Separating The Effects of Knowledge Construction and Group Collaboration in Learning Outcomes of Web-Based Courses

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Information & Management 43 (2006) 778–793

www.elsevier.com/locate/im

Separating the effects of knowledge construction and group


collaboration in learning outcomes of web-based courses
Raquel Benbunan-Fich a,*, J.B. Arbaugh b,1
a
Zicklin School of Business, Box B11-220, Baruch College, CUNY, New York, NY 10010, United States
b
College of Business Administration, University of Wisconsin Oshkosh, Oshkosh, WI 54901, United States
Received 17 May 2005; accepted 8 September 2005

Abstract
This research investigates the effects of knowledge construction and group collaboration in outcomes of MBA courses entirely
delivered on the web. The data comes from a sample of courses conducted from summer 2000 through summer 2002. The results
indicate that students achieve higher perceptions of learning in courses where knowledge is transmitted through the system, as
opposed to constructed by the students, and when students are engaged in collaborative assignments. In terms of final grades, our
findings show that the absence of both factors (knowledge construction and group collaboration) has a detrimental effect on student
performance. Students earn on average better grades in courses where either one of these factors is present. Contrary to our
expectations, the presence of both factors does not have positive synergistic effects on student performance. These findings suggest
that the success of web-based courses entirely delivered online depends upon providing collaborative learning activities, and/or
challenging environments where the students are responsible for the creation of their own knowledge.
# 2006 Elsevier B.V. All rights reserved.

Keywords: Knowledge acquisition; Knowledge construction; Group collaboration; Learning outcomes; Technology-mediated learning; Web-
based courses

1. Introduction [19,23]. However, in order to fully understand the


drivers of better outcomes, we need to make a
The growing popularity of web-based courses at all distinction between how the knowledge is delivered
levels has prompted a multitude of research efforts to in a web-based environment and whether the learning is
test which online course design and delivery strategies expected to occur in terms of individual activities or as a
are most successful. Based on evidence provided by result of group-based processes.
multi-course/multi-section studies and single course Research in technology-mediated learning environ-
studies comparing online and traditional classrooms, ments has received increasing attention in the
the consensus seems to be that collaborative approaches Information Systems literature in the past decade [2].
are better in online courses than individualistic methods IS scholars are uniquely qualified to study how IT
influences learning, not only due to their understanding
of the technology but also because of the long tradition
in the field to study how the technology affects
* Corresponding author. Tel.: +1 646 312 3375.
E-mail addresses: Raquel_BenbunanFich@baruch.cuny.edu
individual cognitive processes and group collabora-
(R. Benbunan-Fich), arbaugh@uwosh.edu (J.B. Arbaugh). tion. In particular, several empirical studies have
1
Tel.: +1 920 424 7189. analyzed the learning benefits of using Group Support

0378-7206/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.im.2005.09.001
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 779

