014; Mikolov et al.,

2013). Having identified that NLP systems historically atomized words, Mikolov et al. (2013)

sought to explore the notion of word similarity by developing a tool which vectorized words.

Word vectorization refers to an NLP method which maps words from a corpus to a

corresponding vector of real numbers, which is then used for word similarity or prediction

analysis. Vector representations which are close to one another indicate word similarity

(Mikolov et al., 2013).

Like most NLP methods, Word2Vec heavily relies upon three major steps when

conducting similarity analyses: 1) data preparation, 2) model training, and 3) method analysis.

Data preparation refers to the identification of a data source, collection and cleaning techniques,

and tokenization the words for the model. The second step refers to how the model is trained.

Within this step, model parameters are reviewed and determined based on the needs of the

research objective. Lastly, method analysis refers to the use of a method unique to the model’s

library to conduct an analysis. For the sake of the later mentioned framework, the most_similar()

method is referenced. This tool finds the top-N most similar words that contribute positively

toward similarity. Essentially, the user trains the Word2Vec model on a corpus of words model

from which the user calls upon the similarity method to identify which N words were most

73
similar to a designated word (or competency). Figure 1 summarizes the word embedding steps

for the Word2Vec model.

Considering the aforementioned points, we determined understanding the context of a

competency exceeds the interests of this research endeavor. The proposed framework seeks to

create dictionaries of words that are similar to established competencies. Specific context of the

phrase and/or competency is irrelevant and exceeds the interests of this endeavor. That said, it

could be a future direction of this research endeavor. Given the prominent nature of Word2Vec

for NLP tasks (Adewumi et al., 2021), we proceeded with this being our primary model within

the proposed framework.

Figure 1. Summarized Steps of Word2Vec

74
 Text Frequency Analyses

In addition to word embeddings, text frequency is another common method throughout

job description research. Where word embeddings help identify prominent words throughout a

corpus, text frequency analyses provide an additional empirical metric for measuring importance.

When studying competencies within job descriptions, it is crucial to measure the degree of

importance surrounding each established competency (Tang et al., 2022). Job descriptions

developed by HRM professionals reflect the needs of the organization and associated

competency qualities that are in great demand through the enterprise, which may be repeatedly

mentioned within job descriptions. Therefore, Term Frequency-Inverse Document Frequency

(TF-IDF) is traditionally used to measure word importance within a corpus.

Per Qaiser and Ali (2018), the total length of documents can vary, which leads to the

possibility that any term may occur more frequently in longer word-count documents when

compared to smaller documents. Thus, the occurrence of any term in a document is divided by

the total terms present in that document. This division calculates the actual term frequency.

Regarding inverse document frequency (IDF), this component acknowledges that not all

words can be treated equally, given that each keyword has a different importance level (Qaiser

and Ali, 2018). Thus, IDF assigns lower weights to frequent words and assigns greater weights to

words that are not as frequent. TF-IDF score is simply the multiplication of term frequency (TF)

with the inverse document frequency (IDF).

75
 CHAPTER 6: THE JOB COMPETENCY RESEARCH FRAMEWORK

DEVELOPMENT

With the help of C-HRM perspectives of Kandula (2013), Chouhan and Srivastava

(2014), and Mansfield (1996), we have identified three requirements for our research model,

namely initiating competency development, establishing and analyzing competencies, and

socializing the findings. In this section, we describe the components of our proposed framework

by describing them in detail.

Initiating Competency Development

The first section of this model involves initiating job description research. If a research

project wishes to succeed, it must be directly connected to a goal established by the research

party, such that there is a clear link in research intentions and problem-solving initiatives

(Probert et al., 2003). Within competency development, a weak linkage will not contribute to the

desired results (Kandula, 2013). Therefore, the first section of this framework encourages the

initiation of research by establishing research goals to guide analysis. This can be achieved by

developing research questions, highlighting problematic trends, or developing hypotheses to

explore. Once researchers have developed their research goals, establishing, and analyzing the

competencies can commence.

Establish Competencies

Keeping the research goal in mind, we consider Mansfield’s (1996) and Chouhan and

Srivastava’s (2014) competency collection techniques: 1) Single-Job Competencies, 2) One-

Size-Fits-All Competencies, and 3) Multiple-Job Competencies. Within the proposed research

model, we have adopted the “One-Size-Fits-All” modeling technique. Justification as to why we

did not include the first or third modeling technique can be found in the next paragraphs.

76
 As defined in previous sections, the “Single-Job Competency Model” is an approach

which leverages employees and management input to derive competencies. Although engaging,

this approach is considered costly with regards to time and money, especially during the data

collection phase. Interviewing and/or surveying stakeholders (employees, customers, and

leadership), especially if participation is incentivized, costs the organization time and money

when taking the initial steps to developing competencies. Furthermore, once the data has been

collected, it must be analyzed and derive common themes throughout the stakeholder feedback.

Furthermore, it is not uncommon for organizations to consult with third-party professionals to

validate identified competencies.

Regarding the “Multiple-Job Model”, this method is traditionally used when the jobs in

question have nothing in common. For example, if an organization has data science and human

resource management vacancies. Therefore, this this method relies on the construction of general

“building block” competencies to facilitate matching individuals to jobs (Mansfield, 1996).

“Building block” competencies tend to reflect wholistic non-technical skills that are universal,

each with a unique definition and ascribed behaviors detailing how an employee can demonstrate

proficiency. Where the other two approaches focus either on a specific job or category of jobs,

the Multiple-Job Approach proposes the development of universal competencies with job-

specific expectations. For instance, if an organization determines communication is a “building

block” competency, the job-specific expectations to demonstrate proficiency may differ for data

scientists and human resource employees. Understanding this method is job agnostic, it falls out

of our scope of framework research interest. The intent of the proposed framework is to help

academic and industry professionals research and better understand a career field.

77
 As mentioned in section 3.2., Chouhan and Srivastava’s (2014) and Mansfield’s (1996)

second approach is known as the “One-Size-Fits-All” competency model. Considering the

objective of the proposed research framework is to arm both academic and industry researchers

with a systematic approach to thematically exploring job postings for a specified career field

(e.g. data scientists), we found the “Once-Size-Fits-All” methodology to be the most cost

effective and efficient way to derive competencies. That said, there are limitations to choosing

this model. For instance, it is heuristic in nature. Competencies may fluctuate or change based

on the literature, job descriptions, and resources the research team reviews may limit consistency

throughout literature. However, the inconsistencies could provide a boon to the academic and

industrial community because it opens an opportunity for rich discussions and perspective to be

shared. Using the “One-Size-Fits-All” approach, established competencies are backed by

published sources. Thus, creating a rich collection of scholarly insight on a designated job

category. Thus, the initiation component of the model adopts the second “One-Size-Fits-All”

model.

Analyze Competencies

Once the competencies have been established, analyzing the proposed competencies

follows. These three-phases consist of data preparation, model training, and similarity analysis

(see Figure 1). The following subsections provide deeper insight into each component.

Data Preparation

The preparation phase is the backbone of the analysis. Within this phase it is imperative

to establish where the job description data is going to come from, how the researchers plan to

extract the data, and what measures the research team will take to clean and prepare the data for

tokenization.

78
 With regards to where the data will come from, scholars have identified a wealth of

sources to acquire job information. Namely, websites that allow applicants to search keywords

and set personalized search parameters such as LinkedIn, Indeed, and GlassDoor have grown in

popularity when conducting job description research (Christensen et al., 2005; Gupta, 2021;

Hosain and Liu, 2020; Landers et al., 2019; Marinescu et al., 2020; McCabe, 2017; Muthyala et

al., 2017). Thus, identifying a source to conduct research is dependent on how the research team

plans to extract the needed data. There are open-source data depots like GitHub or Kaggle which

allow scholars to share data for other research endeavors. Additionally, there is also the option

of scraping data directly from the desired websites. Although, both techniques come with

benefits and concerns. With regards to open-source data, the research team is trusting the source

to have not tampered with or defame the data. Although it is the easiest option to acquire job

description information, researchers are faced with the assumption that it is an honest

representation of what was pulled. Regarding web scraping, many websites have policies and

measures which prevent machine learning techniques from extracting information from them.

Furthermore, memberships and money-walls are also a factor as some companies monetize the

data they are responsible for. There is the option of reaching out directly to a desired company

and acquiring a developer API token; however, not all job posting websites offer this option.

Although web scraping may yield the most data, it faces many challenges to acquire it.

The last consideration of this phase is the preprocessing and tokenization component.

Assuming the data source has been selected and job postings have been collected, it is imperative

to prepare the data for the Word2Vec model. This requires data cleansing such as eliminating

null data and duplicate entries. Furthermore, leveraging the Natural Language Took Kit (NLTK)

python library to efficiently prepare text data for analysis has proven to be an effective tool for

79
NLP analysis (Millstein, 2020; Mohan, 2015). The NLTK library is a robust and renowned tool

within the NLP community that contains a variety of functions to help clean. This research

model proposes the use of regular expressions to conduct our initial preprocessing and the

established stopwords, PortStemmer, and WordNetLemmatizer within NLTK (Mohan, 2015;

Wisdom and Gupta, 2016). The objective of cleaning this information is to remove noisy data

from the job postings to generate key meaningful words (tokens). By doing this, the Word2Vec

model used in phase 2 can better vectorize clear word associations, given our established

competencies (Ma and Zhang, 2015; Vijayarani et al., 2015).

Additionally, we leveraged an online tutoring platform called GeeksforGeeks, which

hosts free tutorials, online courses, coding competitions, and webinars by industry experts to

assist in developing our preprocessing mechanisms.

Table 4 summarizes our preprocessing steps and associated literature and sources. To

effectively preprocess the information, we leveraged published articles and books to guide the

development of our preprocessing.

Table 4. Data Pre-Processing Steps

Pre-Processing Step Example References

Remove newlines and !.,?“:; Embarak, 2018

punctuation
Remove URLs www., https:, .com, .edu, GeeksforGeeks,
2018;López and Romero,
2014
Remove Numbers 1,2,3,4,5, 21234 López and Romero, 2014;
Rajagopalan, 2021
Remove leading and “ word.” | “word .” Embarak, 2018
training whitespaces
Remove Unicode ☻�○êçñ López and Romero, 2014;
characters Rajagopalan, 2021
Tokenize Input “computers”, “as”, “an”, Vijayarani and Janani,
“example” 2016
Remove Stop words “computers”, “as”, “an”, GeeksforGeeks, 2018;
from tokens “example” Hardeniya et al., 2016

80
 Table 4. Data Pre-Processing Steps Continued
Stem Words (strip end of “comput”, “exampl” Anandarajan et al., 2018
word)
Lemmatize Words “computer”, “example” Anandarajan et al., 2018
(identify root word)

Model Training
The once the tokenized words have been established, training Word2Vec model within

the Genism library (Ozkur et al., 2022) becomes essential in developing dictionaries of similar

words based on the established competencies. Tuning the parameters of the model become

imperative within this component, as the wrong combination of hyper-parameters are known to

produce poor quality vectors (Adewumi et al., 2021). Thus, reviewing the number of vector

directions (size), context word window size (window), minimum word count (min_count), and

training algorithm (sg) are a few of many optional parameters to take into consideration when

tuning a model for research.

Model Analysis

By using the Word2Vec.most_similar method, a targeted word (competency) can be

inputted, and the function will return the 10 closest words from the target word (Kurnia, 2020).

This function establishes the competency dictionaries required for thematic analysis.

Understanding the tokens have been stemmed and lemmatized, the input words must reflect a

similar format. Once the dictionaries have been organized and established, the analysis phase can

commence. Within this proposed research framework, we observe Term Frequency-Inverse

Document Frequency (TF-IDF) as the analysis technique. Empirical in nature, TF-IDF is a

prominent and commonly used term weighing methodology which provides indication to word

importance (Chen et al., 2016; Jalilifard et al., 2021). TF-IDF scores range between zero and

one. If the word or dictionary receives a score that is closer to zero, this indicates it is not

81
relevant within the document. On the other hand, if the word or dictionary receives a score that is

closer to one, it is considered more important/relevant.

Within this framework, each established competency dictionary will have a score

between zero and one within each job posting. Once all TF-IDF scores have been established for

each dictionary against each job posting, an average can be taken to better understand which

competencies are more prevalent. However, Post-Hoc statistical testing is imperative to confirm

differences which may appear, Phase 4 provides a space to interpret the results of the TF-IDF

analysis and conduct post-hoc statistical tests.

Socialization

The last phase of this proposed research model enables researchers to interpret and

communicate the results of the TF-IDF competency analysis. By investigating the calculated

averages of TF-IDF scores, researchers can determine which competency dictionaries were more

prevalent. Furthermore, statistical testing and means comparisons can be conducted to confirm

and validate findings. For instance, ANOVA testing can be used for comparing group means and

confirm differences which may exist (Park, 2009). Additionally, if comparing two data sources, a

two-tailed test can be leveraged in determining significance as it challenges the null hypotheses

that the datasets are similar (UCLA: Statistical Consulting Group, n.d.). Once all statistical

considerations have been considered, researcher can then review the results of their analysis and

report their findings to enrich the existing pool of literature. Although, the socialize component

goes beyond communicating the results of the analysis. Reflecting on limitations and future

directions of research is crucial in guiding new developments to continuously improve upon, as

research is an iterative process.

82
 CHAPTER 7: THE JC-COMPASS FRAMEWORK FOR JOB DESCRIPTION

RESEARCH

The purpose of this section is to tie theory and analysis components together to establish

a research framework. Due to the variety of ways to analyze job descriptions, a standardized and

cross-disciplinary research framework for conducting competency-based job description research

would be of great value for industry and academic professionals. The Job Competency Initiate,

Establish, Analyze, and Socialize (JC-Compass) framework described here serves to guide

researchers through their own exploration of job descriptions given a research objective. This

framework leverages the research theory which combines C-HRM and NLP practices and is

intended to help professionals see a project through its entire life cycle when using Word2Vec

modeling analysis techniques (Figure 2).

Figure 2. The JC-Compass Framework

There are four primary sections of this cross-disciplinary model. The first is identified as

initiation, which refers to the creation of research purpose. This can be achieved via creating

research goals, identifying a research problem, and/or developing research questions regarding a

83
designated phenomenon. The second section of this research framework is the establishing of

competencies. This is achieved by targeting a specific job demographic (human resource

managers, data scientists, etc.) and reviewing existing frameworks, literature, and guidance

surrounding the targeted demographic to establish competencies. Once the competencies are

created, the framework shifts to the analysis phase where the researchers are directed extensively

analyze the competencies via reviewing job descriptions. Within this phase, data preparation

commences. This refers to the identification job description data sources, collection, cleaning,

and tokenization of words. Additionally, the Word2Vec model is trained on the preprocessed

words (tokens) to enable the last component, similarity analysis. The method analysis enables the

development of competency dictionaries to conduct further analyses. Lastly, the socialization

component is comprised of communicating the results of the analyses and statistical comparisons

encourages the while the second is methodology. Given this framework marries C-HRM theory

with NLP practice, we observe the NLP components are embedded within the competency

dictionary development and job description analysis. The initiation section ignites the research

project, placing the responsibility of establishing and reviewing research goals and developing

job competencies on the scholar. The methodology section is broken into four phases: 1)

preparation, organization, analysis, and socialization (Meyer, 2019). This section is used to guide

competency-based job description analysis via Python libraries. Subsequent sections discuss

these sections and phases in further detail.

84
 CHAPTER 8: EMPIRICAL DEMONSTRATION OF JC-COMPASS

The purpose of this chapter is to empirically demonstrate the capabilities of the

framework and dictionaries. This chapter is broken down into four sections. The first section

discusses the initiation phase of this example. The second section covers the establishment of

competencies to guide the research. The third section discusses the analysis methodology.

Lastly, the final section discuses the results and socializes the findings.

Initiate Competency Research

To evaluate our job description analysis framework, we demonstrate the value of the

proposed research guidelines by conducting a comparative competency analysis on archival job

description data from the private sector and United States Federal government. The premise of

this example serves two objectives. The first objective is to demonstrate how this framework can

be used when working with archival data. The second is to explore how the United States

Federal Government thematically differs from the private sector when communicating job

requirements for data science vacancies. The United States Bureau of Labor Statistics projects a

36% growth of the data science career field between 2021 and 2031, which is much faster than

the average for all other occupations within the federal and private sector. This evolution has

consequently affected how human resource professionals manage the necessary skill demands

that come with the evolution of technology and work environments. Historically, the federal

sector relies heavily on contracted data science personnel rather than full-time staff, which has

stemmed the Office of Personnel Management (OPM) to derive a new data science job series to

better recruit more fulltime data science personnel. This shift in interest consequently creates a

competitive environment for recruiting top-tier talent within the data science field. Thus, we seek

to fortify recent research initiatives by diverting from the trend of analyzing specific skills

85
desired of data scientists and investigate how the federal and private sector communicate their

respective desires via natural language processing techniques to establish nomenclature and

lexicons related to data science competencies derived from theory and industry practice. In doing

so, this example seeks to showcase the proposed research model by educating the industrial and

academic community on how the private and federal sector thematically communicate their

needs from data scientists rather than just analyze desire skills. By utilizing our proposed

research model, this example draws attention to thematic differences within job descriptions,

which in turn could bolster recruitment strategies such that HRM professionals can remain agile

when faced with competition.

Establish Competencies

As newer AI and ML technologies penetrate industries and data science jobs worldwide,

securing the right talent to manage and develop these tools has become a premium for

organizations within the private and federal sectors. Renowned organizations like the National

Security Agency, Google, IBM, and the Department of Homeland Security leverage data

scientists to conduct predictive analysis on various complex problems that challenge the

organization's success and national security. That said, establishing job competencies has

become a necessary tool for human resource managers across the federal and private sectors

(Gangani et al., 2006; Hattingh et al., 2019; Rodriguez et al., 2002). As the data science career

field has evolved, newer technologies and requirements have presented new responsibilities for

data scientists. This has created a multitude of data science frameworks and varying

competencies across the industry. To guide this research example, we leverage the practices of

the aforementioned “One-Size-Fits-All” C-HRM strategy and synthesize two different

86
competency frameworks published by Deloitte and IBM to propose 5 core competencies related

to data science and artificial intelligence.

Deloitte's "Trustworthy AI™" multidimensional AI framework postures organizations to

develop ethical safeguards across six key dimensions (Ammanath, 2022; Mökander and Floridi,

2021). The six key dimensions of this framework are as follows: 1) transparent and explainable

AI, 2) fair and impartial, 3) robust and reliable, 4) privacy, 5) safe and secure, and 6) responsible

and accountable AI. Understanding ethical practices of data science are imperative to building a

successful organization. We extracted two core competencies from this Deloitte framework to

gauge how job descriptions solicit skills. Specifically, this framework addresses the unique

capabilities of these intelligent technologies and proposes ways to safeguard technology

advancements from committing unfair or unsafe practices. Thus, the first competency we

extracted for our data competency framework is ethics. This competency encompasses the social

responsibilities and skills necessary to uphold within the work environment. The framework

explicitly addresses the unique capabilities of these smarter technologies and proposes ways to

safeguard technology advancements from committing unfair or unsafe practices. Secondly, we

extracted "Artificial Intelligence" as a core competency, as this framework specifically addresses

the unique capabilities of these intelligent technologies and proposes ways to safeguard

technology advancements from committing unfair or unsafe practices. Artificial Intelligence (AI)

is an umbrella term encompassing various topics related to machine vision, NLP,

recommendation engines, and deep learning techniques (to name a few) (Votto et al., 2021).

Incorporating it as a data science competency within our model ensures that sub-interests related

to AI have the opportunity for equal representation.

87
 Figure 3. Adopted Data Science Competencies

The second core framework we leverage to build our competency network stems from

IBM's Data Science Skills Competency Model. Within this framework, there are 28 critical

competencies broken into seven groups. These competencies hold data scientists accountable

within the work environment. These key competencies range from problem-solving to

understanding analytical cycles (Gottipati et al., 2021; IBM, 2020). We extracted three additional

primary competencies from this framework to guide our dictionary development. The first

competency extracted from this framework is "machine learning," which relates to employees

demonstrating they understand machine learning algorithms' principles and mathematical

components. The second competency extracted from this framework is "problem-solving," which

refers to an employee's ability to identify and characterize business problems and develop

solutions. The last component we extracted from this framework was an employee's statistical

capabilities, otherwise denoted as "statistics" within our model. The statistics component refers

to an employee's capability to understand sampling, probability theory, and descriptive statistical

concepts (IBM, 2020).

AI is in the narrative leading into skill development. For the sake of this research

example, IBM's framework alone is limited. Although it addresses the rapid growth of AI

88
capabilities and required responsibilities, it does not mention "ethics" within the document, nor

does it include artificial intelligence within the established competencies. Therefore, we leverage

the Deloitte framework to fill in these gaps and propose five primary competencies which guide

this research: data, ethics, problem-solving, and machine learning (Figure 3). These

competencies will later be the foundation for building term dictionaries to run term-frequency

inverse document frequency (TF-IDF) analysis.

AI is in the narrative leading into skill development. For the sake of this research

example, IBM's framework alone is limited. Although it addresses the rapid growth of AI

capabilities and required responsibilities, it does not mention "ethics" within the document, nor

does it include artificial intelligence within the established competencies. Therefore, we leverage

the Deloitte framework to fill in these gaps and propose five primary competencies which guide

this research: data, ethics, problem-solving, and machine learning (Figure 3). These

competencies will later be the foundation for building term dictionaries to run term-frequency

inverse document frequency (TF-IDF) analysis.

Data Science Competencies within the Federal Government and Private Sector

Technological capabilities have allowed for developing new and innovative data science

tools to streamline once tedious processes to simple key-stroke actions. The US federal

government and private sector have experienced growth spurts within new data science

capabilities, requiring new and innovative talent to manage these tools. Regarding the federal

government, institutional barriers regarding training and awareness limit the implementation and

acceptance of these newer technologies (Kreuger et al., 2019). By reinforcing systematic training

requirements and hiring standards, the US federal government has successfully developed

reskilling opportunities related to machine learning, artificial intelligence, and data science to

89
adopt AI (Tarraf et al., 2019). Furthermore, literature surrounding smarter IT and the federal

government reinforces the need for strong-minded Chief Information Officers to enhance the

strategic role of IT within government initiatives, especially as near-peer adversaries and peer

competitors tirelessly strive toward a competitive advantage (Khallaf and Majadalawieh, 2012;

Peppard, 2007; Wagemann Jr., 2020). Although ambitious, there is a delicate balance the federal

government must adhere to the allocated budget. ‘Doing more with less’ initiatives promote

internal data and statistical upskilling opportunities to respond to the growing need for data

scientists. However, these initiatives have also created struggles within decision-making

processes, given the complex political, managerial, and democratic changes which fluctuate

throughout the government (Janessen and Estevez, 2013; Tarraf et al., 2019). Thus, the Office of

Personnel Management (OPM) renovated its data science occupational series, which outlines the

need for multifaceted data scientists with an interdisciplinary background. Through these

guidelines, the federal government identifies a broad and general range of necessary skills that

need identification when considering talent, emphasizing skills directly related to machine

learning and statistics (OPM, 2021). Through this renovated definition of data science

requirements, upskilling initiatives to grow data scientists have ignited, allowing existing

employees to apply for these opportunities and return to their work centers with a deeper

understanding of data science tools and techniques (Heckman, 2021).