Systems to enhance or to replace traditional classrooms large number of online versus traditional classroom
[1,5,28], and how Group Support Systems enable studies, research based on larger samples of courses
knowledge acquisition [26,27]. However, it is not entirely delivered online is limited. This prior multi-
known whether the superior outcomes achieved by course research has found strong evidence for the
members of the GSS teams are due to the capabilities of effectiveness of collaborative learning in asynchronous
the technology-mediated environment to support learning environments [11]. However, two issues often
knowledge dissemination and sharing, and/or due to overlooked in this area are the role of the learning model
the effects of working in groups. [2,29,34], and the interplay between knowledge delivery
Our main research question is whether the knowl- and group collaboration in online courses.
edge delivery approach or the use of collaborative
learning techniques, or the combination of both, 2.1. Knowledge delivery: transfer versus
improves the outcomes of web-based courses. Empiri- construction
cal evidence to date is somewhat mixed, in part because
the knowledge construction and group collaboration In order for learning to occur, the student must
constructs have not been conceptualized and operatio- acquire knowledge. Grover and Davenport [20] define
nalized as orthogonal constructs. In addition, most of knowledge as the most valuable form of content and
the prior empirical research has been focused on describe a continuum that starts at data, encompasses
comparing online versus traditional courses, rather than information and ends in knowledge. The main distinc-
testing the effectiveness of different models in web- tion between knowledge and information is that
based courses. Therefore, we seek to analyze the ‘‘knowledge is information possessed in the minds of
separate and joint effects of knowledge construction and individuals’’ [3] p. 109. Therefore, from a cognitive
group collaboration in a sample of pure online courses. perspective, knowledge is the result of mental proces-
We begin by reviewing the relevant literature in sing generated by new stimuli.2 The central concern of
knowledge construction and group collaboration and technology-mediated learning environments is how to
developing the hypotheses. Then, we describe the leverage the technology and provide the most effective
research methodology, including the experimental stimuli to improve knowledge acquisition.
design, the sample and the data collection methods. One of the key issues in the design and delivery of
We proceed to present the results of our statistical web-based courses is how knowledge is disseminated
analyses, followed by the discussion, the limitations and through the medium. Although the concern is not
the implications. Finally, we present our conclusions exclusive of technology-mediated learning environ-
and the contributions, and outline possible areas for ments, it is perhaps critical when instructors and
future research endeavors. students are separated from each other in terms of space
and time. A parallel concern in education [25], echoed
2. Literature review and hypotheses in some of the Information Systems research [29], is the
development debate around the appropriateness of objectivist and
constructivist approaches. These two paradigms repre-
Research in the Information Systems literature in the sent fundamentally different conceptions of knowledge.
area of technology-mediated learning includes theore- Objectivism is based on the dissemination of a stable
tical and empirical studies. Theoretical studies are those body of knowledge while constructivism is based on
that present conceptual models fitting technology knowledge construction. A useful analogy to under-
applications to various learning theories [10,29]. stand the difference between these two approaches
Empirical studies are those that compare online learning consists of viewing objectivism as the delivery of a
environments to traditional classrooms [34], or inves- complete set of concepts that needs to be learned (for
tigate the use of Group Support Systems to enable example, a page of written text), and constructivism as
collaborative learning activities in a particular course
[1,28], or test the effectiveness of particular synchro-
2
nous and asynchronous online learning environments Alavi and Leidner [3] argue that when knowledge is articulated
[4,11,32,42]. becomes information and when information is cognitively processed
Although there are some inconsistent findings, becomes knowledge. However, as we will discuss below, in a tech-
nology-mediated learning environment, information for instructional
empirical research comparing online and traditional purposes is highly dependent on the assumptions about the nature of
courses has provided evidence of no significant knowledge. Therefore, in this paper, we will refer to instructional
differences in learning outcomes [2]. Compared to the content as knowledge.
780 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

the conveyance of these concepts in pieces from Collaborative activities allow learners greater opportu-
multiple sources that each learner puts together or nities for increased social presence and a greater sense of
constructs by him/herself. online community, both of which have been associated
The objectivist model consists of transferring with positive online course outcomes [36]. When
knowledge from the professor to the students and working with peers instead of alone (or just with the
allowing each student to learn independently. The main instructor), anxiety and uncertainty are reduced as
assumption in this approach is that there is a unique and learners communicate with their teammates and find
objective body of knowledge representing the world that their way together through complex or new tasks [21].
can be articulated and communicated to the students. As In addition to improving the socio-emotional
a result, the instructor controls material and pace of climate, collaborative activities provide opportunities
learning, and the lecture is the typical mode of for enhanced learning. Students learn more when they
instruction [25]. Direct knowledge transfer may be participate in group endeavors because of specific
the most appropriate model for subject matters based on mechanisms that affect cognitive processes. These
factual, technical or procedural knowledge [29]. mechanisms include resolution of conflicts or disagree-
However, some critics of this approach argue that the ments in group discussions, internalization of explana-
quality of learning is impoverished when students are tions provided by more knowledgeable peers and self-
passive recipients of content that is delivered to them. explanation effects because explaining to others
In contrast, the constructivist model assumes that clarifies one’s own understanding [4,11,31,34,41]. At
knowledge is created or constructed by every learner. the same time, working in virtual teams allows students
Instead of an external objective reality, the mind to reflect on the written contributions of others and
produces its own conception [25]. Within constructi- encourages them to elaborate their thoughts before
vism, there are two different schools of thought: expressing them in writing [21]. The success of group
individual, which assumes that knowledge construction collaboration in technology-mediated environments
occurs in the student’s head and social, based on the depends to a great extent on the effectiveness of the
premise that knowledge is socially constructed in the group of learners. Problems such as lack of commitment
world through interactions with others [24]. The by individuals, insufficient knowledge about the
common element of these two views is the idea that activities or coordination problems may reduce the
knowledge is constructed. In addition to emphasizing potential learning advantages of working in groups.
knowledge construction, instead of knowledge transfer, Notwithstanding these potential disadvantages, the use
constructivist environments provide multiple represen- of collaborative activities in web-based courses is
tations of reality and encourage thoughtful reflection particularly important, because the sense of community
based on personal experience. and the typical classroom cues are often absent in this
The distinction between the instructional and self- kind of environment.
paced exploration approaches has also been investi-
gated in the training literature. Instruction-based 2.3. Instructional strategies and the role of
training, typically provided through lectures or discrete instructors
packets of information, is a deductive approach where
learners are presented with the content that needs to be The two learning models (objectivism and con-
learned in its final form [15,16,39]. However, explora- structivism) lead to fundamentally different choices
tion-based learning is inductive, as learners have to regarding methods of knowledge delivery and assess-
discover or construct what needs to be learned by ment of knowledge acquisition. For example, con-
interacting with the content, which is usually presented structivist instructors tend to focus on projects requiring
in the form of specific examples. Some researchers have solutions to problems instead of emphasizing one-way
suggested that exploration is more effective than instruction, which is the core of objectivist approaches.
instruction-based training in promoting the integration While objectivists focus on delivering sequences of
of new knowledge with previous knowledge [12,15,37]. subject-matter topics, constructivists emphasize authen-
tic tasks in a meaningful context rather than abstract
2.2. Group collaboration instruction out of context [30]. Constructivist methods
require technological platforms that provide access to
A key decision in the design and delivery of web- content in a non-linear or non-structured way, and
based courses is the extent to which students will learn in learning tools such as databases, conceptual models,
isolation or through interactions with their classmates. simulations and hypermedia [29,30]. As a result, the
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 781

role of the instructor changes from being the main The premise of constructivism is that learning is an
source of knowledge and answers, to being a coach or active process of constructing rather than acquiring
facilitator in the students’ construction of their own pre-existing knowledge. In this view, knowledge is
knowledge. not just a mental state; it is an experienced relation
Unlike constructivism, which is a learner-centered of things, and learning is the result of a process of
approach where students control the pace of instruction, inquiry at the individual and/or at the social level [24].
objectivism is an instructor-centered approach, where Leidner and Jarvenpaa [29] p. 267 propose that
students are primarily expected to learn from the ‘‘individuals are assumed to learn better when they are
instructor [15]. The role of instructors in a constructivist forced to discover things themselves rather than when
environment shifts from conveying concepts to raising they are told or instructed’’. Thus, the construction of
provocative questions to motivate the students’ search knowledge should foster a deeper understanding and
for answers, a kind of ‘‘Digital Socrates’’ [14]. better learning in students than the mere transfer
However, in objectivist web-based courses, students from the instructor. Based on this conjectures, we
are exposed to frequent instructional interactions hypothesize:
enacted by the instructor, who is in control of the
material and pace of learning [29]. Students in H1. Students engaged in knowledge-construction web-
objectivist courses are more likely to generate questions based courses will achieve better learning outcomes
when they do not understand the material and contact than students in web-based courses where knowledge is
their instructors in their search for answers or directly transferred from the instructor.
clarification. This dependence on the instructor as the
In web-based collaborative courses, students are
main source of knowledge and advice is likely to
engaged in intellectual exchanges with their peers that
generate higher levels of perceived student–instructor
require verbalization of their current understanding of
communication for students in objectivist courses than
the material and the articulation of their cognitive
for students in constructivist environments.
processes. This articulation can be motivated by the
Collaborative courses are designed around constant
need to explain concepts to peers, or to make explicit
and frequent interactions among learners, where most of
their problem-solving strategies [34]. In addition to
the communication takes place. These exchanges can
these cognitive benefits, working with peers also
occur in public discussion forums, open to all the
provides socio-emotional advantages [21,41], both of
students enrolled in a class, or in smaller group sessions,
which are conducive to better learning outcomes. In
where only students of the same team can participate
contrast, students in individual-based courses have
[23]. The role of the instructor in collaborative
limited possibilities for developing these higher-order
environments resembles that of the facilitator that
cognitive processes associated with group interaction
raises questions to engage students in a dialog among
[13,22]. Therefore, we hypothesize:
themselves and periodically steps in to synthesize key
ideas from group discussions [19]. In contrast, students H2. Students in web-based courses structured around
in individual courses are expected to work on their own. collaborative activities will achieve better learning out-
Although contact with other students probably takes comes than students in individualized web-based
place, it is neither required nor necessary. Therefore, courses.
students in collaborative courses are likely to experi-
ence higher levels of learner–learner communication Based on the theorized effectiveness of constructi-
than students in individually oriented courses. vism and the learning advantages of working in
groups, the combination of knowledge construction
2.4. Learning effects with group collaboration is expected to have positive
synergistic effects on learning outcomes because it
Learning involves the acquisition of knowledge, requires students to construct knowledge by interact-
which is evidenced in changes of behavior leading to ing with others. Thus, learning is the result of a
better performance [2]. Since knowledge acquisition is process of inquiry where meaning is socially
not directly observable, the learning effects can be constructed [24]. The potential for achieving a deeper
manifested in terms of improved perceptions of learning understanding and more learning when knowledge is
or better performance in assessment and evaluation constructed [29], in conjunction with the advantages
activities. Hereafter, we refer to these two constructs as of verbalization and articulation when students are
learning outcomes. learning from others in a collaborative environment
782 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