In contrast, the private (for-profit) sector has much smoother decision-making processes,

given the general objective of selling products or services to consumers to generate stakeholder

wealth (Nutt, 2005). With the increasing challenges of collecting, transforming, and analyzing

big data, the private sector seeks to recruit flexible, agile, and digitally skilled individuals with

strong AI and ML backgrounds. They are needed to oversee capabilities (Sart and Yildiz, 2022).

90
Thus, investing in employees with experience in machine learning and artificial intelligence

enhances the development of emerging markets within the private sector (Strusani and

Houngbonon, 2019).

Lastly, the federal and private sectors expect ethical practices and a fundamental

understanding of data. However, ethical issues surrounding privacy, data validity, fairness, bias,

and transparency are not the only issue for the federal or private sector (Ammanath, 2022; Egger

et al., 2022). Both sectors share a responsibility to maintain ethical practices, if not to do the

right thing but also to save face from the public eye, given the media’s capabilities (Ouchchy et

al., 2020). Furthermore, new mining and data collection capabilities emphasize that applicants

understand they need to have experience handling pertinent data within the organization

ethically. Regarding this research, ethical competency refers to the sector’s ability to

communicate skills related to ethical practices of data science.

Analyze Competencies

We analyze the content of job postings from both the United States federal government

and the private sector using our proposed research model. The first phase is the preparation phase

which encompasses data collection methodologies and data processing. The second phase of this

approach consists of organizing the data. Within this phase, we leverage the results from the

preparation phase to build term-similarity dictionaries to guide our analysis. The third phase is

the analysis phase which encompasses the conduction of a term-frequency inverse document

frequency (TF-IDF) analysis. Within this step, we analyzed the job postings' TF-IDF score and

conducted ANOVA and 2-Tail T-Test analysis.

91
 Data Collection

Regarding data collection, we leveraged two data sources to acquire our job posting data

(Kaggle and USAJobs.gov). We collected two archival data sets to train our model and build our

dictionaries. Regarding our private sector training data, we leveraged a public and archival

dataset from Kaggle (Nasar et al., 2021) which yielded 3,567 unique data science job

descriptions from 2021 and 2022, upon cleaning the data set (removing duplicates and null data).

For our federal job postings, we acquired permission and tokens from USAJobs.gov to

pull data science job postings directly from the website. The privileges bestowed upon us

allowed us to pull archival job postings listed in 2021 and 2022. We omitted any positions

currently open to ensure both datasets were archival. With this query, we received 3,086 job

postings and descriptions directly related to data science. A total of 6,653 job descriptions were

used within this analysis of federal and private sector data science job postings. Table 5

summarizes our cleaning procedure for the datasets.

Understanding the date quantity differences between the federal and private sector

datasets, we reviewed analyzed the uniqueness of tokens for each dataset and discuss the results

within the Cleaning and Tokenization section of this paper (see Table 7).

Descriptive Statistics

Before diving into the various phases of our methodology, we showcase our descriptive

statistics for the federal and private sector datasets. Within our descriptive statistics, we

showcase the number of job descriptions collected for each year summarized within Table 5.

Given the interests of this research, we specifically focused on the job description variable, as the

scraped from GlassDoor did not provide location or usable variables for further analysis.

92
 Table 5. Token Comparison

2022

Private Sector 1,494

Federal Sector 649
Total 2,153

Cleaning and Tokenization

Once our data was acquired, we then preprocessed and tokenized the information in

preparation to train and test the natural language processing (NLP) neural network.

Upon tokenizing the job postings, we reviewed how similar the datasets were to gauge

relevance of comparison. Given the larger sample size, the private sector identified 5,842 tokens

with regard to the job descriptions. The federal sector yielded 1,916 total. Although the private

sector had more unique tokens, the federal sector shared 1,768 of the tokens. Thus, the federal

sector was approximately 93% similar to the private sector data. Table 6 summarizes the numeric

similarity and dissimilarity of each data set. Given the strong similarity score of the federal

sector to the private sector, we proceeded with our analysis.

Table 6. Data Similarity Comparison

Total Token Number Unique Number Shared Percentage Similar
Count Tokens Tokens

Private Sector 5,842 4,074 1,768 --

Federal Sector 1,916 148 1,768 92.27%

Model Training

Upon preprocessing, tokenizing, and validating the similarity of the data, we trained a

Word2Vec model for the private and federal sectors to identify similar words in a text. This

unsupervised neural network is widely used in NLP for generating word embeddings (Htait et al.,

2018). Within these two models, the similarity is the distance between keywords, otherwise

93
known as the cosine similarity score. Following Valecha et al.’s (2021) misinformation analysis,

we utilized the skip-gram algorithm, which predictively learned the word embeddings; this is

numeric vector representation. We also trained the algorithm on a window of two, such that the

context spans over two words to the left and two to the right of the target word during training

(Valecha et al., 2021). To avoid overfitting our model, we used the default model setting for the

number of times a word should appear for it to be considered during the training process (five

times). Once our models are trained, the preparation phase of our analysis ends.

Method Analysis

We proceeded to organize our word embeddings to posture for a TF-IDF analysis. Within

the organization phase, one primary step is establishing our dictionaries. To accomplish this, we

continue to use the NLTK library to call upon the most_similar() function to extract similar

words related to our core competencies: ethics, artificial intelligence, machine learning, problem-

solving, and statistics (Poomagal et al., 2022). Our search queries needed to contain stemmed

and lemmatized words to accomplish this. Thus Table 7 summarizes our search queries within

the most_similar() function.

Table 7. Most_Similiar Queries

Data Science Stemmed/Lemmatized

Competency Query
Ethics ‘ethic’
Machine Learning ‘machin’, ‘learn’
Problem Solving ‘problem’, ‘solv’
Artificial Intelligence ‘artifici’, ‘intellig’
Statistics ‘statist’

Upon running the most_similar() function, we generated the top 10 most similar words

related to our established data science competencies (Martinez Soriano et al., 2019). Once we

acquired all similar words for each competency, we built two comprehensive dictionaries, one

94
for the federal sector and the other for the private sector. To provide further insight, Figures 4a-b

showcase separate graphs which summarize the most similar words for the designated data

science competency. Each grouping is unique and does not share a similar word. The left graph

showcases the top 10 words, most like “ethic,” for both the federal and private sectors.

Furthermore, the data points on each graph represent the cosine similarity score assigned

to each word from the Word2Vec model. However, the graph on the right showcases the top 10

words, most like “machine learning,” of which the private and federal sectors share “artificial”.

This is the only competency in which both datasets and models share similar words. To provide

insight into each dictionary, we have included figures within Appendix B and C which

summarize the comprehensive dictionary for the federal and the private sector. Duplicate words

were eliminated from the dictionaries to eradicate redundancy.

Ethics Dictionary Comparison Machine Learning Dictionary

Cosine Similarity Comparison
Private Sector Cosine Similarity Score Consine Similarity
Federal Sector Cosine Similarity Score Private Sector Cosine Similarity Score

Federal Sector Cosine Similarity Score

exhibit
cognit 1 uphold
stringent 0.8 style unsupervis
screen 0.6 urgenc clean1 reinforc
0.4 surfac artifici
materi 0.2
credibl access 0.5 multimod
elit 0 attitud contribut causal
0
suitabl meticul draw convolut
align demeanor lab neural
intervent motiv forc infer
oak keen imageri theori
administ vector

Figure 4a. Most_Similar Comparison

95
 Artificial Intelligence Dictionary Problem Solving Dictionary
Comparison Comparison
Cosine Similarity Cosine Similarity
Private Sector Cosine Similarity
Private Sector Cosine Similarity Score Score
Federal Sector Cosine
Federal Sector Cosine Similarity Score Similarity Score

cognit pragmat
nuclear1 augment valid1 solver
spatial ai overse break
vehicl 0.5 fusion concept 0.5 diagnos
creat chip qualit difficult
visual 0 multimod substant 0 conceptu
dollar exploit trial frame
access signal propos thinker
imageri tree experiment dig
compens audio diver intuit
threat dissemin

Statistics Dictionary
Comparison
Cosine Similarity
Private Sector Cosine
Similarity Score

econometr
content1 theori
semest inferenti
econom 0.5 causal
thesi exploratori
differenti 0 actuari
substanti mathemat
comput theoret
logic
probabl algebra
correl

Figure 4b. Most_Similar Comparison

Once we established the dictionaries, we leveraged MAXQDA to run our quantitative

content analysis. MAXQDA is a powerful software package for analyzing qualitative data and

conducting content analysis (Oliveira et al., 2015). Because this suite interfaces with Windows-

based software and programs, we were capable of uploading our .CSV files for each of our

datasets once we converted them to .XLSX (excel worksheet). Utilizing the dictionary function

96
within MAXQDA, we uploaded our two dictionaries into the software to run two separate

content analyses, one for our federal data and the other for our private sector data. The results

from the MAXQDA content analysis yielded an excel document with the word count of each job

description and the number of words that appeared within the job description relative to the

dictionary for each job posting. Upon receiving this excel document from MAXQDA, we added

the two columns from the master document which portray the location of the job posting as well

as the number of days the position was open for application. To minimize error, we organized the

job postings by year posted and assigned dummy variables to each job description. From here,

we numerically ordered each description for the MAXQDA excel document and the master

document to ensure the locations were with the correct job posting. We then conducted our TF-

IDF analysis. By examining the TF-IDF analysis, we can compare how the federal and private

sectors differ regarding leveraging the established data science competencies (Figure 3) within

their job postings.

We further coded the job postings to indicate the presence of each competency. For

instance, if the job posting contained terms related to the AI Dictionary that the Word2Vec

model produced, we coded it with “1”, otherwise it was coded “0”. The next section discusses

how we validated the measurements of each dictionary.

Validation of Machine Coding

We based our machine coding validation on strategies and systematic methods used by

Valecha et al. (2021a), Larsen et al. (2020), Wang et al. (2018), and Gonzalez and Sol (2012).

Specifically, we performed a preliminary validation on our coding mechanism to detect any of

the data science competencies within the job posting. This process was performed in two steps:

1) having two information technology (IT) experts manually code randomly selected job postings

97
for both the private and federal sector, and (2) correlate the human coding with the machine’s

measures for the random sample.

We set up a preliminary briefing to manage expectations. The research team met with

both IT experts simultaneously to discuss the rules of engagement for coding these job postings.

It was communicated that the work was to be completed independently with no collaboration.

We then discussed definitions for each competency to confirm an understanding and allow the

expert to ask any questions. Once the meeting concluded, both experts were separated and

proceeded to code. Regarding the experts, they confirmed to have over 9 years’ experience

within the IT field, have had exposure to data science, and have both held managerial/hiring

positions.

Regarding step 1, we sought to have the IT experts code job postings in 3 batches. The

job postings within the first 3 batches were identical. Given that this is a systematic process, our

goal was to repeat the coding process until we achieved a total kappa value greater than the .7

threshold (Valecha et al., 2021a; McHugh, 2012; Landis and Koch, 1977). For each batch, we

provided 30 random and identical job postings to each expert. 15 of the thirty were private sector

job postings, while the remaining were federal. Each batch provided a new set of random job

postings which were shared to the coders. The experts reviewed the job postings and looked for

cues relating to ethics, artificial intelligence, machine learning, problem solving, and statistics.

The experts coded the following: “0” in the absence of information about that competency, and

‘1’ otherwise. We compared the coding mechanisms of the validators against one another and

their performance against the machine. The kappa results of step 1 are summarized in tables 8a-

b and 9a-b.

98
 Table 8a. Results of Inter-Coder Reliability (Private Sector)
PRIVATE Batch 1 (N=15) Batch 2 Batch 3 Total
SECTOR (N=15) (N=15) (N=45)

Total 77% 85% 96% 86%

Ethics 73% 80% 87% 80%
Machine Learning 87% 86% 100% 91%
Artificial Intelligence 73% 80% 100% 84%
Problem Solving 73% 87% 93% 84%
Statistics 80% 93% 100% 91%

Table 8b. Results of Inter-Coder Reliability (Federal Sector)

FEDERAL Batch 1 (N=15) Batch 2 Batch 3 Total
SECTOR (N=15) (N=15) (N=45)

Total 77% 79% 89% 82%

Ethics 73% 80% 100% 84%
Machine Learning 73% 87% 87% 82%
Artificial Intelligence 83% 60% 93% 79%
Problem Solving 77% 93% 80% 83%
Statistics 80% 73% 87% 80%

Table 9a. Results of Preliminary Validation (Private Sector)

PRIVATE Batch 1 (N=15) Batch 2 Batch 3 Total
SECTOR (N=15) (N=15) (N=90)

Total 60% 73% 81% 71%

Ethics 60% 66% 87% 71%
Machine Learning 60% 73% 80% 71%
Artificial Intelligence 53% 74% 87% 71%
Problem Solving 67% 80% 73% 73%
Statistics 60% 73% 80% 71%

Table 9b. Results of Preliminary Validation (Federal Sector)

FEDERAL Batch 1 (N=30) Batch 2 Batch 3 Total
SECTOR (N=30) (N=30) (N=90)

99
 Table 9b. Results of Preliminary Validation (Federal Sector) Continued
Total 58% 73% 81% 71%
Ethics 60% 73% 80% 71%
Machine Learning 67% 73% 73% 71%
Artificial Intelligence 53% 73% 87% 71%
Problem Solving 60% 67% 87% 71%
Statistics 50% 80% 80% 70%

Step 2 consisted of comparing each validator alone to the machine. We distributed a

fourth batch of job postings where were random and unique to the validator. This step is unique

from the others as we sought to compare the machine’s coding to that of the human experts. The

research team administered 30 random job postings to each expert (15 federal and 15 private

sector). This step did ensure that the job postings were not identical. This means each expert got

a different set of randomly selected job postings to review. Within Batch 4 we considered

percent agreement and Cohen’s Kappa. Higher kappa scores indicate an agreed understanding

between validators (Valecha et al, 2021a; McHugh, 2012; Landis and Koch, 1977). Table 10a-b

summarizes the kappa findings.

Table 10a. Results of Batch 4 (Validator 1)

Batch 4 VALIDATOR 1 VS MACHINE Private Sector
Dictionaries (N=15)
Percent Match Cohen's Kappa
Total 84% 68%
Ethics 87% 67%
Machine Learning 93% 87%
AI 87% 67%
Prob Solv 80% 60%
Stat 73% 61%
Batch 4 VALIDATOR 1 VS MACHINE Federal Sector
Dictionaries (N=15)
Percent Match Cohen's Kappa
Total 87% 73%
Ethics 87% 73%
Machine Learning 93% 86%
AI 80% 70%

100
 Table 10a. Results of Batch 4 (Validator 1)
Prob Solv 87% 70%
Stat 87% 66%

Table 10b. Results of Batch 4 (Validator 2)

VALIDATOR 2 VS MACHINE Private Sector
Dictionaries
Percent Match Cohen's Kappa
Total 89% 76%
Ethics 87% 72%
Machine Learning 87% 66%
AI 93% 87%
Prob Solv 87% 71%
Stat 93% 87%

VALIDATOR 2 VS MACHINE Federal Sector

Dictionaries
Percent Match Cohen's Kappa
Total 88% 78%
Ethics 93% 86%
Machine Learning 87% 62%
AI 87% 60%
Prob Solv 87% 100%
Stat 87% 84%

TF-IDF Analysis of Competencies

Following validation, we then conducted our TF-IDF analysis. By examining the TF-IDF

analysis, we can compare how the federal and private sectors differ regarding leveraging the

established data science competencies (Figure 3) within the job postings. To create a robust

analysis, we incorporated synonyms for each term the machine output (Bafna et al. 2016; Akay

et al., 2014). The dictionary of each competency and the associated synonyms can be located in

Appendix D and E.

Descriptive Statistics

Tables 11a-b demonstrates a Spearman correlation test which investigates the presence of

each competency for each sector. Each statistically significant correlation had a value of less

101
than .7, indicating no significant multi-collinearity problems exist between the presence of each

variable and word count (Valecha et al., 2021a).

Table 11a. Spearman Correlation (Private Sector)

Private Sector Correlations

word Priv_Et Priv_ Priv_ Priv_P Priv_St
count h_Bin ML_Bi AI_Bi S_Bin at_Bin
n n
Spearman' wordco Correlation --
s rho unt Coefficient
Sig. (2-tailed) .
N 1492
Priv_Et Correlation .212** --
h_Bin Coefficient
Sig. (2-tailed) <.001 .
N 1492 1492
Priv_M Correlation .106 **
-.023 --
L_Bin Coefficient
Sig. (2-tailed) <.001 .379 .
N 1492 1492 1492
Priv_AI Correlation .150 **
.039 .272** --
_Bin Coefficient
Sig. (2-tailed) <.001 .135 <.001 .
N 1492 1492 1492 1492
Priv_PS Correlation .234** .002 .076** .126** --
_Bin Coefficient
Sig. (2-tailed) <.001 .929 .003 <.001 .
N 1492 1492 1492 1492 1492
Priv_St Correlation .042 -.013 .294 **
.124 **
.059* --
at_Bin Coefficient
Sig. (2-tailed) .108 .622 <.001 <.001 .023 .
N 1492 1492 1492 1492 1492 1492
**. Correlation is significant at the 0.01 level (2-tailed).
*. Correlation is significant at the 0.05 level (2-tailed).

102
 Table 11b. Spearman Correlation (Federal Sector)

Federal Sector Correlations

word Fed_Et Fed_M Fed_ Fed_P Fed_Sta
count h_Bin L_Bin AI_Bi S_Bin t_Bin
n
Spearman word Correlation --
's rho count Coefficient
Sig. (2-tailed) .
N 649
Fed_Et Correlation .610** --
h_Bin Coefficient
Sig. (2-tailed) <.001 .
N 649 649
Fed_M Correlation -.057 -.167** --
L_Bin Coefficient
Sig. (2-tailed) .144 <.001 .
N 649 649 649
Fed_AI Correlation -.063 -.043 .136** --
_Bin Coefficient
Sig. (2-tailed) .111 .269 <.001 .
N 649 649 649 649
Fed_PS Correlation .177 **
.113 **
.060 .070 --
_Bin Coefficient
Sig. (2-tailed) <.001 .004 .124 .076 .
N 649 649 649 649 649
Fed_Sta Correlation .716 **
.571 **
-.103 **
-.033 .217** --
t_Bin Coefficient
Sig. (2-tailed) <.001 <.001 .008 .407 <.001 .
N 649 649 649 649 649 649
**. Correlation is significant at the 0.01 level (2-tailed).

Socialization
Within this section, we convey the results of our analysis and discuss the theoretical and

practical implications of this research example.

Within this component we sought to identify which dictionaries were statistically

significant. In order to do so, we conducted a TF-IDF analysis for each job posting relative to

103
each data science competency dictionary for the federal and private sectors to measure

comparisons. In other words, we analyzed each job within the federal (n=649) and private

(n=1,492) sector and computed TF-IDF scores for the data science competencies identified in

Figure 3. By doing so, we could measure the relevance and significance of the specific dictionary

within each sector. Upon computing the TF-IDF score for each job posting, we calculated the

mean of the dictionary’s TF-IDF score for each job postings to run our comparison analysis

between the federal and private sectors. Furthermore, we conducted an ANOVA analysis to

investigate the difference in means between the federal and private sectors and their relative

significance.

The federal sector places a stronger emphasis on competencies like statistics and ethics

within job postings, whereas the private sector favors problem solving and ethics. Within federal

job postings, statistics ranked first place as the most prominent competency represented within

the job postings. The private sector ranks problem solving as its top competency. The ethics

component shares the second-most prominent position within the ranking. Table 12 summarizes

the rankings of TF-IDF scores for both the federal and private sector.

Table 12. Private and Federal TF-IDF Ranking

Ranking (1= Federal TF-IDF Ranking Private TF-IDF
Strongest, 5= Ranking
Weakest)
1 Problem Solving Statistics
2 Ethics Ethics
3 Machine Learning Problem Solving
4 Artificial Intelligence Machine Learning
5 Statistics Artificial Intelligence

104
 Comparative Mapping

We compared the TF-IDF scores for the private and federal sectors to further analyze our

results (Figure 5a-b). Despite the comparatively low numbers of TF-IDF scores within the

private sector, the results indicate a more substantial presence of the statistics competency within

the federal sector when compared to the private. Comparatively, the private sector may place a

stronger emphasis within the job description on communicating the desired ability problem solve

when compared to the federal sector. Lastly, our research identified that the private sector has

significantly lower TF-IDF scores than the federal sector. The keywords most like ethics,

problem-solving, artificial intelligence, machine learning, and statistics within the private sector

were not used frequently throughout job descriptions, despite having strong cosine similarity

scores. The lower score could indicate a lack of representation throughout the job postings

despite being identified as a word that is closely related to the identified competencies.

Private Sector Average TFIDF Federal Sector Average TFIDF

Statistics Statistics
0.0005 0.008
0.0004 0.006
0.0003
0.004
Artificial 0.0002 Artificial
Ethics Ethics
Intelligence 0.0001 Intelligence 0.002
0 0

Machine Problem Machine Problem

Learning Solving Learning Solving

Figure 5. Both Sector TF-IDF Comparisons

105
 Means Difference Analysis

To test the mean differences among the data science competencies of each group (federal

and private), we conducted an ANOVA analysis via IBM’s SPSS software (Kim et al., 2005).

We also conducted a post-hoc t-test analysis comparing the sectors. The significance level for the

ANOVA test and t-test analysis was .05.

Table 13. ANOVA Means Comparison

ANOVA Table
Sum of df Mean F Sig.
Squares Square

Ethics TFIDF Between Groups 0.000 1 0.002 1159.36 <.001

Machine Learning TFIDF Between Groups 0.000 1 0.005 1220.62 <.001

Artificial Intelligence Between Groups 0.000 1 0.000 93.65 <.001

TFIDF

Problem Solving TFIDF Between Groups 0.000 1 0.000 504.65 <.001

Statistics TFIDF Between Groups 0.024 1 0.024 1007.30 <.001

According to the TF-IDF ANOVA results there is sufficient evidence that supports most

of the mean TF-IDF differs among the five competencies for federal and private job sectors at a

.05 level of significance. Regarding all competencies, the value is less than the established

significance level, the differences within the means are statistically significant. Therefore, the

ANOVA test results show that the TF-IDF mean values of the five competencies obtained from

106
the federal and private sectors are not identical, implying that the different competencies among

the groups have different weights (see Table 13).