[11], are expected to result in the best levels of materials and combine information from different
learning. Thus, we hypothesize: sources. This process of investigation and discovery,
usually presented in the form of a complex project or
H3. Students in collaborative constructivist web-based
case study, allows the student to understand the subject
courses will achieve the best learning outcomes.
matter and create knowledge.
A construction-group course is also based on the
3. Research methods premise that students search for meanings instead of
receiving packaged lectures. Rather than engaging in
3.1. Experimental design individual investigations, students in this type of course
work together in small groups to discuss cases and/or to
The combination of knowledge construction with the collaborate on projects. This kind of environment
presence of group collaboration produces a 2  2 requires students to interact with others in the knowl-
framework with four different categories: transfer- edge construction process.
individual, transfer-group, construction-individual and The goal of the transfer-group approach is to allow
construction-group (see Fig. 1). These categories students to work together on problems with objectively
describe four possible web-based learning environ- correct answers that require them to express the specific
ments. facts and concepts they have memorized or learned.
A transfer-individual course is based on the direct Accordingly, the main difference between the transfer-
knowledge transfer from the instructor to the students, group approach and the construction-group category is
typically via prepared lectures posted on the course web that in the former category students do not create new
site weekly or by modules. Students in these courses are meaning, but use knowledge previously transmitted (or
required to learn and master the material individually directly made available) by the instructor. It may be
and demonstrate their assimilation of pre-ordained argued the transfer-group approach requires students to
knowledge by working individually in the completion search for pre-ordained knowledge and therefore should
of assignments or exams. be considered a form of constructivism. Strictly
A transfer-group course is also based in a one-way speaking, however, students in these kinds of courses
transmission of knowledge from the instructor to the are not required to create new meaning or build
students through the use of lecture notes. However, this knowledge in order to solve the problem, but rather
type of course requires the students to work together and reproduce existing concepts, principles or methods. In
to co-operate in the completion of group projects and doing so, students primarily learn with others and
other group-based assignments. The mastery of the experience some of the cognitive benefits of group
material is accomplished not only by individually work.
assimilating content, but also by working in group The novelty of this research framework is the
assignments that reinforce the learning process. combination of instructor-led knowledge transfer with
A construction-individual course provides an envir- the use of collaborative activities. Typically, research
onment where students construct their own knowledge in technology-mediated learning environments associ-
independently, by actively interacting with the subject ates group collaboration with knowledge construction.
matter. Instead of receiving modules of pre-ordained This may not be always the case, however. Even when
knowledge (as in transfer courses), students gather direct instruction is used as the knowledge transfer

Fig. 1. Experimental design.


R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 783

approach, students may be engaged collaborative delivered using the Blackboard CourseInfo platform.
learning activities that require them to demonstrate Fifteen different instructors taught the courses. Thirty-
what they have learned [24]. The addition of group two courses had an on-site meeting to begin the course
collaboration to traditional knowledge transfer meth- while seven had an on-site meeting at the end of the
ods overcomes one of the major drawbacks of this course to administer the final exam. The remaining
approach because students are no longer passive section did not have any face-to-face meetings on-
recipients of content. campus. Class section enrollments ranged from 9 to 35
students. Table 1 provides more detail about the
3.2. Sample courses, the number of sections, enrollment information
and respondents by course.
The experimental framework allows us to analyze
the effects of knowledge construction and group 3.3. Data collection methods
collaboration on learning outcomes. To test our
hypotheses regarding the separate and joint effects of Three data sources were used in this study: students’
these two factors, we conducted a field experiment in an responses to a questionnaire, instructors’ interviews to
MBA program of a Midwestern U.S. university. We determine course characteristics, and secondary sources
surveyed 40 online class sections from summer 2000 to identify course design indicators. Data from students
through summer 2002. The courses covered a range of was collected in a post-course survey. Students enrolled
topics including Strategy, Organizational Behavior, in courses that had a final on-campus meeting
Project Management, International Business, Human completed the questionnaire in the last class. Students
Resources, Finance, Accounting, Management Infor- who did not come to campus for a final meeting were
mation Systems, E-Commerce, and Special Topics in sent the survey via e-mail. The non-responding students
Management. In most cases, more than one section per were mailed a copy of the questionnaire. The usable
course was offered during the time frame of the study response rate was 66% (579 of 872). For each of the
and included in the sample. respondents, we also collected the final grades obtained
These courses were delivered entirely online. Six of in the course.
them were administered via its own web site using To classify the sections in the study on the basis of
Lotus LearningSpace software and the others were the knowledge construction/group collaboration frame-