Table 14. T-Test TFIDF Analysis

Levene's Test for Equality t-test for Equality of Means
of Variances
F Sig. t df Sig. (1-Tailed) Sig. (2-
tailed)

Ethics TFIDF
676.29 <.001 -34.05 2,139 <.001 <.001

Machine Learning 2,139

TFIDF 219.97 <.001 -34.94 <.001 <.001

Problem Solving 2,139

TFIDF 27.86 <.001 -9.68 <.001 <.001

Artificial Intelligence 2,139 <.001 <.001

TFIDF 440.01 <.001 -22.46

Statistics TFIDF 2,139 <.001 <.001

2661.82 <.001 -31.74

Through our post-hoc TF-IDF t-test analysis, we identified and validated the rejection of

the null such that the variances of the majority of the competencies differ from one another. This

analysis validated all competencies have a difference in means which are statistically

significantly different from zero. The difference in means for the Statistics competency is not

statistically significantly different from zero, which implies the sectors use this dictionary

similarly. Table 14 summarizes the mean difference results.

Post-Hoc Job Title Analysis

Understanding our primary dataset consisted of insights from 2022 job postings, we

sought to conduct a post-hoc analysis regarding job titles to identify any potential differences

between the groups of data leveraging the TF-IDF metrics. This section explores this post-hoc

analysis.

107
 Post-Hoc Private and Federal Sector Job Title Comparison

Understanding job data science titles may differ for both the private and federal sector,

we leveraged Mbah et al. (2017) and Saltz and Grady’s (2017) strategy in categorizing each job

posting based on synonyms and words contained within each job title. Saltz and Grady (2017)

propose five data science job title categories which consist of: 1) Data Scientist, 2) Data Science

Researcher, 3) Data Science Architect, 4) Data Science Programmer, and Data/Business.

Regarding data scientist, this reflects jobs that seek out individuals who can find and interpret

rich and specialized data sources and manage large amounts of data. This practice also includes

being able to visualize the data to aid in understanding complex data. Data science researcher

refers to jobs that seek out researchers/academic interested in exploring actionable knowledge

related to specific scientific problems, business processes, or hidden relationships. With regards

to the data science architect, these jobs refer to hiring an individual capable of designing and

maintaining the architecture of data science applications and/or facilities. Tangential, data

science programmer refers to jobs which seek out individuals who can code and develop

software for analytic applications and support enterprise-level processes. Where an architect

designs and creates blueprints, the programmer acts as the construction unit responsible for

building the desired construct/blueprint. Lastly, data/business analyst refers to jobs which seek

out individuals who can analyze large varieties of data and extract information about systems and

services with regards to organizational performance. We included a subsection dedicated to

management/leadership positions. Keywords within duty titles include: 1) director, head, lead,

manager (Alsudais et al., 2022).

108
 Table 15. Private and Federal Job Titles

Row Labels Private Sector Federal Sector

Data Science
Architect 265 156
Data Science
Programmer 14 5
Data Science
Researcher 17 9
Data Scientist 835 172
Data/Business
Analyst 115 256
Management 246 51
Total 1,492 649

With regards to the job titles, the highest data science job title within the private sector is

reflected by data science architect. Contrarily, data/business analyst is reflected for the federal

sector. This implies the federal sector seems more business-related data science skills while the

private sector communicates needs for specialized data scientists. Furthermore, both the private

and federal sector share their lowest numbers being in data science programmers and data

science researchers. This indicates that both sectors have fewer demands for academic and

programming-specific jobs. Table 15 summarizes these details.

Post-Hoc Private and Federal Sector Job Title Comparison

Once we coded the job titles and garnered an understanding of the demographics, we

conducted a comparative analysis of the TF-IDF scores related to each job title category. Within

this comparison, we the average TF-IDF scores calculated within our primary analysis and

ranked them to see how prominent each competency is related to each job category. Regarding

the private sector, problem solving reigned the most prominent within each duty title category.

Comparatively, the statistics competency was the most prominent throughout the federal sector.

When comparing both sectors, this analysis identified that although the statistics competency

was the most represented within the federal, it was the least represented within the private sector

109
job title categories. This could imply that a stronger emphasis is placed on the desire for

applicants to have stronger problem solving skills within the private sector, whereas the federal

sector favors more technical skills related to the statistics field. Figure 6a-b summarizes these

findings.

Federal Sector Job Category TF-IDF Comparison

0.035

0.03

0.025

0.02

0.015

0.01

0.005

0
Data Science Data Science Data Science Data Scientist Data/Business Management
Architect Programmer Researcher

Average of Ethics TFIDF Average of ML TFIDF Average of AI TFIDF

Average of PS TFIDF Average of STAT TFIDF

Figure 6a. Federal Sector TF-IDF Comparison

Private Sector Job Category TF-IDF Comparison

0.0006

0.0005

0.0004

0.0003

0.0002

0.0001

0
Data Science Data Science Data Science Data Scientist Data/Business Management
Architect Programmer Researcher

Average of Ethics TFIDF Average of MLTFIDF Average of AI TFIDF

Average of PS TFIDF Average of STAT TFIDF

Figure 6b. Federal Sector TF-IDF Comparison

110
 Future Directions and Limitations

By leveraging the competencies derived from IBM and Deloitte, we conducted a content

analysis on job postings from the United States federal government and private sector to gain

insight into the emphasis of each data science competency. Through this analysis, we identified

three themes for future research. The first theme we identified is that there is a lack of research

investigating the differences in job postings between the federal government and private sectors.

Future research endeavors could dive deeper into specific skill requirements (R-programming,

analytics, python) and how they differ from the federal and private sectors. Secondly, statistics

within the private sector represented the lowest competency in job postings. Our analysis

encourages future research endeavors to investigate the relationship between the statistical

expectations of employees and the semantics within job descriptions. Lastly, our research

identified that the private and federal sector’s TF-IDF scores fluctuate in presence which could

indicate shifts in demand for prospective employees. Given the global COVID pandemic, future

research could continue the temporal exploration of labor force demands within the data science

community.

Considering these future research themes, we acknowledge that it is essential to identify

the limitations of this research. Although systematic, the methodology for data collection limited

us in the variables we could explore. If we had access to temporal considerations (how long the

posting was open) or attractiveness of the posting (number of applications received), we could

have dived deeper into how the dictionaries could affect the recruitment efforts of the federal and

private sectors. Furthermore, the sample sizes of the job postings were limited, despite each

being unique and from an archival source. The research team eliminated any duplicate job

postings to reduce redundancy, which caused the federal job postings to drop drastically. This is

111
due to the federal sector recycling job descriptions. Other job posting portals could be leveraged

to expand this research endeavor; however, we chose to keep the sources the same for the sake of

symmetry and standardization. Other venues to acquire job posting data may produce a larger

sample size such as stronger web scraping programs and other coding mechanisms. Moreover,

future research could investigate how data science job descriptions and associated data

competency dictionaries have evolved for the federal and private sectors. Given the datasets

leveraged for this analysis, we did not have the data to conduct that in the depth of research.

Additionally, our data source reflects a snapshot in time from when the data was pulled.

Lastly, there are other pre-trained topic modeling models within Genism that could be

leveraged instead of Word2Vec. Namely, Doc2Vec, Latent Dirichlet Allocation, Author Topic

Modeling, and Hierarchical Dirichlet Processing are a few modeling techniques which fall under

the Genism library. Moreover, the post-hoc statistical models can also be enhanced or exchanged

for more effective techniques as newer developments occur. Through these limitations, we seek

to create a starting point for future research endeavors to grow upon.

112
 CHAPTER 9: CONCLUSION

Understanding big data, AI, and ML have accelerated business intelligence capabilities

for the United States federal government and private sector, research endeavors have also grown

in this interest. However, there is a lack of IS research frameworks within the HRM domain

which leverage NLP tools. This paper seeks to address the IS research gap by proposing the JC-

Compass framework to help guide future multidisciplinary competency-based job description

research. Figure 7 summarizes our proposed method to guide future research competency-based

research endeavors.

Figure 7. Federal Sector TF-IDF Comparison

We acknowledge there are limitations to this research and framework. For instance, there

are other models which are optimized for context analysis such as the models addressed in Table

4. This could provide new considerations for contextually understanding competencies within

job postings rather than relying on word similarities. This framework serves as a foundation, in

hopes that new frameworks come to light and enrich the IS and HRM research domain.

Moreover, we intend for this framework to be a starting point for scholars from academic and

industry backgrounds alike. As NLP capabilities become more sophisticated, the need for tools

and models to guide researchers in using these capabilities becomes imperative to exploring this

research domain.

113
 Where previous literature has homed explicitly in on skill requirements, we wanted to

divert from this trend to focus on a competency approach to analyzing job descriptions. Through

this framework proposal, empirical example, we demonstrated how this framework can benefit

business experts ranging from a variety of different career fields given its systematic and

transparent foundation. We hope the proposed competency model helps HR and IS professionals

expand the growing pool of literature surrounding job description analysis.

114
 APPENDICES

Appendix A: Other Text Embedding Analyses Techniques

Within HRM research, text embeddings are defined as a distributed representation of

terms which are obtained from a neural-network model (Bengio et al., 2000; Fernández-Reyes

and Shinde, 2019). There are two main categories of embedding text: 1) word embedding and 2)

sentence embedding. Word embedding is the practice of encoding of each word present in the

text into a numerical vector, while sentence embedding is the encoding of each sentence present

in the text. Given the interest of exploring specified competencies within job descriptions, word

embeddings empower researchers with the ability to look at specific words and identify

similarities surrounding them. Whereas sentence embeddings focus on sentence similarities.

Thus, this research focuses its attention on word embedding capabilities. Given the recency of

NLP expansion and development, new word embedding models have come into existence. For

instance, Word2Vec (Mikolov et al., 2013; Fernández-Reyes and Shinde, 2019), FastText

(Bojanowski et al., 2017), GloVe (Pennington et al., 2014), ELMo (Peters et al., 2018) and

BERT (Delvin et al., 2018). Table 4 summarizes each word embedding model, its general

purpose, and limitations.

Table 1. Word Embedding Model Summaries

Word Embedding Model Purpose
Name
Word2Vec Semantic & Similarity Tasks
(word vectorization, local
context), prediction-based
model
GloVe Semantic & Similarity Tasks
(Global Context), count-based
method
FastText Syntactic Tasks (n-gram
character word embeddings)
Limitations Source
Single vector representation Mikolov et al.
for word (regardless of (2013)
context); does not handle out
of vocabular words
High memory cost; Does not Pennington et
handle out of vocabulary al. (2014)
words; does not account for
context word is used
High memory cost; single Bojanowski
vector representation for word et al. (2017)
(regardless of context)

115
 Table 1. Word Embedding Model Summaries Continued
ELMo Word Context Tasks Not designed for transfer Peters et al.
learning, needs to be trained (2018)
for specific tasks
BERT Word Context Tasks High memory cost; slow to Delvin et al.
train; not stand alone system (2018)
(designed to be input into
other systems)

116
Appendix B: Private Sector Dictionary (Word2Vec Output)

Ethics Dictionary Machine Learning Problem Artificial Statistics Dictionary

Dictionary Solving Intelligence (AI)
Dictionary Dictionary
Exhibit Unsupervised Pragmatic Cognit Econometric
Uphold Reinforce Solver Augment Theor/etical/y
Style Artificial Break Ai Infer/ential
Urgency Multimodal Diagnose Fusion Causal
Credible Causal Difficult Chip Exploratory
Ethic Convolution Conceptual Multimodal Actuary
Meticulous Neural Frame Exploit Mathematic/al
Demeanor Infer/entail/ce Thinker Signal Algebra
Motiv/e/ation Theor/etical/y Dig Artificial Statistic
Intelligence
Keen Vector Problem Audio Correlate
solv/e/ing
Attitude Machine Learn/ing Intuit Tree Quantitative

117
Appendix C: Federal Sector Dictionary (Word2Vec Output)

Ethics Machine Learning Problem Solving Artificial Intelligence (Ai) Statistics

Dictionary Dictionary Dictionary Dictionary Dictionary
Administer Artificial Disseminate Threat Probable
Oak Imagery Diver Compensate/Ion Logic
Interevent Force Experiment Imagery Compute
Align Lab Propose Access Substantial
Suitable Draw Trial Dollar Differential
Elite Contribute Substant Visual Thesis
Material Access Quality Create Economy
Problem
Screen Surface solv/e/ing Vehicle Semester
Ethic Clean Oversee Spatial Statistic
Stringent Farm Valid Artificial Intelligence Geography
Cognit Machine Learn/ing Concept Nuclear Site

118
Appendix D: FULL Private Sector Dictionary (TF-IDF Analysis)

Ethics Machine Problem Solving Artificial Intelligence Statistics Dictionary

Dictionary Learning Dictionary (AI)
Dictionary Dictionary
Exhibit Unsupervised Pragmatic Importance Econometric
Display Unruled Businesslike Seriousness Theoretic/Theory
Uphold Liberated Down-To-Earth Add To Approach
Defend Emancipated Efficient Amplify Argument
Justify Released Hardheaded Boost Assumption
Maintain Unconquered Logical Build Up Code
Style Delivered Practical Develop Concept
Importance Freed Realistic Enhance Doctrine
Mode Redeemed Sober Enlarge Idea
Tone Reinforce Crack Expand Ideology
Manner Confirm Break Heighten Method
Urgency/Urgent Bolster Gap Intensify Philosopher
Magnitude Corroborate Hole Multiply Plan
Exigency Establish Diagnose Reinforce Position
Credible Support Analyze Strengthen Premise
Insistence Verify Determine Inflate Proposal
Seriousness Buttress Investigate Ai Provision
Significance Artificial Pinpoint Artificial Intelligence Rationale
Gravity Unnatural Pronounce Robotics Scheme
Plausible Mock Recognize Fusion Speculation
Attitude False Difficult Amalgam Suspicion
Stance Simulated Ambitious Blend System
Posture Hypothesis Arduous Synthesis Thesis
Careful Affected Burdensome Chip Understanding
Meticulous Strained Challenging Slice Statistic
Conscientious Fake Crucial Wafer Inferential
Loving Mechanical Demanding Multimodal Analytical
Diligent Pseudo Laborious Exploit Conjecture
Demeanor Exaggerated Onerous Accomplishment Interpretation
Behavior Multimodal Painful Adventure Presumption
Action Causal Problematic Deed Reasoning
Motive Creative Severe Escapade Supposition
Matter Constructive Strenuous Feat Causal
Topic Convolution Tough Signal Creative
Question Complexity Troublesome Conspicuous Imaginative
Theme Complication Conceptual Momentous Innovative
Motif Difficulty Theoretical Noteworthy Inventive

119
 Appendix D: FULL Private Sector Dictionary (TF-IDF Analysis) Continued
Subject Neural Visionary Noticeable Numerical
Fair Infer Frame Salient Seminal
Model Derive Body Tree Unconventional
Performance Decide Cage Sapling Unusual
Show Understand Philosopher Forest Exploratory
Advocate Theoretic/Theory Fabric Seedling Preparatory
Confirm Thesis Framework Shrub Prior
Encourage Vector Physique Timber Actuary
Endorse Machine Learn Scaffolding Wood Analyst
Hold To Robotics Structure Auditor
Quantitative Thinker Bookkeeper
Promote Dig Clerk
Side With Gibe Mathematic
Stand By Intuit Theoretic
Vindicate Hunch Scientific
Ethic Instinct Abstract
Integrity Problem Academic
Solv/E/Ing
Moral Calcul/us/ate/ation
Principle
Virtue

120
Appendix E: FULL Federal Sector Dictionary (TF-IDF Analysis)

Ethics Machine Learning Problem Solving Artificial Statistics Dictionary

Dictionary Dictionary Dictionary Intelligence (Ai)
Dictionary
Ethic Machine Learn Disseminate Threat Probable/ly
Administrate Artificial Advertise Blackmail Anticipation
Carry Out Robotics Circulate Hazard Contingency
Conduct Imager Disperse Intimidation Chance
Direct Symbolism Publicize Menace Feasibility
Execute Force Propagate Peril Expectation
Govern Effort Publish Risk Odds
Oversee Strength Propose Artificial Intelligence Possibility
Supervise Violence Diver Compensate/ion Prospect
Oak Lab Experiment Allowance Logic
Intervention Workshop Analysis Premium Philosophy
Interference Draw Attempt Benefit Rationale
Mediation Tie Enterprise Coverage Sanity
Align Contribute/ion Examination Earnings Sense
Adjust Add Exercise Fee Comput/e/er/tation
Satisfactory Commit Experimentation Indemnity Figure Out
Coordinate Devote Measure Pay Gauge
Regulate Give Observation Profit Substantial
Suitable Grant Operation Redress Big
Advisable Pony Up Ask Reimbursement Consequential
Applicable Share Practice Remuneration Extraordinary
Apt Subsidize Probe Remittance Considerable
Convenient Supply Procedure Restitution Hefty
Correct Approach Research Reparation Generous
Fitting Access Search Reward Large
Good Connection Study Salary Massive
Good Enough Entry Trial Settlement Meaningful
Handy Surface Undertaking Stipend Serious
Proper Exterior Venture Wage Significant
Sufficient Facial Come Up With Imager Sizable
Reasonable Outer Introduce Symbolism Solid
Relevant Outward Nominate Access Steady
Suited Shallow Offer Connection Strong
Useful Superficial Exploratory Approach Valuable
Elite Clean Recommend Dollar Vast
Exclusive Blank Request Entry Different
Silk-Stocking Bright Submit Currency Disparate

121
 Appendix E: FULL Federal Sector Dictionary (TF-IDF Analysis) Continued
Mater Clear Urge Visual Contrasting
Actual Elegant Preliminary Greenback Distant
Appreciable Fresh Substant Create Distinct
Earthly Graceful Big Invent Divergent
Perceptible Hygienic Consequential Build Diverse
Physical Immaculate Considerable Conceive Offbeat
Substantial Neat Extraordinary Constitute Other
Screen Orderly Hefty Construct Economy
Cover Pure Generous Make Particular
Curtain Squeaky Clean Large Design Peculiar
Net Spotless Massive Devise Various
Stringent Simple Meaningful Discover Thesis
Binding Tidy Serious Establish Contention
Demanding Unblemished Significant Forge Hypothesis
Draconian Washed Solid Organize Budgetary
Exacting Farm Sizable Form Opinion
Forceful Acreage Steady Found Premise
Harsh Estate Strong Generate Proposition
Inflexible Field Valuable Initiate Supposition
Ironclad Garden Vast Plan Theory
Rigorous Grassland Quality Produce Commercial
Severe Homestead Concept Set Up Fiscal
Stiff Lawn Approach Shape Industrial
Strict Meadow Conception Spawn Monetary
Tough Nursery Image Start Semester
Cognit Orchard Notion Vehicle Content
Pasture Perception Automobile Contented
Ethic Plantation Theory Bicycle Comfortable
Administrate Ranch Thought Bus Fulfilled
Carry Out Ai View Boat Gratified
Conduct Opportunity Oversee Cab Pleased
Direct Load Manage Car Happy
Execute Anomaly Command Jeep Satisfied
Govern Detect Supervise Taxi Willing
Oversee Multidimension Survey Truck Mathematic/s/al
Supervise Query Watch Wagon Algebra
Valid Van Calculation
Accurate Spatial Calculus
Authentic Nuclear Geometry
Binding Basic Math
Compelling Essential Statistic
Conclusive Fundamental
Credible Important
Good Key
Lawful Paramount
Legal Pivotal

122
Appendix E: FULL Federal Sector Dictionary (TF-IDF Analysis) Continued
Legitimate Significant
Logical
Original
Persuasive
True
Well-Founded
Problem
Solv/e/ing

123
 REFERENCES

Abbasi, A., & Chen, H. (2008). CyberGate: A Design Framework and System for Text Analysis of
Computer-Mediated Communication. MIS Quarterly, 32(4), 811–837. https://doi.org/10.2307/2514
8873

Abbasi, Albrecht, Vance, & Hansen. (2012). MetaFraud: A Meta-Learning Framework for Detecting
Financial Fraud. MIS Quarterly, 36(4), 1293. https://doi.org/10.2307/41703508

Adewumi, T. P., Liwicki, F., & Liwicki, M. (2021). Word2Vec: Optimal Hyper-Parameters and Their
Impact on NLP Downstream Tasks (arXiv:2003.11645). arXiv. http://arxiv.org/abs/2003.11645

Akay, A., Dragomir, A., & Erlandsson, B.-E. (2015). Network-Based Modeling and Intelligent Data
Mining of Social Media for Improving Care. IEEE Journal of Biomedical and Health Informatics,
19(1), 210–218. https://doi.org/10.1109/JBHI.2014.2336251

Alsudais, A., Aldumaykhi, A., & Otai, S. (2022). Comparison of Job Titles for Specific Terms:
Investigating “Data Science.” In E. Pardede, P. Delir Haghighi, I. Khalil, & G. Kotsis (Eds.),
Information Integration and Web Intelligence (pp. 98–103). Springer Nature Switzerland. https:
//doi.org/10.1007/978-3-031-21047-1_

Amaar, A., Aljedaani, W., Rustam, F., Ullah, S., Rupapara, V., & Ludi, S. (2022). Detection of Fake
Job Postings by Utilizing Machine Learning and Natural Language Processing Approaches. Neural
Processing Letters, 54(3), 2219–2247. https://doi.org/10.1007/s11063-021-10727-z

Ammanath, B. (2022). Trustworthy AI: A Business Guide for Navigating Trust and Ethics in AI.
John Wiley & Sons.

Anandarajan, M., Hill, C., & Nolan, T. (2019). Text Preprocessing. In M. Anandarajan, C. Hill, & T.
Nolan (Eds.), Practical Text Analytics: Maximizing the Value of Text Data (pp. 45–59). Springer
International Publishing. https://doi.org/10.1007/978-3-319-95663-3_4

Anton, E., Behne, A., & Teuteberg, F. (2020). The Humans Behind Artificial Intelligence–An
Operationalisation of AI Competencies.

Bafna, P., Pramod, D., & Vaidya, A. (2016). Document clustering: TF-IDF approach. 2016
International Conference on Electrical, Electronics, and Optimization Techniques (ICEEOT), 61–66.
https://doi.org/10.1109/ICEEOT.2016.7754750

Barducci, A., Iannaccone, S., La Gatta, V., Moscato, V., Sperlì, G., & Zavota, S. (2022). An end-to-
end framework for information extraction from Italian resumes. Expert Systems with Applications,
210, 118487. https://doi.org/10.1016/j.eswa.2022.118487

Bengio, Y., Ducharme, R., & Vincent, P. (2000). A Neural Probabilistic Language Model. Advances
in Neural Information Processing Systems, 13. https://proceedings.neurips.cc/paper/2000/hash/728f2
06c2a01bf572b5940d7d9a8fa4c-Abstract.html

124
Benjamin, V., Valacich, J. S., & Chen, H. (2019). DICE-E: A Framework for Conducting Darknet
Identification, Collection, Evaluation with Ethics. MIS Quarterly, 43(1), 1–22. https://doi.org/10.
25300/MISQ/2019/13808

Bojanowski, P., Grave, E., Joulin, A., & Mikolov, T. (2017). Enriching Word Vectors with Subword
Information. Transactions of the Association for Computational Linguistics, 5, 135–146. https://doi
.org/10.1162/tacl_a_00051

Boon, C., Eckardt, R., Lepak, D. P., & Boselie, P. (2018). Integrating strategic human capital and
strategic human resource management. The International Journal of Human Resource Management,
29(1), 34–67. https://doi.org/10.1080/09585192.2017.1380063

Boyatzis, R. E. (2008). Competencies in the 21st century. Journal of Management Development,

27(1), 5–12. https://doi.org/10.1108/02621710810840730

Brumer, Y., Shapira, B., Rokach, L., & Barkan, O. (2017). Predicting Relevance Scores for Triples
from Type-Like Relations using Neural Embedding.

Carretta, A. (1992). Career and succession planning. In Competency Based Human Resource
Management. HayGroup.

Chen, J., Chen, C., & Liang, Y. (2016). Optimized TF-IDF Algorithm with the Adaptive Weight of
Position of Word. 114–117. https://doi.org/10.2991/aiie-16.2016.28

Chouhan, V. S., & Srivastava, S. (2014). Understanding Competencies and Competency Modeling ―
A Literature Survey. IOSR Journal of Business and Management, 16(1), 14–22. https://doi.org/10.
9790/487X-16111422

Christensen, B. J., Lentz, R., Mortensen, D. T., Neumann, G. R., & Werwatz, A. (2005). On‐the‐Job
Search and the Wage Distribution. Journal of Labor Economics, 23(1), 31–58. https://doi.org/10.1
086/425432

Chuang, Z., Ming, W., Guang, L. C., Bo, X., & Zhi-qing, L. (2009). Resume Parser: Semi-structured
Chinese Document Analysis. 2009 WRI World Congress on Computer Science and Information
Engineering, 5, 12–16. https://doi.org/10.1109/CSIE.2009.562

Cole, M. S., Feild, H. S., & Giles, W. F. (2003). Using Recruiter Assessments of Applicants’ Resume
Content to Predict Applicant Mental Ability and Big Five Personality Dimensions. International
Journal of Selection and Assessment, 11(1), 78–88. https://doi.org/10.1111/1468-2389.00228

Deguchi, T., Seo, S., & Ishii, N. (2022). Meaning of the Clusters on Dimensionality Reduction by
Word Clustering. 2022 12th International Congress on Advanced Applied Informatics (IIAI-AAI),
325–330. https://doi.org/10.1109/IIAIAAI55812.2022.00072

125
Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018). BERT: Pre-training of Deep
Bidirectional Transformers for Language Understanding. https://doi.org/10.48550/arXiv.1810.04805

Embarak, Dr. O. (2018). Data Analysis and Visualization Using Python: Analyze Data to Create
Visualizations for BI Systems. Apress. https://doi.org/10.1007/978-1-4842-4109-7

Enholm, I. M., Papagiannidis, E., Mikalef, P., & Krogstie, J. (2022). Artificial Intelligence and
Business Value: A Literature Review. Information Systems Frontiers, 24(5), 1709–1734.
https://doi.org/10.1007/s10796-021-10186-w

Falatah, R. (2021). The Impact of the Coronavirus Disease (COVID-19) Pandemic on Nurses’
Turnover Intention: An Integrative Review. Nursing Reports, 11(4), Article 4. https://doi.org/1
0.3390/nursrep11040075

Fernández-Reyes, F. C., & Shinde, S. (2019). CV Retrieval System based on job description
matching using hybrid word embeddings. Computer Speech & Language, 56, 73–79. https://doi.or
g/10.1016/j.csl.2019.01.003

Gao, Q., Huang, X., Dong, K., Liang, Z., & Wu, J. (2022). Semantic-enhanced topic evolution
analysis: A combination of the dynamic topic model and word2vec. Scientometrics, 127(3), 1543–
1563. https://doi.org/10.1007/s11192-022-04275-z

Gardiner, A., Aasheim, C., Rutner, P., & Williams, S. (2018). Skill Requirements in Big Data: A
Content Analysis of Job Advertisements. Journal of Computer Information Systems, 58(4), 374–384.
https://doi.org/10.1080/08874417.2017.1289354

GeeksforGeeks. (2018, February 2). Part of Speech Tagging with Stop words using NLTK in python.

GeeksforGeeks. https://www.geeksforgeeks.org/part-speech-tagging-stop-words-using-nltk-python/

Goldberg, Y., & Levy, O. (2014). word2vec Explained: Deriving Mikolov et al.’s negative-sampling
word-embedding method (arXiv:1402.3722). arXiv. http://arxiv.org/abs/1402.3722

Goldfarb, A., Taska, B., & Teodoridis, F. (2023). Could machine learning be a general purpose
technology? A comparison of emerging technologies using data from online job postings. Research
Policy, 52(1), 104653. https://doi.org/10.1016/j.respol.2022.104653

Gonzalez, R., & Sol, H. (2012). Validation and Design Science Research in Information Systems. In
Research Methodologies, Innovations and Philosophies in Software Systems Engineering and
Information Systems (pp. 403–426). https://doi.org/10.4018/978-1-4666-0179-6.ch021

Gottipati, S., Shim, K. J., & Sahoo, S. (2021). Glassdoor Job Description Analytics – Analyzing Data
Science Professional Roles and Skills. 2021 IEEE Global Engineering Education Conference
(EDUCON), 1329–1336. https://doi.org/10.1109/EDUCON46332.2021.9453931

126
Gregor, S., & Hevner, A. R. (2013). Positioning and Presenting Design Science Research for
Maximum Impact. MIS Quarterly, 37(2), 337–355. https://doi.org/10.25300/MISQ/2013/37.2.01

Gupta, A. (2022). Company, Job, and Perceived Advantage Influence on Job Hunting: Evaluating
Glassdoor Job Hunting platform. Journal of Digitovation and Information System, 2(1), Article 1.
https://doi.org/10.54433/JDIIS.2022100011

Ha, T., Lee, M., Yun, B., & Coh, B.-Y. (2022). Job Forecasting Based on the Patent Information: A
Word Embedding-Based Approach. IEEE Access, 10, 7223–7233.
https://doi.org/10.1109/ACCESS.2022.3141910

Hardeniya, N., Perkins, J., Chopra, D., Joshi, N., & Mathur, I. (2016). Natural Language Processing:
Python and NLTK. Packt Publishing Ltd.

Heckman, J. (2021, May 14). First cohort of federal employees graduate from OMB’s data science
upskilling program. Federal News Network. https://federalnewsnetwork.com/workforce/2021/05
/first-cohort-of-federal-employees-graduate-from-ombs-data-science-upskilling-program/

Heidarysafa, M., Kowsari, K., Bashiri, M., & Brown, D. E. (2021). Toward a Knowledge Discovery
Framework for Data Science Job Market in the United States (arXiv:2106.11077). arXiv. http://ar
xiv.org/abs/2106.11077

Hevner, A. R., March, S. T., Park, J., & Ram, S. (2004). Design Science in Information Systems
Research. MIS Quarterly, 28(1), 75–105. https://doi.org/10.2307/25148625

Horvath, A.-S. (2022). ComPara: A corpus linguistics in English of computation in architecture

dataset. Data in Brief, 42, 108169. https://doi.org/10.1016/j.dib.2022.108169

Hosain, S., & Liu, P. (2020). LinkedIn for Searching Better Job Opportunity: Passive Jobseekers’
Perceived Experience. The Qualitative Report, 25(10), 3719–3732.

Htait, A., Fournier, S., Bellot, P., Htait, A., Fournier, S., & Bellot, P. (2018). Unsupervised Creation
of Normalization Dictionaries for Micro-Blogs in Arabic, French and English. Computación y
Sistemas, 22(3), 729–737. https://doi.org/10.13053/cys-22-3-3034

Hu, K., Luo, Q., Qi, K., Yang, S., Mao, J., Fu, X., Zheng, J., Wu, H., Guo, Y., & Zhu, Q. (2019).
Understanding the topic evolution of scientific literatures like an evolving city: Using Google
Word2Vec model and spatial autocorrelation analysis. Information Processing & Management, 56(4),
1185–1203. https://doi.org/10.1016/j.ipm.2019.02.014

IBM. (2020). The Data Science Skills Competency Model. IBM Corporation, 12.

Jalilifard, A., Caridá, V. F., Mansano, A. F., Cristo, R. S., & da Fonseca, F. P. C. (2021). Semantic
Sensitive TF-IDF to Determine Word Relevance in Documents. In S. M. Thampi, E. Gelenbe, M.
Atiquzzaman, V. Chaudhary, & K.-C. Li (Eds.), Advances in Computing and Network
Communications (pp. 327–337). Springer. https://doi.org/10.1007/978-981-33-6987-0_27

127
Janssen, M., & Estevez, E. (2013). Lean government and platform-based governance—Doing more
with less. Government Information Quarterly, 30, S1–S8. https://doi.org/10.1016/j.giq.2012.11.003

Kandula, S. R. (2013). Competency-Based Human Resource Management: A Complete Text with

Case Studies on Competency Mapping, Modelling, Assessing and Applying. PHI Learning Pvt. Ltd.

Khallaf, A., & Majdalawieh, M. (2012). Investigating the Impact of CIO Competencies on IT
Security Performance of the U.S. Federal Government Agencies. Information Systems Management,
29(1), 55–78. https://doi.org/10.1080/10580530.2012.634298

Kim, D. J., Song, Y. I., Braynov, S. B., & Rao, H. R. (2005). A multidimensional trust formation
model in B-to-C e-commerce: A conceptual framework and content analyses of academia/practitioner
perspectives. Decision Support Systems, 40(2), 143–165. https://doi.org/10.1016/j.dss.2004.01.006

Kortum, H., Rebstadt, J., & Thomas, O. (2022). Dissection of AI Job Advertisements: A Text
Mining-based Analysis of Employee Skills in the Disciplines Computer Vision and Natural Language
Processing. Hawaii International Conference on System Sciences. https://doi.org/10.24251/HICSS.
2022.635

Kreuter, F., Ghani, R., & Lane, J. (2019). Change Through Data: A Data Analytics Training Program
for Government Employees. Harvard Data Science Review. https://doi.org/10.1162/99608f92ed35
3ae3

Kurnia, R. I. (2020). Classification of User Comment Using Word2vec and SVM Classifier.
International Journal of Advanced Trends in Computer Science and Engineering, 9(1), 643–648.
https://doi.org/10.30534/ijatcse/2020/90912020

Laber, M. E., & O’Connor, J. K. (2000). Competency modeling: Ready, set, research. The Industrial-
Organizational Psychologist, 37(4), 91–96.

Landers, R., Brusso, R., & Auer, E. (2019). Crowdsourcing Job Satisfaction Data: Examining the
Construct Validity of Glassdoor.com Ratings. Personnel Assessment and Decisions, 5(3). https://doi.
org/10.25035/pad.2019.03.006

Landis, J. R., & Koch, G. G. (1977). The Measurement of Observer Agreement for Categorical Data.
Biometrics, 33(1), 159–174. https://doi.org/10.2307/2529310

Larsen, K., Lukyanenko, R., Mueller, R., Storey, V., Vander Meer, D., Parsons, J., & Hovorka, D.
(2020, June 1). Validity in Design Science Research.

Layug, C. (2018). Extracting Major Topics From Survey Text Responses Using Natural Language
Processing. Naval Post Graduate School.

Le, Q., & Mikolov, T. (2014). Distributed Representations of Sentences and Documents. International
Conference on Machine Learning.

128
López, F., & Romero, V. (2014). Mastering Python Regular Expressions. Packt Publishing Ltd.

Lunn, S., Zhu, J., & Ross, M. (2020). Utilizing Web Scraping and Natural Language Processing to
Better Inform Pedagogical Practice. 2020 IEEE Frontiers in Education Conference (FIE), 1–9.
https://doi.org/10.1109/FIE44824.2020.9274270

Mansfield, R. S. (1996). Building competency models: Approaches for HR professionals. Human

Resource Management, 35(1), 7–18. https://doi.org/10.1002/(SICI)1099-050X(199621)35:1<7::AID-
HRM1>3.0.CO;2-2

Marinescu, I. E., Skandalis, D., & Zhao, D. (2020). Job Search, Job Posting and Unemployment
Insurance During the COVID-19 Crisis (SSRN Scholarly Paper No. 3664265). https://doi.org/
10.2139/ssrn.3664265

Markus, M. L., Majchrzak, A., & Gasser, L. (2002). A Design Theory for Systems That Support
Emergent Knowledge Processes. MIS Quarterly, 26(3), 179–212.

Martinez Soriano, I., Castro Peña, J. L., Fernandez Breis, J. T., San Román, I., Alonso Barriuso, A.,
& Guevara Baraza, D. (2019). Snomed2Vec: Representation of SNOMED CT Terms with
Word2Vec. 2019 IEEE 32nd International Symposium on Computer-Based Medical Systems
(CBMS), 678–683. https://doi.org/10.1109/CBMS.2019.00138

Mbah, R. B., Rege, M., & Misra, B. (2017). Discovering Job Market Trends with Text Analytics.
2017 International Conference on Information Technology (ICIT), 137–142. https://doi.org/10.11
09/ICIT.2017.29

McCabe, M. B. (2017). SOCIAL MEDIA MARKETING STRATEGIES FOR CAREER

ADVANCEMENT: AN ANALYSIS OF LINKEDIN. Journal of Business and Behavioral Sciences,
29(1).

McHugh, M. L. (2012). Interrater reliability: The kappa statistic. Biochemia Medica, 22(3), 276–282.

Meyer, M. A. (2019). Healthcare data scientist qualifications, skills, and job focus: A content analysis
of job postings. Journal of the American Medical Informatics Association, 26(5), 383–391. https://do
i.org/10.1093/jamia/ocy181

Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013). Efficient Estimation of Word Representations
in Vector Space (arXiv:1301.3781). arXiv. http://arxiv.org/abs/1301.3781

Mökander, J., & Floridi, L. (2021). Ethics-Based Auditing to Develop Trustworthy AI. Minds and
Machines, 31(2), 323–327. https://doi.org/10.1007/s11023-021-09557-8

Muthyala, R., Wood, S., Jin, Y., Qin, Y., Gao, H., & Rai, A. (2017). Data-Driven Job Search Engine
Using Skills and Company Attribute Filters. 2017 IEEE International Conference on Data Mining
Workshops (ICDMW), 199–206. https://doi.org/10.1109/ICDMW.2017.33

129
Nash, M. (2018). ‘Let’s work on your weaknesses’: Australian CrossFit coaching, masculinity and
neoliberal framings of ‘health’ and ‘fitness.’ Sport in Society, 21(9), 1432–1453. https://doi.org/
10.1080/17430437.2017.1390565

Nutt, P. C. (2006). Comparing Public and Private Sector Decision-Making Practices. Journal of
Public Administration Research and Theory, 16(2), 289–318. https://doi.org/10.1093/jopart/mui041

Oliveira, M., Bitencourt, C. C., Santos, A. C. M. Z. dos, & Teixeira, E. K. (2015). Thematic Content
Analysis: Is There a Difference Between the Support Provided by the MAXQDA® and NVivo®
Software Packages? Revista de Administração Da UFSM, 9(1), 72–82. https://doi.org/10.5902
/1983465911213

Ouchchy, L., Coin, A., & Dubljević, V. (2020). AI in the headlines: The portrayal of the ethical
issues of artificial intelligence in the media. AI & SOCIETY, 35(4), 927–936. https://doi.org/10
.1007/s00146-020-00965-5

Ozkur, M., Benlier, N., Takan, I., Vasileiou, C., Georgakilas, A. G., Pavlopoulou, A., Cetin, Z., &
Saygili, E. I. (2022). Ginger for Healthy Ageing: A Systematic Review on Current Evidence of Its
Antioxidant, Anti-Inflammatory, and Anticancer Properties. Oxidative Medicine and Cellular
Longevity, 2022, e4748447. https://doi.org/10.1155/2022/4748447

Park, H. M. (2009). Comparing Group Means: T-tests and One-way ANOVA Using Stata, SAS, R,
and SPSS. https://scholarworks.iu.edu/dspace/handle/2022/19735

Pennington, J., Socher, R., & Manning, C. (2014). Glove: Global Vectors for Word Representation.
Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing
(EMNLP), 1532–1543. https://doi.org/10.3115/v1/D14-1162

Peppard, J. (2007). The conundrum of IT management. European Journal of Information Systems,

16(4), 336–345. https://doi.org/10.1057/palgrave.ejis.3000697

Perlman, B. J. (2016). Human Resource Management at the Local Level: Strategic Thinking and
Tactical Action. State and Local Government Review, 48(2), 114–120. https://doi.org/10.1177/016
0323X16659115

Peters, M. E., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018).
Deep contextualized word representations (arXiv:1802.05365). arXiv. http://arxiv.org/abs/1802.0
5365

Poomagal, S., Malar, B., Ranganayaki, E. M., Deepika, K., & Dheepak, G. (2022). Sentiment
Thesaurus, Synset and Word2Vec Based Improvement in Bigram Model for Classifying Product
Reviews. Sn Computer Science, 3(6), 422. https://doi.org/10.1007/s42979-022-01305-8

Probert, D. R., Farrukh, C. J. P., & Phaal, R. (2003). Technology roadmapping—Developing a

practical approach for linking resources to strategic goals. Proceedings of the Institution of

130
Mechanical Engineers, Part B: Journal of Engineering Manufacture, 217(9), 1183–1195. https://doi.o
rg/10.1243/095440503322420115

Qaiser, S., & Ali, R. (2018). Text Mining: Use of TF-IDF to Examine the Relevance of Words to
Documents. International Journal of Computer Applications, 181. https://doi.org/10.5120/ijca20
18917395

Quan, P., Shi, Y., Niu, L., Liu, Y., & Zhang, T. (2018). Automatic Chinese Multiple-Choice Question
Generation for Human Resource Performance Appraisal. Procedia Computer Science, 139, 165–172.
https://doi.org/10.1016/j.procs.2018.10.235

Rajagopalan, G. (2021). Regular Expressions and Math with Python. In G. Rajagopalan (Ed.), A
Python Data Analyst’s Toolkit: Learn Python and Python-based Libraries with Applications in Data
Analysis and Statistics (pp. 77–99). Apress. https://doi.org/10.1007/978-1-4842-6399-0_3

Ratnawat, R. K. (2015). Competency Based Human Resource Management: Concepts, Tools,

Techniques, and Models: A Review. 03(05)

Rodriguez, D., Patel, R., Bright, A., Gregory, D., & Gowing, M. (2002). Developing competency
models to promote integrated human resource practices. Human Resource Management: Published in
Cooperation with the School of Business Administration, The University of Michigan and in Alliance
with the Society of Human Resources Management, 41(3), 309–324.

Saltz, J. S., & Grady, N. W. (2017). The ambiguity of data science team roles and the need for a data
science workforce framework. 2017 IEEE International Conference on Big Data (Big Data), 2355–
2361. https://doi.org/10.1109/BigData.2017.8258190

Sart, G., & Yildiz, O. (2022). Digitalism and Jobs of the Future [Chapter]. Digital Transformation
and Internationalization Strategies in Organizations; IGI Global. https://doi.org/10.4018/978-1-7998-
8169-8.ch001

Shaikh, R., Phulkar, N., Bhute, H., Shaikh, S. K., & Bhapkar, P. (2021). An Intelligent framework for
E-Recruitment System Based on Text Categorization and Semantic Analysis. 2021 Third
International Conference on Inventive Research in Computing Applications (ICIRCA), 1076–1080.
https://doi.org/10.1109/ICIRCA51532.2021.9544102

Shen, Y., & Liu, J. (2021). Comparison of Text Sentiment Analysis based on Bert and Word2vec.
2021 IEEE 3rd International Conference on Frontiers Technology of Information and Computer
(ICFTIC), 144–147. https://doi.org/10.1109/ICFTIC54370.2021.9647258

Sibarani, E. M., Scerri, S., Morales, C., Auer, S., & Collarana, D. (2017). Ontology-guided Job
Market Demand Analysis: A Cross-Sectional Study for the Data Science field. Proceedings of the
13th International Conference on Semantic Systems, 25–32.
https://doi.org/10.1145/3132218.3132228

131
Strusani, D., & Houngbonon, G. V. (2019). The Role of Artificial Intelligence in Supporting
Development in Emerging Markets [Brief]. World Bank. https://doi.org/10.1596/32365

Swift, M. (2021). What Makes a Great Coach? [Corporate]. Https://Www.Crossfit.Com/pro-

Coach/What-Makes-a-Great-Coach. https://www.crossfit.com/pro-coach/what-makes-a-great-coach

Tang, Y., Bai, S., & Cui, L. (2022). An Empirical Study on the Deficiencies and Optimization of the
Management System of Tourist Attractions Based on Human Resource Management. Advances in
Multimedia, 2022, e2133830. https://doi.org/10.1155/2022/2133830

Tarraf, D., Shelton, W., Parker, E., Alkire, B., Carew, D., Grana, J., Levedahl, A., Leveille, J.,
Mondschein, J., Ryseff, J., Wyne, A., Elinoff, D., Geist, E., Harris, B., Hui, E., Kenney, C.,
Newberry, S., Sachs, C., Schirmer, P., Warren, K. (2019). The Department of Defense Posture for
Artificial Intelligence: Assessment and Recommendations. RAND Corporation. https://doi.org/
10.7249/RR4229

UCLA: Statistical Consulting Group. (n.d.). FAQ: What are the differences between one-tailed and
two-tailed tests? Retrieved January 28, 2023, from https://stats.oarc.ucla.edu/other/mult-
pkg/faq/general/faq-what-are-the-differences-between-one-tailed-and-two-tailed-tests/

United Nations Industrial Development Organization. (2002). UNIDO Strengthening Organizational

Core Values and Managerial Capabilities—Part One Competencies. UNIDO Human Resource
Management Branch.

Valecha, R., Sharman, R., Rao, H. R., & Upadhyaya, S. (2013). A Dispatch-Mediated
Communication Model for Emergency Response Systems. ACM Transactions on Management
Information Systems, 4(1), 2:1-2:25. https://doi.org/10.1145/2445560.2445562

Valecha, R., Srinivasan, S. K., Volety, T., Kwon, K. H., Agrawal, M., & Rao, H. R. (2021). Fake
News Sharing: An Investigation of Threat and Coping Cues in the Context of the Zika Virus. Digital
Threats: Research and Practice, 2(2), 1–16. https://doi.org/10.1145/3410025

Valecha, R., Volety, T., Rao, H. R., & Kwon, K. H. (2021). Misinformation Sharing on Twitter
During Zika: An Investigation of the Effect of Threat and Distance. IEEE Internet Computing, 25(1),
31–39. https://doi.org/10.1109/MIC.2020.3044543

Varelas, G., Lagios, D., Ntouroukis, S., Zervas, P., Parsons, K., & Tzimas, G. (2022). Employing
Natural Language Processing Techniques for Online Job Vacancies Classification. In Artificial
Intelligence Applications and Innovations. AIAI 2022 IFIP WG 12.5 International Workshops, 333–
344. Springer International Publishing. https://doi.org/10.1007/978-3-031-08341-9_27

Vathanophas, V. (2007). Competency Requirements for Effective Job Performance in Thai Public
Sector. Contemporary Management Research, 3(1), Article 1. https://doi.org/10.7903/cmr.49

Vijayarani, S., & Janani, R. (2016). Text Mining: Open-Source Tokenization Tools: An Analysis. 3,
37–47.