Table 1
Courses, class sections, enrollments, and respondents
Course No. of class sections Total enrollment No. of respondents
Foundations of finance 2 39 20
International financial management and investments 1 14 10
Managerial accounting/cost management 2 52 38
Management information systems integration 1 31 22
Management information systems analysis and design 1 16 6
Managerial problem solving 1 21 20
Human resources management 3 82 56
International business 4 98 62
Organizational foundations 3 57 30
Organizational leadership and change 2 46 34
Personal and professional development 2 52 42
Strategic management 3 69 48
Marketing strategy 1 26 14
E-commerce to E-business 1 23 8
Classic and contemporary literature in business 3 63 43
Planning for management in the future 2 49 34
Environmental management 1 16 11
Introduction to project management 3 79 46
Advanced topics in project management 2 25 13
Project execution and control 1 12 8
Business environments: law, regulation, and ethics 1 22 14
Total 40 872 579
784 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

work, we used a two-step process. First, we identified a (transfer-group, construction-individual, and construc-
set of four course design indicators, namely lectures, tion-group) according to the instructors’ responses.
small group discussion forums, projects and exams. By To study the nature of the interaction that took place
reviewing the corresponding course web sites, we in the web-based courses, several items were included
determined how many courses used lecture notes (in the in the post-course student questionnaire. Student
form of PowerPoint and in the form of audio lectures), perceptions of learner–instructor communication were
and how many used small discussion forums. We also measured using seven items and student perceptions of
determined by reviewing the course syllabi the use of learner–learner communication were measured using
projects (individual and/or in groups) and the use of four items. Both of these scales were adapted from
exams, as assessment techniques. Based on these instruments used by [38].
indicators, we established an initial course classifica- The main dependent variables used in this study are
tion. However, there were a number of courses whose perceived student learning and actual grades. The
classification was not clear because the information perceived learning scale was originally developed by
from secondary sources was not definitive. [1,22], and adapted for this study. Perceived learning
Therefore, as a second step, we conducted semi- measures the extent to which students believe they
structured interviews with all the instructors. These have learned in the course. Higher levels of perceived
interviews probed their knowledge delivery approach learning are likely to result in higher student
and assessed whether the predominant type of learning satisfaction, which in turn should translate into higher
activity in their courses was individual or group-based. ratings for the course, often used as a measure of
To learn about the instructional methods, faculty were success [28].
asked whether their courses were focused primarily on Factor loadings and reliability results for the learner–
fact/concept dissemination via web-based lectures, or instructor, learner–learner and perceived learning scales
based on knowledge construction by students. To are shown in Table 2. All the items measuring each
investigate the presence of the group-learning para- perceptual variable clearly loaded onto its correspond-
digm, faculty were also asked to compare the proportion ing factor. These loadings range from 0.64 to 0.86
of individual and group learning activities in their indicating a strong correlation between each of the
courses and choose the predominant method. The items and the variable it measured. In addition, there are
interview protocol is presented in Appendix I. We no cross-loadings greater than 0.45. This analysis
adopted a final course classification according to the provides evidence of good convergent and discriminant
information provided by the instructors. validity.
The reliability coefficients of the scales are 0.92,
3.4. Operationalization of variables 0.85 and 0.94 for the learner–instructor, learner–learner
and learning perception scales, respectively. In addition,
Courses in the sample were classified according to the except for one item in the learner–learner scale, all the
knowledge delivery approach (transfer versus construc- other single item coefficients are higher than 0.80.
tion) and the collaboration dimension (individual versus Table 3 shows the reliabilities of each item with the
group). These two dimensions were operationalized with scale and the scale Cronbach’s a coefficients. These
two categorical variables: knowledge construction, KC reliabilities are high and well above the cutoff rate of 0.7
(0 for transfer and 1 for constructed) and group recommended by [33].
collaboration, GC (0 for individual and 1 for groups). The second dependent variable is the final grade that
Based on the instructors’ interview responses, 4 sections students earned in the courses. This is a measure of
were classified as transfer-individual, 14 sections were objective performance, and it is an indicator of actual
categorized as transfer-group, 9 sections were placed in learning. Earlier research indicates that self-reported
construction-individual and 13 sections were categorized learning is often inconsistent with objective perfor-
as construction-group. In some cases, two sections of the mance measured by grades [28]. While learning
same course were classified in different cells of the perception indicates the extent to which students feel
framework, based on the answers provided by each of the they learned in the course, actual learning measures the
instructors. For example, two sections of Strategic degree to which they actually learned based on the
Management were classified in the construction-group performance assessment methods used by the instruc-
category, while the third was placed in the transfer-group tors. Therefore, the use of final grades as a dependent
cell. Similarly, the three sections of human resources variable incorporates a more objective measure of
management were assigned to different categories knowledge acquisition.
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 785