132
Vo, N. N. Y., Vu, Q. T., Vu, N. H., Vu, T. A., Mach, B. D., & Xu, G. (2022). Domain-specific NLP
system to support learning path and curriculum design at tech universities. Computers and Education:
Artificial Intelligence, 3, 100042. https://doi.org/10.1016/j.caeai.2021.100042

Votto, A. M., Valecha, R., Najafirad, P., & Rao, H. R. (2021). Artificial Intelligence in Tactical
Human Resource Management: A Systematic Literature Review. International Journal of Information
Management Data Insights, 1(2), 100047. https://doi.org/10.1016/j.jjimei.2021.100047

Wagemann Jr, J. (2020). The United States Air Force and Artificial Intelligence: Moving Forward by
Learning from Past Technology Implementation. Air Force Fellows Program Maxwell AFB.

Wang, X., Parameswaran, S., Bagul, D. M., & Kishore, R. (2018). Can online social support be
detrimental in stigmatized chronic diseases? A quadratic model of the effects of informational and
emotional support on self-care behavior of HIV patients. Journal of the American Medical
Informatics Association, 25(8), 931–944. https://doi.org/10.1093/jamia/ocy012

Yu, R., Das, S., Gurajada, S., Varshney, K., Raghavan, H., & Lastra-Anadon, C. (2021). A Research
Framework for Understanding Education-Occupation Alignment with NLP Techniques. Proceedings
of the 1st Workshop on NLP for Positive Impact, 100–106. https://doi.org/10.18653/v1/2021.nlp4p
osimpact-1.11

Zhu, J., Zerbe, E., Ross, M. S., & Berdanier, C. G. P. (n.d.). The Stated and Hidden Expectations:
Applying Natural Language Processing Techniques to Understand Postdoctoral Job Postings.

133
ESSAY III. VETERAN TALENT WITHIN DATA SCIENCE: AN EXPLORATORY

RESUME ANALYSIS ON THE EMPLOYABILITY

OF ACTIVE-DUTY VETERANS

134
 CHAPTER 1: ABSTRACT

United States military veterans face employment challenges as they transition from

military service to the civilian sector. For one, communicating one’s skills and worth has

challenged both applicants and recruiters in understanding prior military experience, especially

for those who served within a combat capacity. Coincidentally, Artificial Intelligence (AI) and

Machine Learning (ML) capabilities have grown across the human resource career field to assist

in talent management endeavors. This has enabled Human Resource (HR) professionals and

applicants to encounter newer capabilities to assist in automating the quantification of

psychometric properties (analytic thinking, tone, authenticity, and communication style)

presented within resumes. Accordingly, newer approaches to reviewing veteran resumes have

come to fruition. Understanding these capabilities are relatively new, this paper seeks to add to

the growing literature interest in leveraging hybrid approaches to analyze military experience

within veteran resumes claiming to have skills within the data science career field. Specifically,

this paper leverages LIWC-22, human-coding, and Word2Vec analyses to investigate veteran

resumes and their employability.

135
 CHAPTER 2: INTRODUCTION

Decades of war and conflict within Iraq, Afghanistan, and other forward-deployable

countries has brought renewed attention to employment challenges of United States veterans.

Historical troop-level drawdowns yielding the lowest number of service members in Afghanistan

have vectored many veterans from a variety of career fields homeward toward the United States,

some pursuing endeavors outside of the military (Garamone, 2021). In the wake of a of troop-

level drawdown in Afghanistan and Iraq, it is likely the number of veterans leaving the service

and pursuing civilian employment will increase (Keeling et al., 2019).

Although the Bureau of Labor Statistics provides inferences on the state of

unemployment for veterans, Keeling et al. (2019) observe that these statistics may be misleading.

Statistics reported by the Bureau of Labor Statistics have indicated a record 2.5% low-rate of

veteran unemployment (January 2023) and marginal difference in unemployment rates between

veterans and nonveterans. The survey methodology investigates 60,000 households per month

and provides a cross-sectional comparison which could represent a small portion of veterans

(Loughran, 2014; Keeling et al., 2019). Furthermore, when observing the different populations of

veterans, the percentage of unemployment varies from generation to generation. For instance, by

averaging the 2022 “Employment Situation” data from the Bureau of Labor Statistics, Gulf War

II Era veterans (any who have served after September 11th, 2001) represent the largest

percentage within the veteran population making up 44.49% (see Figure 1).

136
 2022 Average Unemployment Population Percentage

18.86

44.49
10.63

26.15

Gulf War II Era Gulf War I Era WWII, Korean War, Vietnam Era Other Service Periods

Figure 1. 2022 Average Unemployment Population Percentage

The total unemployment percentage averaged to about 2.83% in 2022, Gulf War II Era

veterans’ average unemployment calculated to about 3.11%. Understanding other generations

may have had higher unemployment rates, the focus of this research seeks to investigate the Gulf

War II Era veterans seeking employment within the civilian sector given its population size and

relevancy toward veterans entering the civilian labor force. Thus, within this exploratory

research, we seek to further understand factors which may contribute to this population’s

unemployment by investigating how Gulf War II Era veterans market their experience.

Commonly cited reasons for high unemployment durations for these veterans, relative to non-

veteran unemployment, include employer discrimination, skill mismatching, mental health

conditions, physical disabilities, and inept job search strategies (Loughran, 2014).

Upon entering the civilian labor force, veterans are faced with the challenge of

communicating the unique skills, experiences, and achievements earned throughout their military

career in such a way that is marketable to a civilian employer within their resume. Although the

Department of Defense provides a Transition Assistance Program and Veteran Affair's benefits

137
to equip veterans with tools for communicating their skills via resume modules and curriculum,

veterans have claimed that these programs have done little to assist in truly preparing them for

drafting effective resumes and translating military experience (Harrod et al., 2023).

Coincidentally, the unemployment rate of Gulf War II Era veterans continues to remain above

the average veteran unemployment, and there have been claims that there is very little research

surrounding United States military veterans and employability within the civilian sector and that

there is potential for this research interest to grow (Goss, 2020; Stone, 2016).

With the Bureau of Labor Statistics projecting a 36% growth of the data science career

field between 2021 and 2031, veterans departing the military may see opportunity to start

marketing themselves to this career field. This growth within the data science career field is

considered significantly faster than the average for all other occupations listed throughout the

United States within O*Net. Understanding military veterans departing Active Duty may pursue

a variety of different career fields, we seek to explore Gulf War II Era resumes that have

deliberately communicated data science skills. In doing so, this paper investigates the

employability of military veterans that have communicated data science skills by analyzing the

civilian unemployment durations throughout resumes. To accomplish this analysis, we consider

psychometric variables to gauge potential psychological considerations in the veteran’s writing

style. We further incorporate a veteran’s career identity by observing data science competencies

(Votto et al., 2023). Lastly, we factor in control variables related to the veteran with regards to

their military rank, years served on active duty, holding a security clearance, college education,

managerial experience, data science experience, and total number of civilian jobs held. Once

these variables are calculated, we proceed to explore any relationships that may exist between

138
the psychometric variables, data science competencies, controls, and civilian unemployment

durations.

By exploring these relationships, we aspire for this paper to produce three primary

contributions. First, we add to the growing pool of research surrounding employability and

United States military veteran. We also demonstrate the capabilities of leveraging an academic

IT artifact to conduct research on a unique and important demographic within the United States

labor force. Secondly, this research fortifies academic interests in exploring psychometric factors

and data science competencies which may contribute to employability of veterans. Lastly, this

paper provides insight to veterans and employers within the data science community on potential

strategies to better communicate and interpret military experience.

The arrangement of this paper is organized into the following sections. Section 2

discusses previous literature which has investigated veteran unemployment. We also highlight

literature that has leveraged LIWC-22 to conduct language analysis. Additionally, we introduce

the theoretical background and research model which guides the hypotheses development of this

research. Section 3 discusses the methodology of our analysis. Within this section we discuss

data collection and cleaning techniques. We further discuss analysis techniques leveraged to

explore the relationships between unemployment durations and psychometric variables within

resumes. The next section examines the results the analyses and further discusses post-hoc

considerations. Following this, a discussion section provides insight into implications,

limitations, and future research directions of this research. Lastly, the final section provides a

conclusion which summarizes this work and encourages further exploration of veteran

unemployment.

139
 CHAPTER 3: LITERATURE REVIEW AND THEORIETICAL BACKGROUND

Within this section, we introduce the definition of a military veteran and published

research that has contributed to the exploration of veteran employability. We further Fugate et

al.’s (2004) proposed construct of employability which guides our research model.

Previous Military Veteran Employability Research

The United States Department of Veteran Affairs defines a military veteran as an

individual who served within the reserve, guard, or federal active military and were discharged

or released therefrom under conditions other than dishonorable (Department of Veteran Affairs,

2019). Understanding these individuals are equipped with unique skills based on their military

experiences, veteran employability has been a research interest given concerns surrounding

unemployment and challenges to assimilate back into the civilian workforce. Therefore, we

comprehensively reviewed the top-ten pieces of peer-reviewed literature surrounding veteran

employability to understand how this has already been explored and how our research could

potentially bolster this research interest.

Within Table 1, we have summarized nine pieces of literature that have thoroughly

investigated veteran employability. This table highlights the authors and year of publication. It

further provides insight to the sample analyzed, research methods, and findings. When

identifying these pieces of literature, we leveraged a four-step systematic approach modeling

Votto et al. (2021). We first established the database from which the literature would be pulled

from. For this research, we leveraged Google Scholar given its multidisciplinary publication

sources and open-source nature. We then determined search parameters consisting of a date

range and search string. Our scope of interest was publications within the Gulf War II era

timeframe. Therefore, the established date range was any peer-reviewed publication between

140
September 11th, 2001, and March 2023. The search string we used further narrowed the scope by

identifying peer-reviewed literature with the following keywords: “United States”, “Military

Veteran”, and “Employability”. We used the Boolean operator, “and”, to connect the keywords

together and ensure the results contained each of the established keywords. The third step was

extracting the most relevant pieces of literature from Google Scholar. Understanding hundreds of

results are revealed after searching, we filtered the results by “most relevant” and chose the top-

ten pieces of literature which were perceived as “most-relevant” based on the established

keywords. The last step of our literature review consisted of reading each article to confirm it

was related to United States Military Veterans and employability.

Of the identified literature reviewed, four of ten were doctoral dissertations published

between 2016 and 2020. Additionally, three journal articles and two book chapters were

identified. We disqualified one piece of literature, as it was outside the scope of our interest due

to focusing on the Canadian Armed Forces. It was noted that of all the publications within the

Google Scholar database, the top ten most relevant results yielded more dissertations than

academic journal publications. All articles are summarized below in Table 1.

Table 1. Previous Veteran Employability Research

Author(s) Publication Publication Sample Research Findings

Year Type Method
Niu et al. 2022 Journal Article Student Survey analysis Student Veterans had
Veterans on perceived slightly lower perceived
and employability employability and ambition
Civilians than active duty members
(37 and civilians.
Military
Veterans,
48
Nonvetera
ns)

141
 Table 1. Previous Veteran Employability Research Continued
Stone 2016 Dissertation 511 Survey analysis Perceptions of hiring
MTurk to gauge how managers play a large role
Participant human resource in applicant selection—
s managers military stereotypes were
perceive military supported to affect how
veterans veterans are perceived
James 2017 Dissertation 10 Female Interview 5 prevalent themes
Military Analysis on identified: 1) importance of
Officers military networking, 2) workplace
transition to culture, 3) role adjustment,
civilian jobs 4) military stereotypes, and
5) lessons learned/shared
Fossey et al. 2019 Book Chapter Military- Systematic Developed set of principles
to-Civilian policy review of to assist in developing
policies published NATO effective policies for
developed guidance military transitioning to
by NATO surrounding civilian life
military
transition and
benefits
Goss 2020 Dissertation 15 Post Qualitative Identified four themes:
9/11 Army Descriptive 1)feeling of uncertainty
Officers Study (interviews shifting roles, 2) reliance on
and focus inner self and veteran
groups) on identity to build confidence,
military 3) establish support
transition to network, and 4) remain
civilian mission-focused to prepare
employment, for transition
one-on-one
interviews &
focus groups
Rutter 2020 Dissertation 10-15 Qualitative Identified seven themes:
prior Descriptive 1) planning and preparation,
military Analysis (online 2) work recommendations,
non- questionnaire, 3) cultural change, 4)
commissio semi-structured competencies, 5)
ned officer interview) on networking, 6) role in the
veterans motivation workplace, and 7) support
influencing systems
military-to-
civilian transition
Lampka and 2017 Journal Article 5 Veteran Mixed Methods 3 Qualitative findings which
Kowaleski Resumes (Qualitative hindered successful
analysis for employment within civilian
resumes, workforce: 1) military
quantitative for linguistics, 2)skills were
unemployment embedded, deemphasized
durations) importance, 3) no
management experience
communicated.

142
 Table 1. Previous Veteran Employability Research Continued
3 Quantitative finding:
regression analysis showed
a strong positive
relationship between
civilian and unemployment
rates
McLennan 2021 Book Chapter Bureau of N/A Overview of military
Labor occupations and pay grades
Statistics and their effect on
military employability
occupation
listing
Perkins et al. 2022 Journal Article 854 Survey Analysis Two key findings: 1) Career
Veterans to gain insight on planning/exploration and
participati components that resume writing were
ng in predict significantly important to a
employme employment and veteran in search of a job
nt help veterans after the military, 2)
programs find jobs veterans leveraging a
mentor or coach were 2.2
times more likely to finda
job 6- to 9-months after
separation

Theoretical Background

In an attempt to bolster academic literature with regards to military veteran

employability, we reference Fugate et al.’s (2004) psycho-social construct of employability to

guide our research. This specific framework is considered an aggregate multidimensional

construct (MDC). Law et al. (1998) define MDCs as directionality of the relationships between a

construct and its dimensions. In this instance, this MDC refers to one’s employability (the

construct) and its relationship to associated dimensions. Fugate et al. (2004) view this

relationship as aggregate, such that the dimensions cause or create the construct (one’s

employability). Therefore, within this research, we observe the dimensions of psychological

human capital (PHC) and career identity (CI) for each resume (Fugate et al., 2004; Ngoma and

Ntale, 2016). The employability construct and its dimensions are summarized in Figure 1.

Prior to discussing the individual dimension, we first seek to define the concept of

employability and the control variables associated with each resume. The construct of

143
employability embodies individual characteristics which affect how individuals’ interface within

work environments. Employability is seen as person-centric, such it reflects characteristics that

foster adaptive behavior and cognition with regards to communicating skills and performing

tasks (Fugate et al., 2004). To measure employability, unemployment durations have been

calculated to measure relationships between personal characteristics and one’s employability.

For instance, Lampka and Kowaleski (2017) leverage a mixed methods approach to gain better

understanding of the relationships which exist between veteran unemployment and skills

communicated within resumes.

Figure 2. Employability Construct: Modified from Fugate et al (2004) and Ngoma and
Ntale (2016)

Psychological Human Capital Elements and Employability

Psychological Human Capital (PHC) refers to the positive psychological traits of an

individual which related to their career and job affiliation. Essentially, these are the traits which

connect employees to employment. Literature connects this dimension with a multitude of

qualitative elements such as tonality, analytical thinking, and authenticity when communicating

experience (McEvoy et al., 2005; Nisha and Rajasekaran, 2018). We also consider two

communication styles observed within a resume derived from Osgood et al. (1957). The first is

144
active orientation which refers to diction that imply active engagement and initiative. The second

is week diction which refers to the use of language related to submission, vulnerability and

withdrawl (Lim 2002; Stone and Hunt, 1963; Osgood et al., 1957). Therefore, PHC is broken

into two branches: 1) psychometric variables and 2) communication styles. Within this

subsection, we will dive into literature surrounding each element of PHC and derive hypotheses

based on findings and inferences.

Regarding the PHC element of analytical thinking, we observe two pieces of literature

which highlight its relationship to employability and derive our first hypothesis. Nisha and

Rajaeskaran (2018) conduct a literature review which highlights a variety of traits that are

considered desirable skills for potential employees seeking new employment opportunities.

Among the skills listed, analytical thinking ranked within the most marketable that employers

seek within prospective employees. Furthermore, Kulkarni et al. (2017) designs an employability

skill matrix which studies engineering graduates and their ability to problem solve. Through this

matrix, increased analytical thinking and expertise in technical knowledge yielded improved

employability within the student sample. Thus, we develop the following hypothesis regarding

analytical thinking and employability:

H1a: Analytical thinking is positively related to employability.

The next observed element of PHC is authenticity. This element refers to the perceived

honesty and genuineness of input (Boyd, 2022). Within literature, effective self-branding

requires the applicant and employer to be authentic and clearly understand the “why”

surrounding both sides of the hiring process (Ainspan and Saboe, 2021). Specifically, an

employee must be true and honest with the skills they bring to the labor pool and communicate

145
with them effectively. Contrarily, the employer must effectively communicate desired skills and

expectations of an employee.

H1b: Authenticity is positively related to employability.

Tonality is the third element within PHC. This element refers to the social and

psychological responses to emotional (positive or negative) stimuli (Cohn et al., 2004). In the

context of employability, Fugate and Kinicki (2008) refer to tonality as the cognitive reappraisals

which indicate initiative and proactive behavior, such that positive tones reflect those who are

optimistic and take initiative. Considering the desirability of employees who take initiative and

bring a positive outlook to the organization, we posit the following for further analysis:

H1c: Tone is positively related to employability.

We further consider weak and active orientation with regards to the communication style

of each resume. Active orientation refers to the individual’s ability to react to intellectual

challenges within a space and take initiative (Memarovic et al., 2012). Contrarily, weak

orientation refers to an individual’s inability to seize initiative and lean toward submissive

tendencies (Stone and Hunt, 1963). Given these characteristics, we posit the following regarding

communication styles:

H2a: Active Orientation is positively related to employability.

H2b: Week Orientation is negatively related to employability.

Career Identity Elements and Employability

Fugate et al. (2004) introduce career identity as one’s self-definition within a career

context. This construct is referred to as a cognitive compass which motivates one to adapt and

create opportunities to match work aspirations. This construct provides a strong foundation for

employability through an applicant’s ability to self-reflect on goals and communicate their skills

146
needed to achieve them. Thus, we consider data science competencies addressed by Votto et al.’s

(2023). These competencies were derived from literature and guidance published by IBM and

Deloitte and investigate how veterans communicate their identities within competencies and how

they interact with employability.

With regards to Deloitte, we reference their multidimensional Artificial Intelligence

framework titled “Trustworthy AI™” which postures organizations to develop ethical

technologies (Ammanath, 2022; Mökander and Floridi, 2021). The development of ethical

expectations goes beyond that of technology. The desire for employees to have a grounded sense

of ethics, especially when handling sensitive data, has become a desired competency of

employees. Thus, we extract our first competency, ethics, from this framework. This competency

reflects how employees communicate their ability to enforce equitable and fair practices within

the data science work environment. Additionally, we extract another competency, Artificial

Intelligence, from this framework to address how employees communicate their aptitude to

leverage these intelligent tools. That said, Artificial Intelligence encompasses a variety of topics

related to machine learning and deep learning (Votto et al., 2021). one core competency from

this Deloitte framework to gauge how military veterans communicate their experience and skills

to make themselves employable within the data science career field. Thus, we posit the

following:

H3a: Ethics is positively related to employability.

H3b: Artificial Intelligence is positively related to employability.

The second framework we leveraged to derive competencies was developed by IBM and

is titled “Data Science Skill’s Competency Model”. Although this framework contains twenty-

eight data science competencies across seven groups, we extracted three primary considerations

147
for this analysis. The first competency we adopted was problem solving. This refers to a

prospective employee’s ability to communicate skills pertaining to understand analytic cycles,

identify business problems, and develop solutions. The second competency we extracted from

IBM’s model was “machine learning”. This competency refers to an applicant’s ability to

demonstrate and communicate an understanding of machine learning algorithm principles and

statistical components. Lastly, we extracted “statistics” from the IBM model as our last

competency to consider career identity. This competency refers to an applicant’s ability to

understand probability theory, sampling, and other vital statistical components (IBM, 2020).

From these competencies, we posit the following regarding employability:

From the aforementioned competencies, an applicant’s data science career identity is

derived. The ability to reflect on past experiences and communicate skills related to this career

field bolsters career identity within this career field and promotes potential to enhance

employability. Thus, through the posited hypotheses, we hope to further explore how veterans

communicate their skills and understand how these competencies are related to employability.

Figure 2 summarizes the hypotheses and research model driving this research initiative.

Figure 3. Employability Research Model

148
 CHAPTER 4: METHODOLOGY

Within this section, we describe the process of data acquisition and cleansing, descriptive

statistics, and analytic techniques. We utilize Meyer ‘s (2019) three-phased methodology when

conducting our analysis. The first phase consists of data collection and preparation. The second

phase discusses how the dimensions are organized. The last phase is the analysis and reporting

component. The following subsections provide deeper insight into each of the phases.

Data Collection and Preparation

With regards to the source of our resumes, we manually extracted resumes from Indeed

(Muthyala et al., 2017). Understanding APIs exist to assist and streamline resume extraction

endeavors, we pursued the manual route given the specific demographic (Gulf War II Era

Veterans) we targeted for our analysis.

We acknowledge our desired data may contain sensitive information relating to military

veterans. Resumes may contain personal emails, telephone numbers, and addresses related to the

individual. Therefore, we pursued an Information Review Board to review our proposed research

interests and confirm confidentiality measures would ensure ethical research practices were

maintained and respected. Upon receiving IRB approval, we continued to analyze resumes from

Indeed.