Table 2
Factor loadings and cross loadings of the indexes
Learner–instructor Learner–learner Perceived learning
Learner–instructor
The instructor frequently offered opinions to students 0.80 0.04 0.24
The instructor frequently asked the students questions 0.78 0.19 0.24
The instructor frequently attempted to elicit student interaction 0.78 0.25 0.29
Interaction between the instructor and the class was high 0.76 0.20 0.23
Students often asked the instructor questions 0.74 0.03 0.16
Students often stated their opinions to the instructor 0.71 0.22 0.21
In general, the instructor was effective in motivating the 0.64 0.33 0.39
students to interact in this course
Learner–learner (all reverse scored)
Students seldom answered each other’s questions 0.17 0.79 0.11
In this class, students seldom stated their opinions to each other 0.15 0.78 0.15
There was little interaction between students 0.28 0.71 0.16
The students seldom asked each other questions 0.29 0.69 0.18
Perceived learning
Good understanding of basic concepts 0.22 0.11 0.86
Identified central issues in field 0.21 0.11 0.85
Course quality compared favorably to other courses 0.24 0.07 0.82
Improved ability to integrate facts 0.20 0.14 0.80
Learned to see relationships between ideas 0.25 0.14 0.77
Improved ability to communicate clearly 0.33 0.12 0.68
Course served my needs well 0.27 0.18 0.60
Satisfied with the course 0.31 0.15 0.58
High quality class discussions 0.38 0.43 0.47

3.5. Statistical analyses 4. Results

The unit of data collection and the unit of analysis 4.1. Course classification and manipulation checks
for testing our hypotheses was the individual student.
However, since the students were enrolled in sections We validated the course classification derived from
when they received the treatment (i.e. experienced the the instructors’ interviews, using the four course design
web-based course), it is likely that the interaction indicators (lectures, discussion forum, projects and
between students and their common experiences for exams) and our own original classification for each
being in the same class will influence their percep- course. First, we coded each indicator with dummy
tions. This influence or class level effect implies that variables and analyzed these variables vis-à-vis the
the error terms contain a systematic component [35]. course classification. According to the premises of the
The violation of independence is one of the main knowledge delivery approaches, there should be a
reasons for not using traditional regression models. preference for lectures in direct knowledge transfer
Ignoring the group effect could lead us to falsely courses, and a preference for projects in knowledge
reject the null hypothesis (no effect of treatment) construction courses. In addition, individually oriented
because the degrees of freedom are artificially courses should be structured in terms of individual
increased [40]. projects and exams, instead of group-based exercises. In
To take into account that students are nested units contrast, collaborative courses should use group
within classes, and these sections are in turn, grouped in projects and small discussion forums to complement
different treatments, we used multilevel analysis of (or perhaps replace) individual assessment methods.
variance, also known as hierarchical analysis of Consistent with the premises of each approach, we
variance (HANOVA). This technique allows studying noticed that none of the transfer-individual courses in
the effects of course characteristics on individual level our sample used team projects, while all of the
outcomes, while controlling for the fact that students collaborative sections used small group discussion.
received the treatment in a group setting or web-based We also tested for statistical differences between the
course [18]. proportions of courses with each of these indicators,
786 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