Following similar strategies within systematic literature reviews (Votto et al., 2021), we

conducted a nine-step filtering process. This was a nine-step process (summarized in Table 2)

ultimately yielded 395 resumes for our analysis. Within the first step, we typed “data science”

into Indeed’s search field yielding 1,467,289 results. We then limited the search to “Job Title”

and “Skills” within the United States. These parameters yielded 130,094 resumes. We further

narrowed our parameters down to resumes that had been updated within the last six months,

149
excluding any candidates that had been contacted within the last 30 days. This filtration yielded

50,105 resumes. For the fifth step, we observed resumes that were “read to work now”, which

yielded 27,725 resumes. We then selected candidates that were “active or former military” which

yielded 1,125 resumes. We then manually reviewed each resume and extracted prior active-duty

Gulf War II Era veterans and excluding any members who were still serving on active duty. This

yielded 504 resumes. For the eighth step, we continued to manually review each resume and

eliminate any National Guard or Reserve members, yielding 430 resumes. For the last step, we

eliminated any personally identifiable information from each resume to ensure anonymity and

recruited an active-duty military member to help audit the resumes to ensure they veterans served

on active duty and were not currently in the reserves or national guard. This yielded our final

resume pool of 395 for our analysis.

Data Operationalization

To compute the PHC variables, we first leveraged LIWC-22 (Parameswaran et al., 2022)

to quantify the presence of analytic thinking, authenticity, and tone within a resume. This

program leverages a dictionary-based approach to quantify predefined psychological categories

whose constructs have both been tested for validity and reliability (Boyd, 2022; Parameswaran et

al., 2022; Boyd and Pennebaker, 2015). Additionally, we used the General Inquirer to find the

psychological indexes of Osgood et al.’s (1957) active and weak lexicon. In doing so, we

constructed a machine algorithm to observe the presence of Osgood et al.’s (1957) active and

week orientation lexicons and derive a percentage, on the same scale as the LIWC-22 variables.

The reliability and validity of Osgood et al.’s (1957) has been tested throughout literature

(Stoklasa et al. 2017; Stone and Hunt, 1963). All PHC variables are represented as a percentage,

150
relative to the word count of each resume. Table 3 summarizes all continuous variables observed

within this analysis. The last two variables located within Table 3 are later discussed.

Table 2. Continuous Variables

Variable Name Description Formula/Dictionary Variable Influencing
Description Scale Literature

Analytic Metric of formal and Based on several categories of 0.00 – Pennebaker et

logical thinking function words. Low analytic 100.00 al.(2014);
score indicates using language Boyd and Pennebaker
that is intuitive and personal (2015);
(friendly). High analytic score Sterling et al. (2019)
is correlate with reason and
grades (cold/rigid).
Authentic Perceived honesty and Refers to how spontaneous 0.00 – Pennebaker et al.
genuineness of writing someone speaks. Derived form 100.00 (2003)
studies surrounding deception Kalichman and Smyth
and honesty. Low score (2021)
indicates social caution (self- Markowitz et al. (in
monitoring). High score press)
indicates spontaneous with
little-to-no social inhibitions
Tone The degree of positive Higher score indicates more 0.00 – Mehl and Pennebaker
or negative positive words. Lower score 100.00 (2004)
inflection/tone within indicates more negative tones. Vergani et al. (2021)
writing
Weak Language depicting Higher score indicates 0.00 – Lim (2002)
potency of weakness, semantic diction gravitates 100.00 Stone and Hunt.
submission of toward the use of words that (1963)
authority/power, convey weakness or capacity Osgood et al. (1957)
dependence on others, of inaction
vulnerability to others,
or withdrawal
Active Diction that imply an Higher score indicates 0.00 – Memarovic et al.
active semantic diction gravitates 100.00 (2012)
orientation/engagement toward the need for Lim (2002)
intellectual encounters or Osgood et al. (1957)
physically challenges within a
space
Total Number Indicates the Total Quantified how many civilian 0.00 – --
of Civilian Jobs Number of Civilian jobs the veteran indicated 22.00
(CONTROL) Jobs Held within resume

Regarding the CI variables, we consider the private sector data science competency

dictionaries established and validated by Votto et al. (2023). Although, this research leverages

these dictionaries to evaluate resumes instead of job descriptions. These dictionaries were

derived from data science job descriptions scraped from Glassdoor. Votto et al. (2023) utilized a

151
Word2Vec model to extract the top-10 most similar related to each data science competency

(ethics, machine learning, artificial intelligence, problem solving, and statistics). We leveraged

MAXQDA to run our quantitative content analysis to identify which resumes had a presence of

each data science competency (Oliveira et al., 2015). If the resume contained words from the

dictionary, it was coded with a “1”; otherwise “0”.

To computer our dependent and control variables, we hired a research assistant to

calculate the required variables. Regarding the dependent variable, the research assistant

calculated how many weeks each individual resume indicated an employment gap between

civilian jobs. Upon completion, each resume had a number to reflect the number of weeks the

individual indicated an employment gap within their resume.

Additionally, the research assistant extracted the control variables for our model. There

were seven controls total we considered with our model. The resumes did not include

information regarding the veteran’s gender. Therefore, we extracted other demographic

considerations involving education and work experience (Cole et al. 2004). The first control we

considered was whether the veteran indicated earning a college degree (1 = degree

communicated, 0 = otherwise). The second control variable consisted of acknowledging whether

the veteran communicated experience with holding a security clearance (1 = clearance

communicated, 0 = otherwise). The third control variable military rank of the veteran (1 =

officer, 0 = enlisted). The fourth control consisted of quantifying how many years the individual

served on active duty (1 = over 10 years, 0 = less than 10 years). The fifth and sixth control

variables examined the most recent job the individual held. The fifth confirmed if their most

recent job was managerial in nature (1 = managerial, 0 = not managerial), while the sixth

confirmed the veteran’s most recent job was related to data science (1 = data science related, 0 =

152
otherwise). The last control variable we considered was how many civilian jobs the veteran held.

This variable is expanded upon in Table 3, while the remaining categorical variables are

summarized in Table 4.

Table 3. Categorical Variables

Variable Name Description Variable Nomenclature

College Graduate Indicated College Degree Earned 0 = No Degree Indicated

Clearance Indicated Clearance Held
Rank The paygrade indicated within
resume
Years Active Duty How long the resume candidate
served on Active Duty
Previous Job (Managerial) Indicated previous job held was
managerial in nature
Previous Job (Data Science) Indicated previous job held was
related to data science
Data Science Competencies Presence of each competency within
(Ethical, Machine Learning, AI, entire resume
Problem Solving, Statistics)
1 = Degree Earned
0 = No Clearance Indicated
1 = Secret Clearance
0 = Enlisted (E1 – E9)
1 = Officer (O1 – O6+)
0 = 0.00 – 9.99 Years Time in
Service
1 = 10.00+ Years Time in Service
0 = Not Managerial
1 = Managerial
0 = Note Data Science Related
1 = Data Science Related
0 = Not Present
1 = Present

Descriptive Statistics

Regarding the descriptive statistics, we leveraged SPSS to analyze our frequency tables

and correlation matrix ( Kim et al., 2005). Table 5 and Table 6 report the descriptive statistics of

the variables of our interest. We also conducted a Spearman Correlation Test to gauge the

correlations between the variables of our interest. There were no variables that shared a high

correlation as the highest was .392 (Parameswaran et al., 2022; Valecha et al., 2021; Venkatesan

et al., 2021). The Spearman Correlation graph can be found in Appendix A.

153
 Table 4. Descriptive Statistics: Continuous Variables
Descriptive Statistics
Minimum Maximum Mean Variance
Analytic 83.53 98.76 94.2468 6.507
Authentic 4.14 94.67 31.6974 201.691
Tone 3.61 98.56 55.8252 394.867
Weak .00 1.662 .25 .064
Active .58 9.04 3.12 1.54
Total_Num_Civ_Jobs 0 22 3.63 6.12

Table 5. Descriptive Statistics: Categorical Variables

Statistics
Active Duty

Intelligence
Clearance

Job (Data
Graduate

Artificial
Learning

Statistics

Previous

Previous
Machine

Problem
Security

Solving
College

Ethics
Years
Rank

Job
Mean .81 .08 .44 .15 .45 .34 .29 .50 .40 .30 .45
Variance .15 .07 .25 .13 .25 .23 .21 .25 .24 .21 .25

Analytic Techniques

In order to test our established hypothesis, we followed previous research techniques.

Due to the violation of normality within our variables, negative binomial regression has been

suggested as a method to analyze our variety of variables given the robust nature of this analysis

(Valecha et al., 2021; Osgood, 2000). Therefore, each of the hypotheses were tested using

negative binomial regression. The equation and results of this analysis are as follows:

𝐿𝐿𝐿𝐿𝐿𝐿 (𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈𝑈 𝐺𝐺𝐺𝐺𝑝𝑝, 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊)

= 𝛽𝛽0 + 𝛽𝛽1 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 𝑇𝑇ℎ𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛽𝛽2 𝐴𝐴𝐴𝐴𝐴𝐴ℎ𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒𝑒 + 𝛽𝛽3 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇

+ 𝛽𝛽4 𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊𝑊 + 𝛽𝛽5 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 + 𝛽𝛽6 𝐸𝐸𝐸𝐸ℎ𝑖𝑖𝑖𝑖𝑖𝑖 + 𝛽𝛽7 𝑀𝑀𝑀𝑀𝑀𝑀ℎ𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖𝑖

+ 𝛽𝛽8 𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴𝐴 + 𝛽𝛽9 𝑃𝑃𝑃𝑃𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜 + 𝛽𝛽10 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆

+ 𝛽𝛽11 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 + 𝛽𝛽12 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 + 𝛽𝛽13 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 + 𝛽𝛽14 𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌𝑌

+ 𝛽𝛽15 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 + 𝛽𝛽16 𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷𝐷 + 𝛽𝛽17 𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇𝑇

154
 Table 6. Main Model Results

Psychological Human Capital (Psychometric Variables)

Analytic Thinking (-.065)
~.0480
Authenticity (.009)
~.0074
Tone (-.048)***
~.0061
Psychological Human Capital (Communication Style)
Weak Disposition (-.315)
~.4706
Active Lexicon (-.517)***
~.1083
Ethics (-.544)*
(Glassdoor) ~.2261
Machine Learning (-1.317)***
(Glassdoor) ~.2474
Artificial Intelligence (-1.231)***
(Glassdoor) ~.2534
Problem Solving (-.169)
(Glassdoor) ~.2116
Statistics (2.886)***
(Glassdoor) ~.2665
Control Variables
College Graduate (2.917)***
~.3229
Military Rank (-2.263)***
~.4606
Security Clearance (.770)***
~.2170
Years Active Duty (-2.481)***
~.3595
Previous Job (-.905)**
(Managerial) ~.2893
Previous Job (Data (.094)
Science) ~.719
Total Number of (.386)***
Civilian Jobs ~.0659
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001
Note: () covariate; ~standard error

155
 CHAPTER 5: RESULTS

The results of the negative binomial regression analysis are summarized in Table 7.

These results indicate significant negative effects on civilian unemployment for several key

variables. Regarding PHC, tone (p-value < .001) and active orientation (p-value < .001) were

statistically significant and had a negative effect on civilian unemployment. This indicates the

more positive and optimistic the member’s tone is throughout their resume, the more employable

they could be. Similarly, leveraging an active disposition when communicating one’s worth

positively affects employability. Regarding CI, three of the five data science competencies

negatively impacted unemployment durations. These competencies are ethics (p-value < .05),

machine learning (p-value < .001), and artificial intelligence (p-value < .001). This indicates that

the incorporation of ethical behavior and skills related to artificial intelligence and machine

learning bolster one’s employability within their resume when pursuing data science jobs.

Robustness Checks

To expand upon our initial findings, we conducted three robustness checks. This

subsection provides deeper insight into these robustness checks. The first robustness check

considered was an alternate measure test. This investigates the consistency of results (Balance et

al., 1987) within an analysis by considering a different dependent variable. Within this test, we

observed the number of weeks unemployed since their last job listed within the resume instead of

the total number of weeks unemployed as a civilian. Through this test, we confirmed a member’s

tone (p-value < .05), and mention of ethics (p-value < .001), machine learning (p-value < .001),

and artificial intelligence (p-value < .001) continued to negatively affect unemployment duration.

This implies consistency between data science competencies and the psychometric variable of

tone. Table 8 summarizes the findings.

156
 Table 7. Alternate Measures Analysis
Psychological Human Capital (Psychometric Variables)
Analytic Thinking (-.070)
~10.0846
Authenticity (-.047)
~.0193
Tone (-.025)*
~.0061
Psychological Human Capital (Communication Style)
Weak Disposition (-13.370)***
~.2.5266
Active Lexicon (-.154)
~.1083
Career Identity (Data Science Competencies)
Ethics (-.1.647)***
(Glassdoor) ~.5033
Machine Learning (-6.160)***
(Glassdoor) ~1.0844
Artificial Intelligence (-5.026)***
(Glassdoor) ~.9975
Problem Solving (-.560)
(Glassdoor) ~.5089
Statistics (6.603)***
(Glassdoor) ~.7967
Control Variables
College Graduate (3.627)***
~.7541
Military Rank (4.446)***
~1.1200
Security Clearance (-4.501)***
~.6536
Years Active Duty (-37.498)
~ --
Previous Job (2.069)**
(Managerial) ~.7176
Previous Job (Data (-3.849)***
Science) ~.5671
Total Number of (-1.054)***
Civilian Jobs ~.1804
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001
Note: () covariate; ~standard error

Secondly, we conducted a dictionary comparison analysis where we leveraged new data

science competency dictionaries from a different private-sector job description source, LinkedIn.

Within this test, we sought to explore how using a different source for our dictionaries could

affect the significance of our established variables from the main model. We followed the same

Word2Vec method mentioned in Votto et al. (2023) and generated a new list of the top-10 most

157
similar words (Deguhci et al., 2021; Brumer et al., 2017). After eliminating duplicate job

descriptions and tokenizing the words, we trained the model on 1,000 data science job

descriptions from LinkedIn. The results of the Word2Vec “most_similar” method can be found

in the Appendices B and C the Glassdoor and LinkedIn datasets. Once the dictionaries were

created, we coded each resume based on the presence of each dictionary (1 = present, 0 =

otherwise). Upon successfully coding the resumes, we conducted another Spearman Correlation

test to ensure the new variables were not multicollinear with the established controls and other

independent variables. There were no variables that shared a high correlation as the highest was

.282 (Parameswaran et al., 2022; Valecha et al., 2021; Venkatesan et al., 2021). We then reran

our original model, substituting the original private sector dictionaries from Glassdoor with the

dictionaries created from LinkedIn. The results of this secondary analysis yielded considerable

and consistent results. To summarize, both tone (p-value < .001) and active orientation (p-value

< .001) stayed true to negatively impacting unemployment durations. Furthermore, the CI

variables of machine learning (p-value < .05) and artificial intelligence (p-value < .001)

remained the same by negatively impacting unemployment durations. Thus, supporting our

original findings within the main model. Table 9 summarizes these findings.

Table 8. Alternate Dictionary Analysis

Psychological Human Capital (Psychometric Variables)
Analytic Thinking (-.043)
~..0487
Authenticity (.012)
~.0073
Tone (-.040)***
~.0054
Psychological Human Capital (Communication Style)
Weak Disposition (.162)
~.4743
Active Lexicon (-.639)***
~.1139
Career Identity (Data Science Competencies)
Ethics (-.326)
(LinkedIn) ~.3128

158
 Table 8. Alternate Dictionary Analysis Continued
Machine Learning (-.562)*
(LinkedIn) ~.2474
Artificial Intelligence (-1.413)***
(LinkedIn) ~.2534
Problem Solving (.693)*
(LinkedIn) ~.2742
Statistics (2.113)***
(LinkedIn) ~.2401
Control Variables
College Graduate (2.917)***
~.3229
Military Rank (-2.263)***
~.4606
Security Clearance (.770)***
~.2170
Years Active Duty (-2.481)***
~.3595
Previous Job (-.905)**
(Managerial) ~.2893
Previous Job (Data (.094)
Science) ~.719
Total Number of (.386)***
Civilian Jobs ~.0659
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001
Note: () covariate; ~standard error

The last robustness check we conducted was an endogeneity (Hausman) test. This test is

designed to investigate the any correlations that may exist between our exploratory independent

variables and error terms (Lu et al., 2018). To examine the endogeneity issue, we acknowledge

it is important to choose an instrumental variable for each model (Venkatesan et al., 2021).

Within this analysis, we consider the entire word count of each resume as the instrumental

variable and consider the Glassdoor data science competencies from Votto et al. (2023).

Leveraging SPSS, we first generated our residuals (Sargan-Hansen J-Statistic) for each model.

This was accomplished by establishing our dependent variable as the data science competency

and word count as our independent variable. We then ran a regression analysis to generate the

error residual for each competency. Once the residuals were generated, we plugged them

residual into our original model, one at a time, to better understand some of the relationships

which may exist. This resulted in five different models which are summarized in Table 10.

159
Models 1 (ethics) and 4 (problem solving) confirmed the variables were not interdependent with

one another by the statistically insignificant results. This confirms that there are no omitted

variable biases caused by the effect word count has on each data science competency. Models 2

(machine learning), 3 (artificial intelligence), and 5 (statistics) did not meet the endogeneity

criteria due to resulting in statistically significant results. That said, all five models shared the

same recurring statistically significant results from the original model of this analysis. Tone,

active orientation, ethics, machine learning, and artificial intelligence continued to negatively

impact unemployment durations, despite the residuals being added to the model. Despite several

of the models not meeting the endogeneity criteria, there was consistency demonstrated within

the results when compared to the original model of this analysis.

Table 9. Results of Endogeneity Test

Model 1: Model 2: Model 3: Model 4: Model 5:
Resume Word Resume Resume Resume Resume
Count on Word Word Word Word
Ethics Count on Count on Count on Count on
ML AI PS Stat
Psychological Human Capital (Psychometric Variables)
Analytic Thinking (-.045) (-.062) (-.045) (-.048) (-.045)
~.0492 ~.0505 ~.0491 ~.0495 ~.0491
Authenticity (.009) (.004) (.009) (.010) (.009)
~.0076 ~.0075 ~.0076 ~.0075 ~.0076
Tone (-.042)*** (-.048)*** (-.042)*** (-.044)*** (-.042)***
~.0068 ~.0061 ~.0066 ~.0068 ~.0067
Psychological Human Capital (Communication Style)
Weak Disposition (-.384) (-.558) (-.386) (-.355) (-.391)
~.4756 ~.4870 ~.4759 ~.4735 ~.4763
Active Lexicon (-.528)*** (-.531)*** (-.524)*** (-.528)*** (-.527)***
~.1114 ~.1133 ~.1117 ~.1106 ~.1116
Career Identity (Data Science Competencies)
Ethics (Glassdoor) (-3.251)* (-.729)** (-.594)** (-.582)* (-.600)**
~1.4050 ~.2386 ~.2269 ~.2275 ~.2272
Machine Learning (Glassdoor) (-1.420)*** (-1.582)*** (-1.423)*** (-1.386)*** (-1.436)***
~.2585 ~.2630 ~.2582 ~.2560 ~.2596
Artificial Intelligence (Glassdoor) (-1.259)*** (-1.271)*** (-4.167)*** (-1.245)*** (-1.262)***
~.2564 ~.2620 ~1.3947 ~.2558 ~.2565
Problem Solving (Glassdoor) (.072) (-.210) (.077) (-1.00) (-.090)
~.2451 ~.2089 ~.2408 ~.6169 ~.2433
Statistics (Glassdoor) (2.882)*** (3.123)*** (2.877)*** (2.881)*** (-1.943)
~.2670 ~.2753 ~.2672 ~.2667 ~2.2438
Control Variables

160
 Table 9. Results of Endogeneity Test Continued
College Graduate (.3.072)*** (2.970)*** (3.083)*** (3.032)*** (3.088)***
~.3386 ~.3388 ~.3390 ~.3369 ~.3394
Military Rank (-2.066)*** (-2.173)*** (-2.052)*** (-2.118)*** (-2.042)***
~.4623 ~.4619 ~.4612 ~.4648 ~.4618
Security Clearance (.956)*** (.756)*** (.969)*** (.907)*** (.973)***
~.2393 ~2237 ~.2383 ~.2685 ~.2389
Years Active Duty (-2.482)*** (-2.596)*** (-2.473)*** (-2.509)*** (-2.476)***
~.3695 ~3793 ~.3695 ~.3691 ~.3697
Previous Job (Managerial) (-1.711)*** (-1.411)*** (-1.731)*** (-1.673)*** (-1.723)***
~.2639 ~.2571 ~.2649 ~.2625 ~.2640
Previous Job (Data Science) (.363) (.305) (.347) (.352) (.368)
~2450 ~.2417 ~.2437 ~.2454 ~.2451
Total Number of Civilian Jobs (.617)*** (.610)*** (.624)*** (.600)*** (6.25)***
~.0715 ~.0685 ~.0720 ~.0705 ~.0720
Residual on Model
Competency Residual (-2.655) (-2.583)*** (-2.898)* (-1.017) (-4.826)*
~1.3592 ~.5569 ~1.3513 ~.7098 ~2.2313
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001
Note: () covariate; ~standard error

Hypotheses Results

As mentioned previously, PHC was represented by the veteran’s tone and active

orientation when communicating their worth within the resume. Simultaneously, CI was

represented by the presence of ethics, machine learning, and artificial intelligence. Each of these

variables negatively impacted unemployment durations within the resume, thus indicating the

potential for enhancing a veteran’s employability. Regarding the control variables, there were

three which negatively impacted unemployment durations. The first was military rank which

implied that if the veteran was an officer, their unemployment duration was less than that of an

enlisted member. The next was the number of years served on active duty. Based on this

analysis, if the veteran served over 10 years in the military, the unemployment duration within

the resume was negatively affected. Lastly, if the military veteran’s most recent job was

managerial in nature, then the unemployment duration was also negatively affected. There is one

control variable worth mentioning that positively affected unemployment durations. The number

of civilian jobs held by the veteran positively affected unemployment durations. This implies that

it hindered the employability of the veteran. This can be explained by the veteran holding

161
numerous jobs and bouncing around from opportunity to opportunity, thus creating more gaps

within their experience pool. Such behavior could have employers shy away from employing this

individual as they’ve demonstrated an inability to remain at a job for a desired duration of time

that would indicate fidelity.

Analytic thinking, authenticity, weak orientation, and problem solving yielded

insignificant results. This implies that these variables may not affect a veteran’s employability.

Furthermore, we had one data science competency which positively affected unemployment

durations. This went against our initial hypothesis that the presence of statistics would positively

impact employability. We speculate that the cause of this is related to the increase of AI and

machine learning capabilities within the data science career field. Although statistics may be a

core component of machine learning practices, it may not be as marketable on its own as the

demand for machine learning engineers and scientists grows. To summarize our findings, Table

11 demonstrates which hypotheses were supported and which were not.

Table 10. Hypotheses Results

Category Variable Hypothesis Significance Hypothesis Supported
Value Hypothesis Not Supported
Statistically Insignificant
Psychological Psychometric Analytic H1a .007 Statistically Insignificant
Variables Thinking
Human
Capital
Authenticity H1b <.001 Statistically Insignificant

Tone H1c .925 Hypothesis Supported

Communication Weak H2a .009 Statistically Insignificant

Styles

Active H2b <.001 Hypothesis Supported

162
 Table 10. Hypotheses Results Continued
Career Data Science Ethics H3a .003 Hypothesis Supported
Identity Competency

Machine H3b .779 Hypothesis Supported

Learning

Artificial H3c .008 Hypothesis Supported

Intelligence

Problem H3d .424 Statistically Insignificant

Solving

Statistics H3e .301 Hypothesis Not Supported

163
 CHAPTER 6: POST-HOC ANALYSIS

Within this chapter, we discuss the two post-hoc analysis we performed to open the AI

black box within military veteran resumes. The first analysis explores the Top-25 data science

technologies listed on O*Net and investigates their presence within veteran resumes. Regarding

the second post-hoc analysis, we conduct a topic modeling analysis to gain insight to potential

topics which are prevalent within resumes.

The Occupational Information Network (O*Net) is a database created by the Department

of Labor to showcase occupations and communicate desired skills related to said occupation

(Peterson et al., 2001). This website seeks to digitize information regarding professions that exist

within the United States to help prospective applicants understand the rising demands (Lennon et

al., 2023). Considering the data science career field is expected to grow 36% between 2021 and

2031 (Bureau of Labor Statistics, 2023), we sought to leverage the Top-25 data science skills

posted within the O*Net database. Upon collecting these variables, we separated them into two

categories: 1) software and 2) programming languages. Table 12 summarizes these variables.

Table 11. Top 25 Data Science Technologies from O*Net

Software (In order of Programming Language
demand) (In order of demand)
Apache Spark Python
Amazon Web Services SQL
Tableau R
Apache Hadoop TensorFlow
SAS Oracle Java
Scala Scikit-Learn
Microsoft Azure PyTorch
Apache Hive C++
Git Pandas
MATLAB Keras
Docker NumPy
Microsoft Excel NoSQL
- PySpark

164
 We then created dictionaries for each category and calculated the existence of each

technological skill within each resume. We quantified how many resumes contained each

technological skill and averaged the number of months each veteran was unemployed, as seen in

Table 13 and Figure 3. To avoid redundancy, we only counted the skill once per resume to

ensure duplication was mitigated.

Table 12. Top Software Related to Resumes

Software (In order of Percentage of Resume Number of Resumes Average Months Unemployed
demand)
Apache Spark 6.58 26 12
Amazon Web Services 16.71 66 6.83
Tableau 22.78 90 5.96
Apache Hadoop 1.77 7 0.58
SAS 6.84 27 10.71
Scala 1.01 4 19.80
Microsoft Azure 8.86 35 6.48
Apache Hive 0.51 2 0
Git 25.06 99 7.17
MATLAB 7.34 29 8.57
Docker 3.04 12 10.62
Microsoft Excel 52.66 208 10.08

165
 O*Net Data Science Software vs Average Unemployment
(Months)
60.00
50.00
40.00
30.00
20.00
10.00
0.00
Apache Amazon Tableau Apache SAS Scala Microsoft Apache Git MATLAB Docker Microsoft
Spark Web Hadoop Azure Hive Excel
Services

Percentage of Resume Average Months Unemployed

Figure 4. Data Science Software Vs Unemployment

Regarding software, Amazon Web Services, Tableau, Git, and Microsoft Excel were the

most prevalent within veteran resumes, each variable demonstrating a lower unemployment

duration when compared to some of the other software. This indicates veterans have more

experience with these technologies and do well to advertise these skills within resumes,

potentially reducing unemployment durations. Of the four mentioned, Microsoft Excel was the

strongest represented with over 50% of veteran resumes communicating experience with this

suite.

In addition to software, we considered the desired programming languages from O*Net.

Table 14 and Figure 4 summarize the findings.

Table 13. Top Programming Languages Related to Resumes

Programming Language Percentage of Resumes Number of Resumes Average Months Unemployed

(In order of demand)
Python 52.41 207 8.42
SQL 56.46 223 9.42
R 2.78 11 12.19
TensorFlow 8.35 33 8.92
Oracle Java 26.08 103 12.48

166
 Table 13. Top Programming Languages Related to Resumes Continued
Scikit-Learn 4.56 18 9.96
PyTorch 1.77 7 7.74
C++ 21.27 84 12.53
Pandas 11.65 46 7.22
Keras 4.05 16 3.26
NumPy 9.37 37 8.08
NoSQL 3.29 13 10.40
PySpark 1.27 5 1.41

Technology vs Average Unemployment Duration

60.00

50.00

40.00

30.00

20.00

10.00

0.00

Perecentage of Resumes Average Months Unemployed

Figure 5. Data Science Programming Languages Vs Unemployment

The most prominent programming languages that exist within veteran resumes include

Python, SQL, and Oracle Java. There is an inverse relationship between these languages and

unemployment. As observed in Figure 4, the aforementioned languages have a stronger presence

within the percentage of resumes and lower unemployment durations. This would imply that

veterans expressing experience with these skills are more marketable with regards to securing a

job within the data science career field.

Upon generating the descriptive statistics, we follow previous research techniques by

observing 3 software and programming languages to investigate the effects they would have on

167
the main model of our analysis (Shah et al., 2022; Li et al., 2021; Kumar et al., 2018). The top-

three software identified (Li et al., 2021), in order of resume presence, are as follows: 1)

Microsoft Excel (data management software), 2) Git (system software), and 3) Tableau (data

visualization software). In addition, the top-three programming languages (in order of presence)

are as follows: 1) SQL, 2) Python, and 3) Oracle Java. The following subsections detail the

results of adding these new variables into the model.

Post-Hoc Investigation: Data Science Software and Programming Languages vs

Unemployment Durations

This subsection provides a deeper insight into the empirical investigation we conducted

to understand the potential effects the top-three software and programming languages may have

on veteran employability.

As seen in Table 15, Microsoft Excel and Git are statistically significant and have a

negative effect on unemployment durations. This implies that there is potential for the inclusion

of these skills within a resume to enhance a veteran’s employability. Contrarily, Tableau was

statistically significant; although, its inclusion within the model positively impacted

unemployment durations. This indicates that the inclusion of this software within a resume may

not benefit veterans when communicating data science skills.

Table 14. Data Science Software Model Results

Psychological Human Capital (Psychometric Variables)

Analytic Thinking (-.075)
~.0554
Authenticity (.014)
~.0083
Tone (-.045)***
~.0074
Psychological Human Capital (Communication Style)
Weak Disposition (2.016)***
~.4743

168
 Table 14. Data Science Software Model Results Continued
Active Lexicon (-.859)***
~.1379
Career Identity (Data Science Competencies)
Ethics (-.813)**
~.2762
Machine Learning (-.457)
~.3106
Artificial Intelligence (-1.248)***
~.3247
Problem Solving (.670)*
~.2776
Statistics (3.533)***
~.3332
Control Variables
College Graduate (4.155)***
~.4310
Military Rank (-2.215)***
~.5253
Security Clearance (.1.909)***
~.3022
Years Active Duty (-3.303)***
~.4365
Previous Job (Managerial) (-2.398)***
~.2893
Previous Job (Data (.006)
Science) ~.2938
Total Number of Civilian (.568)***
Jobs ~.0756
Software
Microsoft Excel (-1.284)***
~.3308
Tableau (.652)*
~.3175
Git (-2.972)***
~.4254
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001
Note: () covariate; ~standard error

We also considered the effects of programming languages within our main model. As

indicated in Table 16, the inclusion of Python and Oracle Java negatively affects unemployment

169
durations within our model. This result indicates that veterans who communicate skills related to

these languages may make themselves more employable, given the negative effects on

unemployment durations. Although SQL was the most prevalent within resumes, it was not

considered statistically significant. This indicates that there is potential the inclusion of this

programming language may not be as marketable than the former-mentioned languages. The

results of this analysis are summarized in Table 16.

Table 15. Data Science Programming Language Model Results

Psychological Human Capital (Psychometric Variables)

Analytic Thinking (-.108)*
~.0497
Authenticity (.672)
~.0080
Tone (-.055)***
~.0067
Psychological Human Capital (Communication Style)
Weak Disposition (-.014)
~.4992
Active Lexicon (-.551)***
~.1131
Career Identity (Data Science Competencies)
Ethics (-.536)*
~.2459
Machine Learning (-1.459)***
~.2898
Artificial Intelligence (-1.341)***
~.3116
Problem Solving (-.241)
~.2745
Statistics (3.634)***
~.3006
Control Variables
College Graduate (3.157)***
~.3537

170
Table 15. Data Science Programming Language Model Results Continued
Military Rank (-2.359)***
~.4802
Security Clearance (.746)***
~.2236
Years Active Duty (-3.343)***
~.4348
Previous Job (Managerial) (-1.565)***
~.2589
Previous Job (Data (.686)**
Science) ~.2471
Total Number of Civilian (.494)***
Jobs ~.0726

Programming Languages
Python (-1.556)***
~.2627

SQL (.565)
~.3109

Oracle Java (-1.317)***

~.2923
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001
Note: () covariate; ~standard error

171
 CHAPTER 7: DISCUSSION

This study makes three two contributions to literature. First, we contribute to the broad

texts surrounding data science, artificial intelligence, and human resource management by taking

a multidisciplinary approach to investigating a unique demographic of resumes. This research

leverages natural language processing techniques using Word2Vec to generate data science

competency dictionaries, which are then used to investigate veteran resumes. This approach

shines a light on potential capabilities that exist which could be used to better understand how

AI and machine learning techniques could be used within the human resource management

domain. To better understand how Gulf War II veterans communicate their data science skills,

we hypothesized that psychological human capital and career identity affect their employability

(Ngoma and Ntale, 2016; Fugate et al., 2004). By leveraging technologies like LIWC-22,

Word2Vec, and MAXQDA we successfully investigated potential variables which could affect a

veteran’s employability. This demonstrates the potential for future research to continue the

exploration of leveraging human resource data within AI and machine learning initiatives.

The second contribution of this research shines a light on factors which may affect a

veteran’s employability when seeking a job within the data science career field. Specifically, our

data and analysis suggest the positive tone of one’s resume decreases unemployment durations

throughout resumes. This implies that if a veteran uses optimistic diction or words that convey a

positive outlook on their experience, then the potential for lower unemployment durations exists.

Furthermore, if the veteran’s communication style is active, implying they take initiative, this

also has a positive influence on employability. Regarding the data science competencies, ethics,

artificial intelligence, and machine learning were also positive influences on a veteran’s

employability. AI and machine learning capabilities competencies have become a large

172
component of conducting analyses, especially with organizations that manage big data

initiatives. Communicating experience with ethical practices, such as conducting/building

transparent algorithms or bias mitigation techniques, could potentially enhance a veteran’s

employability. Moreover, communicating specific experience with machine learning and

artificial intelligence technologies could improve a veterans employability, as these skills are

seen as cutting-edge and competitive (Squicciarini and Nachtigall, 2021).

Future Work and Limitations

Through this analysis, we hope to inspire new literature to explore the various

components of human resource management and data science, especially by highlighting some

of the limitations which exist within this research. First, human resource professionals are

charged with managing a rich source of data relating to employees and their respective

organizations. At times, this data is qualitative in nature, which creates potential opportunities for

newer NLP technologies to step in to help analyze and manage them.

We leveraged Word2Vec as our primary NLP model to conduct similarity analysis;

however, there are more word embedding algorithms which exist that could provide insight to

contextual factors within resumes and job descriptions. This research was limited by only

leveraging Word2Vec. Future research initiatives could explore how other word embedding

algorithms like FastText (Bojanowski et al., 2017), GloVe (Pennington et al., 2014), ELMo

(Peters et al., 2018) and BERT (Delvin et al., 2018) analyze resumes and job descriptions.

Additionally, we acknowledge that this research is limited by the small sample size of our

resume pool. Understanding we leveraged a manual approach to cleaning and auditing these

resumes to ensure the demographic we desired was captured, there are sophisticated web

scraping capabilities which exist that may streamline the process and produce a larger sample-

173
size for analysis. Future research endeavors could leverage these newer technologies to acquire a

larger data set to analyze and interpret.

To expand upon this research initiative, other communication styles could be explored

when considering resumes. This research only considered Osgood et al.’s (1957) active and

weak lexicon, given the intent and nature of resumes. Other potential communication styles

could be explored such as Laswell’s Value dictionaries or Harvard’s IV-4 dictionaries (Lim,

2002).

Additionally, we consider future research initiatives could explore the temporal

differences which may exist between veterans seeking jobs within data science. This research

investigated any veteran that was considered a Gulf War II veteran. This could be broken down

into tertiary components. For example, exploring how well a veteran who was recently

discharged from the military communicates their skills versus veterans who have been out for

several years. This could potentially give insight to how well veterans are postured to write their

resumes upon leaving the military and provide an avenue for NLP capabilities to step in and help

investigate and analyze this phenomenon.

Lastly, future research initiatives could further investigate the various transition

assistance programs offered to military veterans. As discussed in Harrod et al.’s (2023) interview

research, veterans have claimed that the current curriculum has not prepared them well to

translate their skills or properly structure their resume. Through this research on veteran

employability, newer research directions could explore resumes of veterans who have recently

attended a transition assistance program and investigate the structure, content, and marketability

of their resumes. In doing so, this may provide deeper insight to potential curriculum

174
enhancements for the resume-building component of transition assistance programs offered to

military veterans entering the civilian labor force.

175
 CHAPTER 8: CONCLUSION

Within this paper, we investigated Gulf War II veteran resumes that communicated skills

within data science. We leveraged a multidisciplinary approach, allowing Fugate et al. (2004)

and Ngoma and Ntale’s (2016) theory on employability to guide our NLP investigations.

Through this analysis, we demonstrated factors which may contribute to positively affecting a

veteran’s employability when pursuing a career within the data science community. Through the

limitations addressed in the previous section, we clearly denote that there are rich avenues worth

exploring with regards to human resource management and the data science community. Outside

of the mentioned contributions, this research hopes to provide two additional perspectives. First,

we hope this research shines a light on how multidisciplinary approaches could benefit research

initiatives. Secondly, we hope this research inspires future scholars to continue the conversation

of where machine learning capabilities could exist within human resource management function.

We acknowledge there are considerable risks involving biases and transparency when

incorporating these newer technologies. With that, we hope this research invites more

conversation and dialogue on how to appropriately leverage and manage these newer

technologies to reduce tedious processes.

176
 APPENDICES

Appendix A: Spearman Correlation (Valecha et al., 2021)

1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 7
1 --
2 -0.106* --
-
0.165* .238*
3 ** ** --
-
.130*
4 .063 ** .028 --
5 .072 -.086 -.064 .116* --
.144* .214*
6 -0.97 ** ** -.045 -.050 --
.167** .202*
7 * -.50 -.121* .108* ** -.113* --
.288*
8 .083 -.121* .049 .022 .015 .015 ** --
.180* .209*
9 -.038 .035 .035 .045 .015 ** .071 ** --
1 .163* .170* .392* .204* .188*
0 .068 .076 .041 ** ** -.060 ** ** ** --
1 .163* .193*
1 .117* .009 .013 ** .066 -.037 .113* .117* -.027 ** --
1 .231** .185* .176* .136* .138*
2 * -.100* -.041 ** .074 -.047 ** ** .058 .098 ** --
-
1 .136* .138*
3 .50 ** -.008 ** .036 .040 .099* .122* .104* .044 .016 .057 --
-
1 .151* .188* .131* .172* .212*
4 .079 ** -.114* ** .099* -.041 .022 .044 .001 -.029 ** ** ** --
1 .282** .241* .148* .230* .217*
5 * .020 -.013 ** .090 -.054 ** .125* .101* .032 .115* ** .085 ** --
-
1 .219** .219* .159* .151* .291* .192* .216* .174* .259* .154*
6 * -.048 -.020 ** ** ** ** ** .104* ** ** ** .097 .014 ** --
1 .111** .214* .154* .146* .160* .05
7 .107* * .108* -.053 .047 * .004 ** ** .082 .123* .040 -.056 -.069 ** 6 --
Note: 1= Analytic; 2=Authentic; 3=Tone; 4=Acitve_Lexicon; 5=Weak_Lexicon; 6=Ethics; 7=Machine
Learning; 8=Artificial Intelligence; 9=Problem Solving; 10=Statistics; 11=College Graduate; 12=Rank;
13=Clearance; 14=Years Active Duty; 15=Managerial; 16=DataSciJob; 17=TotalNumJobs
Note: p-value <0.10; *p-value <0.05; **p-value <0.01; ***p-value <.001

177
Appendix B: Glassdoor Dictionary (Votto et al., 2023)

Ethics Machine Learning Problem Solving Artificial Intelligence (AI) Statistics

Dictionary Dictionary Dictionary Dictionary Dictionary
Exhibit Unsupervised Pragmatic Cognit Econometric
Uphold Reinforce Solver Augment Theor/etical/y
Style Artificial Break Ai Infer/ential
Urgency Multimodal Diagnose Fusion Causal
Credible Causal Difficult Chip Exploratory
Ethic Convolution Conceptual Multimodal Actuary
Meticulous Neural Frame Exploit Mathematic/al
Demeanor Infer/entail/ce Thinker Signal Algebra
Motiv/e/ation Theor/etical/y Dig Artificial Intelligence Statistic
Keen Vector Problem Audio Correlate
solv/e/ing

178
Appendix C: LinkedIn Dictionary

Ethics Machine Learning Problem Solving Artificial Intelligence (AI) Statistics

Dictionary Dictionary Dictionary Dictionary Dictionary
AI Audio Analyst Acumen Assumption
Blend Classic Bridge Elegant Concentration
Driverless Convolution Cognit Intuition Correlate
Especial Fascinate Drill Pain Diffusion
Intersect Fusion Fascinate Pragmatic Ensemble
Proximity Metrologic Fusion Solver Quantitative
Supercharge Outlier Increasingly Tenacious Static
Transparent Reinforcement Quantum Thinker Survivor
Trustworthy Sophisticated Signal Uncover Theory
Unbound Unsupervised Synthetic Unrelated Underly

179
 REFERENCES

Adamczyk, R. (2014). Selected Aspects of Emotionality in Medical English Lexis–Application of the

Osgood Semantic Differential. Ostrava Journal of English Philology, 6(2), 95–110.

Ainspan, N. D., & Saboe, K. N. (2021). Military Veteran Employment: A Guide for the Data-Driven
Leader. Oxford University Press.

Ammanath, B. (2022). Trustworthy AI: A Business Guide for Navigating Trust and Ethics in AI. John
Wiley & Sons.

Bakhshi, H., Downing, J. M., Osborne, M. A., & Schneider, P. (2017). THE FUTURE OF SKILLS.
Pearson.

Ballance, R. H., Forstner, H., & Murray, T. (1987). Consistency Tests of Alternative Measures of
Comparative Advantage. The Review of Economics and Statistics, 69(1), 157–161. https://doi.or
g/10.2307/1937915

Bartlett, J. E., Kotrlik, J. W., & Higgins, C. C. (n.d.). Organizational Research: Determining
Appropriate Sample Size in Survey Research.

Bergman, M., & Herd, A. (2017). Proven Leadership = College Credit: Enhancing Employability of
Transitioning Military Members Through Prior Learning Assessment. Advances in Developing
Human Resources, 19(1), 78–87. https://doi.org/10.1177/1523422316682949

Beristain, S. I., Barbosa, R. R. L., & Barriocanal, E. G. (2022). Improving Jobs-Resumes

Classification: A Labor Market Intelligence Approach. International Journal of Information
Technology & Decision Making, 1–17. https://doi.org/10.1142/S0219622023500013

Berntson, E., Näswall, K., & Sverke, M. (2008). Investigating the relationship between employability
and self-efficacy: A cross-lagged analysis. European Journal of Work and Organizational
Psychology, 17(4), 413–425. https://doi.org/10.1080/13594320801969699

Bhambri, S. (n.d.). Employability Skills: Ticket to Employment. 11(4).

Bojanowski, P., Grave, E., Joulin, A., & Mikolov, T. (2017). Enriching Word Vectors with Subword
Information. Transactions of the Association for Computational Linguistics, 5, 135–146. https://doi
.org/10.1162/tacl_a_00051

Boser, S. (2007). Power, Ethics, and the IRB: Dissonance Over Human Participant Review of
Participatory Research. Qualitative Inquiry, 13(8), 1060–1074. https://doi.org/10.1177/107780
0407308220

Boyd, R. L. (2022). The Development and Psychometric Properties of LIWC-22. Austin, TX:
University of Texas at Austin.

180
Boyd, R. L., & Pennebaker, J. W. (2015). Did Shakespeare Write Double Falsehood ? Identifying
Individuals by Creating Psychological Signatures With Text Analysis. Psychological Science, 26(5),
570–582. https://doi.org/10.1177/0956797614566658

Brumer, Y., Shapira, B., Rokach, L., & Barkan, O. (2017). Predicting Relevance Scores for Triples
from Type-Like Relations using Neural Embedding.

Bureau of Labor Statstics. (2023). Data Scientists: Occupational Outlook Handbook: : U.S. Bureau of
Labor Statistics. https://www.bls.gov/ooh/math/data-scientists.htm

Cohn, M. A., Mehl, M. R., & Pennebaker, J. W. (2004). Linguistic Markers of Psychological Change
Surrounding September 11, 2001. Psychological Science, 15(10), 687–693. https://doi.org/10.11
11/j.0956-7976.2004.00741.x

Cole, M. S., Feild, H. S., Giles, W. F., & Harris, S. G. (2004). Job Type and Recruiters’ Inferences of
Applicant Personality Drawn from Resume Biodata: Their Relationships with Hiring
Recommendations. International Journal of Selection and Assessment, 12(4), 363–367. https://doi.
org/10.1111/j.0965-075X.2004.00291.x

Cole, M. S., Rubin, R. S., Feild, H. S., & Giles, W. F. (2007). Recruiters’ Perceptions and Use of
Applicant Résumé Information: Screening the Recent Graduate. Applied Psychology, 56(2), 319–343.
https://doi.org/10.1111/j.1464-0597.2007.00288.x

Curry Hall, K., Harrell, M., Bicksler, B., Stewart, R., & Fisher, M. (2014). Veteran Employment:
Lessons from the 100,000 Jobs Mission. RAND Corporation. https://doi.org/10.7249/RR836

Deguchi, T., Seo, S., & Ishii, N. (2022). Meaning of the Clusters on Dimensionality Reduction by
Word Clustering. 2022 12th International Congress on Advanced Applied Informatics (IIAI-AAI),
325–330. https://doi.org/10.1109/IIAIAAI55812.2022.00072

Department of Labor. (2023). Latest Employment Numbers. DOL. http://www.dol.gov/agencies/

vets/latest-numbers

Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2018). BERT: Pre-training of Deep
Bidirectional Transformers for Language Understanding. https://doi.org/10.48550/arXiv.1810.04805

Drouin, M., Boyd, R. L., Hancock, J. T., & James, A. (2017). Linguistic analysis of chat transcripts
from child predator undercover sex stings. The Journal of Forensic Psychiatry & Psychology, 28(4),
437–457. https://doi.org/10.1080/14789949.2017.1291707

Enholm, I. M., Papagiannidis, E., Mikalef, P., & Krogstie, J. (2022). Artificial Intelligence and
Business Value: A Literature Review. Information Systems Frontiers, 24(5), 1709–1734. https://
doi.org/10.1007/s10796-021-10186-w

Fossey, M., Lazier, R., Neil Lewis, M., Williamson, N., & Caddick, N. (2019). Chapter 4—Military-
to-civilian transition policies, processes, and program efforts. In C. A. Castro & S. Dursun (Eds.),

181
Military Veteran Reintegration (pp. 51–74). Academic Press. https://doi.org/10.1016/B978-0-12-
815312-3.00004-8

Fox, A. K., & Royne Stafford, M. B. (2021). Olympians on Twitter: A Linguistic Perspective of the
Role of Authenticity, Clout, and Expertise in Social Media Advertising. Journal of Current Issues &
Research in Advertising, 42(3), 294–309. https://doi.org/10.1080/10641734.2020.1763521

Fugate, M., & Kinicki, A. J. (2008). A dispositional approach to employability: Development of a

measure and test of implications for employee reactions to organizational change. Journal of
Occupational and Organizational Psychology, 81(3), 503–527. https://doi.org/10.1348/09631
7907X241579

Fugate, M., Kinicki, A. J., & Ashforth, B. E. (2004). Employability: A psycho-social construct, its
dimensions, and applications. Journal of Vocational Behavior, 65(1), 14–38. https://doi.org/10.
1016/j.jvb.2003.10.005

Garamone, J. (2021). U.S. Completes Troop-Level Drawdown in Afghanistan, Iraq [Government].