Table 3
Scale reliability
Scale Items Cronbach’s a coefficient
Learner–instructor 0.92
The instructor frequently offered opinions to students 0.91
The instructor frequently asked the students questions 0.91
The instructor frequently attempted to elicit student interaction 0.90
Interaction between the instructor and the class was high 0.91
Students often asked the instructor questions 0.92
Students often stated their opinions to the instructor 0.91
In general, the instructor was effective in motivating the students to interact in this course 0.91
Learner–learner 0.85
Students seldom answered each other’s questions 0.82
In this class, students seldom stated their opinions to each other 0.82
There was little interaction between students 0.79
The students seldom asked each other questions 0.81
Perceived learning 0.94
Good understanding of basic concepts 0.93
Identified central issues in field 0.93
Course quality compared favorably to other courses 0.93
Improved ability to integrate facts 0.93
Learned to see relationships between ideas 0.93
Improved ability to communicate clearly 0.93
Course served my needs well 0.93
Satisfied with the course 0.93
High quality class discussions 0.94

using the course-section as the unit of analysis. Table 4 between construction and transfer courses. In terms of
presents these results along with the proportion of individual and collaborative assignments, there is a
courses in each condition where the dummy variable for marginally significant difference between the propor-
the indicator is 1. tion of individual sections using exams when compared
When compared to knowledge construction sections, to collaborative courses (54% versus 30%, z-value 1.48;
a larger proportion of knowledge transfer courses use p < 0.10). Also marginally significant is the proportion
lecture notes (z-value = 2.40, p < 0.01), and audio of collaborative courses using group projects in
lectures (z-value = 2.08; p < 0.05) as vehicles for comparison to individual-based courses (37% versus
disseminating knowledge. No significant differences 15%; z-value = 1.4; p < 0.1). Finally, the preference for
are found in the use of individual or group projects small discussion forums is more noticeable in group-
based courses than in individualistic ones. Every
collaborative section used small group discussion,
Table 4 while only 62% of the individual sections did. Thus,
Tests of differences in proportions of course design indicators
when analyzed separately, these course design indica-
Knowledge construction Transfer Construction z-Value tors based on syllabi information and web site
course classification (%) (%) inspection, are consistent with the information provided
Lecture notes 100 73 2.40** by the instructors regarding the nature of their courses.
Audio lectures 78 45 2.08* For a second validation, we compared our original
Individual projects 89 95 0.78 author-derived course classification based on the course
Group projects 22 36 0.39
design indicators with the final one determined from the
Group collaboration Individual Collaborative z-Value instructors’ interviews. The level of agreement between
course classification the two classifications is 0.8 (out of the 40 sections, 8
Individual projects 85 96 1.31
were placed in different cells of the framework). The
Exams 54 30 1.48$ disagreements were solved in favor of the instructor-
Group projects 15 37 1.4$ provided classification. Taken together, these analyses
Small groups 62 100 3.44*** validate the results of the interviews and increase the
Significance levels: *p < 0.05; **
p < 0.01; ***
p < 0.001; $p < 0.10. reliability of our course classification.
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 787