U.S. Department of Defense. https://www.defense.gov/News/News-
Stories/Article/article/2473884/us-completes-troop-level-drawdown-in-afghanistan-
iraq/https%3A%2F%2Fwww.defense.gov%2FNews%2FNews-
Stories%2FArticle%2FArticle%2F2473884%2Fus-completes-troop-level-drawdown-in-afghanistan-
iraq%2F

Goss, K. K. (n.d.). How Post-9/11 Army Officers Describe Their Successful Transition from the
Military to Civilian Employment in the United States [Ed.D., Grand Canyon University]. Retrieved
March 23, 2023, from https://www.proquest.com/docview/2465715130/abstract/A77EF185065
5400EPQ/1.

Graham, B., & Paul, C. (2010). Does higher education really lead to higher employability and wages
in the RMI? US Census Bureau Report.

Habbat, N., Anoun, H., & Hassouni, L. (2022). Exploration, Sentiment Analysis, Topic Modeling,
and Visualization of Moroccan Twitter Data. In J. Kacprzyk, V. E. Balas, & M. Ezziyyani (Eds.),
Advanced Intelligent Systems for Sustainable Development (AI2SD’2020) (pp. 1067–1083). Springer
International Publishing. https://doi.org/10.1007/978-3-030-90639-9_87

IBM. (2020). The Data Science Skills Competency Model. IBM Corporation, 12.

Harrod, M., Miller, E. M., Henry, J., & Zivin, K. (2017). “I’ve never been able to stay in a job”: A
qualitative study of Veterans’ experiences of maintaining employment. Work, 57(2), 259–268.
https://doi.org/10.3233/WOR-172551

James, A. (n.d.). A Phenomenological Study: Female Veteran Commissioned Officers’ Successful

Transition to the Civilian Workforce [Ed.D., Grand Canyon University]. Retrieved March 23, 2023,
from https://www.proquest.com/docview/1889186685/abstract/47284E922D5442CFPQ/1

182
Jarrahi, M. H. (2018). Artificial intelligence and the future of work: Human-AI symbiosis in
organizational decision making. Business Horizons, 61(4), 577–586. https://doi.org/10.101
6/j.bushor.2018.03.007

Jennings, R. E., Lanaj, K., Koopman, J., & McNamara, G. (2022). Reflecting on one’s best possible
self as a leader: Implications for professional employees at work. Personnel Psychology, 75(1), 69–
90. https://doi.org/10.1111/peps.12447

Jordan, K. N., Sterling, J., Pennebaker, J. W., & Boyd, R. L. (2019). Examining long-term trends in
politics and culture through language of political leaders and cultural institutions. Proceedings of the
National Academy of Sciences, 116(9), 3476–3481. https://doi.org/10.1073/pnas.1811987116

Kacewicz, E., Pennebaker, J. W., Davis, M., Jeon, M., & Graesser, A. C. (2014). Pronoun Use
Reflects Standings in Social Hierarchies. Journal of Language and Social Psychology, 33(2), 125–
143. https://doi.org/10.1177/0261927X13502654

Kalichman, S. C., Link to external site, this link will open in a new window, & Smyth, J. M. (2021).
“And you don’t like, don’t like the way I talk”: Authenticity in the language of bruce springsteen.
Psychology of Aesthetics, Creativity, and the Arts. https://doi.org/10.1037/aca0000402

Kaur, R., & Singh, H. (2018). A Literature Review on the Employability and the Effects of Ex-
Military Personnel in Corporate Boardrooms. International Affairs and Global Strategy, 61, 50–58.

Keeling, M. E., Ozuna, S. M., Kintzle, S., & Castro, C. A. (2019). Veterans’ Civilian Employment
Experiences: Lessons Learnt From Focus Groups. Journal of Career Development, 46(6), 692–705.
https://doi.org/10.1177/0894845318776785

Kim, D. J., Song, Y. I., Braynov, S. B., & Rao, H. R. (2005). A multidimensional trust formation
model in B-to-C e-commerce: A conceptual framework and content analyses of academia/practitioner
perspectives. Decision Support Systems, 40(2), 143–165. https://doi.org/10.1016/j.dss.2004.01.006

Kulkarni, V., Bewoor, A. K., Malathi, P., & Balapgol, B. S. (2017). Employability Skill Matrix for
Engineering Graduates of Tier-II Institutes. Journal of Engineering Education Transformations,
30(3).

Kumar, P., Kumar, P., Zaidi, N., & Rathore, V. S. (2018). Analysis and Comparative Exploration of
Elastic Search, MongoDB and Hadoop Big Data Processing. In M. Pant, K. Ray, T. K. Sharma, S.
Rawat, & A. Bandyopadhyay (Eds.), Soft Computing: Theories and Applications (pp. 605–615).
Springer. https://doi.org/10.1007/978-981-10-5699-4_57

Lampka, E., & Kowalewski, S. J. (2017). Veterans in the Workplace: An Analysis of Military Veteran
Underutilization in the Civilian Workforce. 2(1).

Lehner, O. M., Ittonen, K., Silvola, H., Ström, E., & Wührleitner, A. (2022). Artificial intelligence
based decision-making in accounting and auditing: Ethical challenges and normative thinking.

183
Accounting, Auditing & Accountability Journal, 35(9), 109–135. https://doi.org/10.1108/AAAJ-09-
2020-4934

Lennon, C., Zilian, L. S., & Zilian, S. S. (2023). Digitalisation of occupations—Developing an

indicator based on digital skill requirements. PLOS ONE, 18(1), e0278281.
https://doi.org/10.1371/journal.pone.0278281

Lewis, P. V. (1985). Defining ‘business ethics’: Like nailing jello to a wall. Journal of Business
Ethics, 4(5), 377–383. https://doi.org/10.1007/BF02388590

Li, G., Yuan, C., Kamarthi, S., Moghaddam, M., & Jin, X. (2021). Data science skills and domain
knowledge requirements in the manufacturing industry: A gap analysis. Journal of Manufacturing
Systems, 60, 692–706. https://doi.org/10.1016/j.jmsy.2021.07.007

Lim, E. T. (2002). Five Trends in Presidential Rhetoric: An Analysis of Rhetoric from George
Washington to Bill Clinton. Presidential Studies Quarterly, 32(2), 328–366.

Liu, W. (2018). High-involvement human resource practices, employee learning and employability.
Career Development International, 23(3), 312–326. https://doi.org/10.1108/CDI-10-2017-0177

Loughran, D. S. (2014). Why Is Veteran Unemployment So High? Rand Corporation.

Lu, G., Ding, X. (David), Peng, D. X., & Hao-Chun Chuang, H. (2018). Addressing endogeneity in
operations management research: Recent developments, common problems, and directions for future
research. Journal of Operations Management, 64, 53–64. https://doi.org/10.1016/j.jom.2018.10.001

Margaret Little, B. (2011). Employability for the workers – what does this mean? Education +
Training, 53(1), 57–66. https://doi.org/10.1108/00400911111102360

Markowitz, D., Kouchaki, M., Gino, F., Hancock, J., & Boyd, R. (2022). Authentic First Impressions
Relate to Interpersonal, Social, and Entrepreneurial Success. Social Psychological & Personality
Science. https://journals-sagepub-com.libweb.lib.utsa.edu/doi/full/10.1177/19485506221086138

McAllister, C. P., Mackey, J. D., Hackney, K. J., & Perrewé, P. L. (2015). From Combat to Khakis:
An Exploratory Examination of Job Stress With Veterans. Military Psychology, 27(2), 93–107.
https://doi.org/10.1037/mil0000068

McEvoy, G. M., Hayton, J. C., Warnick, A. P., Mumford, T. V., Hanks, S. H., & Blahna, M. J.
(2005). A Competency-Based Model for Developing Human Resource Professionals. Journal of
Management Education, 29(3), 383–402. https://doi.org/10.1177/1052562904267538

McKenny, A. F., Short, J. C., & Payne, G. T. (2013). Using Computer-Aided Text Analysis to Elevate
Constructs: An Illustration Using Psychological Capital. Organizational Research Methods, 16(1),
152–184. https://doi.org/10.1177/1094428112459910

184
McLennan, W. (2021). Understanding Veteran’s Resumes and Conducting Veteran Interviews. In N.
D. Ainspan & K. N. Saboe, Military Veteran Employment: A Guide for the Data-Driven Leader.
Oxford University Press.

Memarovic, N., Langheinrich, M., Alt, F., Elhart, I., Hosio, S., & Rubegni, E. (2012). Using public
displays to stimulate passive engagement, active engagement, and discovery in public spaces.
Proceedings of the 4th Media Architecture Biennale Conference: Participation, 55–64.
https://doi.org/10.1145/2421076.2421086

Minerley, S. (2018). Mind The Gap: Unemployment Discrimination. Government Law Review, 1(11).
Minnis, S. E. (n.d.). A phenomenological exploration of combat veterans’ experiences as they
transition to civilian employment using higher education as career development [Ph.D., Texas A&M
University]. Retrieved March 6, 2023, from https://www.proquest.com/docview/1650237914/
abstract/7E5B68B3D714235PQ/1

Mökander, J., & Floridi, L. (2021). Ethics-Based Auditing to Develop Trustworthy AI. Minds and
Machines, 31(2), 323–327. https://doi.org/10.1007/s11023-021-09557-8

Monzani, D., Vergani, L., Pizzoli, S. F. M., Marton, G., & Pravettoni, G. (2021). Emotional Tone,
Analytical Thinking, and Somatosensory Processes of a Sample of Italian Tweets During the First
Phases of the COVID-19 Pandemic: Observational Study. Journal of Medical Internet Research,
23(10), e29820. https://doi.org/10.2196/29820

Muthyala, R., Wood, S., Jin, Y., Qin, Y., Gao, H., & Rai, A. (2017). Data-Driven Job Search Engine
Using Skills and Company Attribute Filters. 2017 IEEE International Conference on Data Mining
Workshops (ICDMW), 199–206. https://doi.org/10.1109/ICDMW.2017.33

Newman, M., Pennebaker, J. W., Berry, D. S., & Richards, J. M. (2003). Lying Words: Predicting
Deception from Linguistic Styles. Personality and Social Psychology Bulletin, 25(5), 665–675.

Ngoma, M., & Ntale, P. (2016). Psychological capital, career identity and graduate employability in
Uganda: The mediating role of social capital. International Journal of Training and Development,
20(2), 124–139. https://doi.org/10.1111/ijtd.12073

Nisha, S. M., & Rajasekaran, V. (2018). Employability Skills: A Review. IUP Journal of Soft Skills,
12(1), 29–37.

Niu, Y., Zhu, Y., Xu, X., & Hunter-Johnson, Y. (2022). Exploring Self-Perceived Employability and
Ambition of Student Veterans in a Higher Education Institution in the United States. Journal of
Interdisciplinary Studies in Education, 11(2), Article 2.

Norlander, P., Ho, G. C., Shih, M., Walters, D. J., & Pittinsky, T. L. (2020). The Role of
Psychological Stigmatization in Unemployment Discrimination. Basic and Applied Social
Psychology, 42(1), 29–49. https://doi.org/10.1080/01973533.2019.1689363

185
Oliveira, M., Bitencourt, C. C., Santos, A. C. M. Z. dos, & Teixeira, E. K. (2015). Thematic Content
Analysis: Is There a Difference Between the Support Provided by the MAXQDA® and NVivo®
Software Packages? Revista de Administração Da UFSM, 9(1), 72–82. https://doi.org/10.5902
/1983465911213

O*Net. (n.d.). National Center for O*Net Development: Data Scientists 15-2051.000. O*Net Online.
Retrieved June 13, 2023, from https://www.onetonline.org/link/summary/15-2051.00

Oosthuizen, R. M., Coetzee, M., & Mntonintshi, F. (2014). Investigating the relationship between
employees’ career anchors and their psychosocial employability attributes in a financial company. SA
Journal of Human Resource Management, 12(1), Article 1. https://doi.org/10.4102/sajhrm.v12i1.650
Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The Measurement of Meaning. University of
Illinois Press.

Osgood, D. W. (2000). Poisson-Based Regression Analysis of Aggregate Crime Rates. Journal of

Quantitative Criminology, 16(1), 21–43. https://doi.org/10.1023/A:1007521427059

Parameswaran, S., Mukherjee, P., & Valecha, R. (2022). I Like My Anonymity: An Empirical
Investigation of the Effect of Multidimensional Review Text and Role Anonymity on Helpfulness of
Employer Reviews. Information Systems Frontiers. https://doi.org/10.1007/s10796-022-10268-3

Pennebaker, J. W., Chung, C. K., Frazee, J., Lavergne, G. M., & Beaver, D. I. (2014). When Small
Words Foretell Academic Success: The Case of College Admissions Essays. PLOS ONE, 9(12),
e115844. https://doi.org/10.1371/journal.pone.0115844

Pennington, J., Socher, R., & Manning, C. (2014). Glove: Global Vectors for Word Representation.
Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing
(EMNLP), 1532–1543. https://doi.org/10.3115/v1/D14-1162

Perkins, D. F., Davenport, K. E., Morgan, N. R., Aronson, K. R., Bleser, J. A., McCarthy, K. J., Vogt,
D., Finley, E. P., Copeland, L. A., & Gilman, C. L. (2022). The influence of employment program
components upon job attainment during a time of identity and career transition. International Journal
for Educational and Vocational Guidance. https://doi.org/10.1007/s10775-022-09527-1

Peters, M. E., Neumann, M., Iyyer, M., Gardner, M., Clark, C., Lee, K., & Zettlemoyer, L. (2018).
Deep contextualized word representations (arXiv:1802.05365). arXiv. http://arxiv.org
/abs/1802.05365

Peterson, N. G., Mumford, M. D., Borman, W. C., Jeanneret, P. R., Fleishman, E. A., Levin, K. Y.,
Campion, M. A., Mayfield, M. S., Morgeson, F. P., Pearlman, K., Gowing, M. K., Lancaster, A. R.,
Silver, M. B., & Dye, D. M. (2001). Understanding Work Using the Occupational Information
Network (o*net): Implications for Practice and Research. Personnel Psychology, 54(2), 451–492.
https://doi.org/10.1111/j.1744-6570.2001.tb00100.x

186
Puad, M. H. M., & Desa, H. M. (2020). Dissecting Perceptions of New Graduates on Work
Orientation and Self-Confidence in Employability Skills Training Program. Universal Journal of
Educational Research, 8(1A), 70–75. https://doi.org/10.13189/ujer.2020.081310

Robertson, H. C., & Brott, P. E. (2013). Male Veterans’ Perceptions of Midlife Career Transition and
Life Satisfaction: A Study of Military Men Transitioning to the Teaching Profession. Adultspan
Journal, 12(2), 66–79. https://doi.org/10.1002/j.2161-0029.2013.00016.x

Rosenberg, S., Heimler, R., & Morote, E. (2012). Basic employability skills: A triangular design
approach. Education + Training, 54(1), 7–20. https://doi.org/10.1108/00400911211198869

Ross, C. M., & Young, S. J. (2005). Resume Preferences: Is It Really “Business as Usual”? Journal of
Career Development, 32(2), 153–164. https://doi.org/10.1177/0894845305279162

Routon, P. W. (2014). The Effect of 21st Century Military Service on Civilian Labor and Educational
Outcomes. Journal of Labor Research, 35(1), 15–38. https://doi.org/10.1007/s12122-013-9170-4

Royyan, A. R., & Setiawan, E. B. (2022). Feature Expansion Word2Vec for Sentiment Analysis of
Public Policy in Twitter | Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi).
http://www.jurnal.iaii.or.id/index.php/RESTI/article/view/3525

Rudin, C. (2019). Stop explaining black box machine learning models for high stakes decisions and
use interpretable models instead. Nature Machine Intelligence, 1(5), 206–215. https://doi.org/10.1038
/s42256-019-0048-x

Rutter, C. M. (n.d.). Military Veteran Transition to the Civilian Workforce: A Descriptive Study
[Ph.D., Grand Canyon University]. Retrieved March 23, 2023, from https://www.proquest.com
/docview/2465726378/abstract/C4BE6605BC164710PQ/1

Shah, N., Willick, D., & Mago, V. (2022). A framework for social media data analytics using
Elasticsearch and Kibana. Wireless Networks, 28(3), 1179–1187. https://doi.org/10.1007/s11276-
018-01896-2

Smith, P. A. C., & Sharma, M. (2002). Developing personal responsibility and leadership traits in all
your employees: Part 1 – shaping and harmonizing the high‐performance drivers. Management
Decision, 40(8), 764–774. https://doi.org/10.1108/00251740210441018

Squicciarini, M., & Nachtigall, H. (2021). Demand for AI skills in jobs: Evidence from online job
postings. https://doi.org/10.1787/3ed32d94-en

Stoklasa, J., Talášek, T., & Stoklasová, J. (2019). Semantic differential for the twenty-first century:
Scale relevance and uncertainty entering the semantic space. Quality & Quantity, 53(1), 435–448.
https://doi.org/10.1007/s11135-018-0762-1

Stone, C. B. (2016). The veteran myth: An experimental investigation of human resource managers’
perceptions of United States military veterans [Ph.D., The University of Texas at San Antonio]. In

187
ProQuest Dissertations and Theses.
https://www.proquest.com/docview/1793670665/abstract/78224DDD49F844C2PQ/1

Stone, P. J., & Hunt, E. B. (1963). A computer approach to content analysis: Studies using the
General Inquirer system. Proceedings of the May 21-23, 1963, Spring Joint Computer Conference,
241–256. https://doi.org/10.1145/1461551.1461583

Taherdoost, H. (2018). How to Design and Create an Effective Survey/Questionnaire; A Step by Step
Guide (SSRN Scholarly Paper No. 3224226). https://papers.ssrn.com/abstract=3224226

Valecha, R., Srinivasan, S. K., Volety, T., Kwon, K. H., Agrawal, M., & Rao, H. R. (2021). Fake
News Sharing: An Investigation of Threat and Coping Cues in the Context of the Zika Virus. Digital
Threats: Research and Practice, 2(2), 1–16. https://doi.org/10.1145/3410025

Valli, R. (2016). Creating a questionnaire for a scientific study. International Journal of Research
Studies in Education, 6(4). https://doi.org/10.5861/ijrse.2016.1584

Venkatesan, S., Valecha, R., Yaraghi, N., Oh, O., & Rao, H. R. (2021). Influence in Social Media: An
Investigation of Tweets Spanning the 2011 Egyptian Revolution. MIS Quarterly, 45(4), 1679–1714.
https://doi.org/10.25300/MISQ/2021/15297

Votto, A. M., Manuel, D., Valecha, R., Keeton, K., Rao, H. R. (2023). JC-Compass: A Framework
for Conducting Competency-Based Job Posting Research and Analysis. Manuscript in preparation.

Votto, A. M., Valecha, R., Najafirad, P., & Rao, H. R. (2021). Artificial Intelligence in Tactical
Human Resource Management: A Systematic Literature Review. International Journal of
Information Management Data Insights, 1(2), 100047. https://doi.org/10.1016/j.jjimei.2021.100047

188
 VITA

Captain Alexis “Ali” Votto joined the UTSA Ph.D. program in Information Technology

as a student in August of 2020. She is one of the few Active-Duty members serving in the Air

Force to be selected for the highly competitive Secretary of the Air Force STEM Ph.D. program.

This program allows Active-Duty officers to attend a civilian institution in pursuit of a STEM-

related Ph.D. as their primary job for three years. She is a Force Support Officer (FSO) by trade,

which translates to Human Resource Management, Mortuary Affairs, and Manpower analytics as

her profession.

Having earned her Bachelor of Science in Mathematics and Master of Science in Human

Resource Management, she sought to marry these two academic backgrounds within her Ph.D. to

see how she can enhance the HRM profession and potential processes within the USAF. To her

knowledge, she is one of the very few FSOs in her career field to achieve a Ph.D. within a STEM

field. She hopes this achievement inspires others to compete for programs to further explore how

STEM exists in business management processes. Prior to joining the Ph.D. program, she served

in various leadership positions ranging from Section Command to Deputy Flight Command. She

has had administrative oversight for over 1,600 personnel as an executive officer and was

privileged with G-Series orders to issue administrative actions on behalf of the commander as a

section commander.

On a personal note, she calls San Antonio her home. This is the city where her career

started and where she hopes to return to once it’s time to retire from the USAF. Future endeavors

include serving at the Air Force Personnel Center at Randolph Air Force Base as a personnel

analyst. She is happily married to Joseph Votto and they have three fur-children, Koda, Hank,

and Butters.
 ProQuest Number: 30575705

INFORMATION TO ALL USERS

The quality and completeness of this reproduction is dependent on the quality
and completeness of the copy made available to ProQuest.

Distributed by ProQuest LLC ( 2023 ).

Copyright of the Dissertation is held by the Author unless otherwise noted.

This work may be used in accordance with the terms of the Creative Commons license
or other rights statement, as indicated in the copyright statement or in the metadata
associated with this work. Unless otherwise specified in the copyright statement
or the metadata, all rights are reserved by the copyright holder.

This work is protected against unauthorized copying under Title 17,

United States Code and other applicable copyright laws.

Microform Edition where available © ProQuest LLC. No reproduction or digitization

of the Microform Edition is authorized without permission of ProQuest LLC.

ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, MI 48106 - 1346 USA