As an additional manipulation check, we analyzed two means is small, this result is statistically significant
the learner–instructor and learner–learner communica- due to the separation of the nested factor from the error
tion scales using a hierarchical ANOVA model with two term in our analyses. When the systematic component
main factors (knowledge construction and group created by the class level effect is removed from the
collaboration), their interaction and one nested factor residual error, the magnitude of the error term is
due to the course/section. According to the premises of reduced. If the nested factor is treated as the error term,
knowledge delivery, higher levels of learner–instructor the significance of knowledge construction is p = 0.06,
communication should take place in knowledge transfer as indicated in the footnote of the table. Both analyses
courses than in knowledge construction courses, and suggest that the perception of learner–instructor
higher levels of learner–learner communication should communication is higher for students in knowledge
occur in collaborative courses when compared to transfer courses than for students in knowledge
individually oriented courses. Table 5 shows the means construction courses.
by condition and the results of these tests. The group collaboration factor in the learner–learner
The knowledge construction factor is significant in HANOVA model is significant (F = 112.11; p < 0.001)
the learner–instructor HANOVA model (F = 34.15; and the marginal means of the collaborative versus
p < 0.001), and the marginal means indicate that individual students are in the expected direction (5.33
students in courses where knowledge is transferred versus 4.29). The significance of this factor does not
(as opposed to constructed) report higher levels of change when the nested factor is used as the error term.
communication with their instructor (4.95 versus 4.85, Thus, students in group-based courses report higher
respectively). Although the difference between these perceptions of communication with their classmates,
than students in individually oriented courses.
Table 5
HANOVA results on interaction variables 4.2. Test of hypotheses
HANOVA results Learner–instructor Leaner–learner
communication, Fa communication, Fb Our hypotheses predicted relationships between the
Model 11.58 ***
9.50 *** main factors (knowledge construction and group
Knowledge construction 34.15*** 39.81*** collaboration) and the learning outcomes (H1 and
(KC) H2, respectively) and an interaction effect (H3). We test
Group collaboration 81.04*** 112.11*** each hypothesis using both learning perception and
(GC)
grades as dependent variables.3 The results of these tests
Interaction (KC  GG) 73.22*** 30.91***
Nested factor 9.70*** 6.37*** are shown in Table 6. We report in the table the analysis
(course–section) when the nested factor (course/section) is separated
R2 0.46 0.41 from the error term and include footnotes to show the
results when the nested factor is used as the error term.
n Means Means For both dependent variables, hierarchical MANOVA
Cell means model results on the nested factor (course/section) are
Transfer/individual 36 3.45 (1.23) 3.31 (1.37) significant at 0.1%. These results indicate the presence
Transfer/collaborative 191 5.23 (1.04) 5.30 (1.15)
of group level effects due to the particular character-
Construction/individual 128 4.86 (1.27) 4.56 (1.61)
Construction/collaborative 221 4.84 (1.31) 5.35 (1.34) istics of each section. The inclusion of the nested factor
controls for the variations in the particular implementa-
Marginal means
Transfer 227 4.95 (1.26) 4.99 (1.39)
tions of each course/section.
Construction 349 4.85 (1.29) 5.06 (1.38) According to H1, students engaged in knowledge
Individual 164 4.55 (1.39) 4.29 (1.64) construction were expected to achieve better learning
Collaborative 412 5.02 (1.21) 5.33 (1.14) outcomes than students in courses where knowledge is
Notes: Standard deviations in parentheses. Significance levels:
*
p < 0.05; **p < 0.01; ***p < 0.001.
a 3
If the nested factor (course–section) is used as the error term, The correlation between the two dependent variables (grades and
instead of the residual error, the significance of knowledge construc- perception of learning) is 0.13, and this correlation is significant at
tion is 6%. The other factors retain the same levels of significance. 1%. Its low level confirms that objective and perceived learning are
b
If the nested factor (course–section) is used as the error term in this two different aspects of learning. However, since both dependent
analysis, instead of the residual error, group collaboration is signifi- variables are significantly correlated, we use a hierarchical MANOVA
cant at 1% and knowledge construction and the interaction factor are models. The results of separate HANOVA analyses are similar to
each significant at 5%. those reported here.
788 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

Table 6
Hierarchical MANOVA results on perceived learning and grades
HANOVA results Wilks’ l Perceived Grade, Fb
learning, Fa
Model 3.71*** 4.91***
Knowledge construction 2.08 4.25* 14.84***
(KC)
Group collaboration (GC) 6.17** 32.80*** 11.46***
Interaction (KC  GG) 3.71* 18.08*** 9.87**
Nested factor 2.97*** 4.63*** Fig. 2. Interaction effects on learning perception.
(course–section)
R2 0.21 0.26
second dependent variable, the marginal means of
n Perceived learning Grade
grades in collaborative courses are higher than those
means means
in the individual courses (3.73 versus 3.66) and the
Cell means group collaboration factor is significant (F = 11.46;
Transfer/individual 36 4.01 (1.94) 3.38 (0.83)
p < 0.001). Thus, H2 is supported for both dependent
Transfer/collaborative 190 5.20 (0.97) 3.72 (0.52)
Construction/individual 128 4.63 (1.39) 3.75 (0.38) variables.
Construction/collaborative 221 4.86 (1.31) 3.74 (0.38) The interaction effect hypothesis (H3) predicted that
Marginal means
students in construction-group courses were expected to
Transfer 226 5.01 (1.25) 3.66 (0.59) achieve the highest levels of perceived learning across
Construction 349 4.77 (1.34) 3.74 (0.38) conditions. With respect to learning perception, the
Individual 164 4.49 (1.54) 3.66 (0.53) interaction effect of knowledge construction and group
Collaborative 411 5.01 (1.17) 3.73 (0.45) collaboration is significant (F = 18.08; p < 0.001), but
Notes: Standard deviations in parentheses. Significance levels: not in the expected direction. The means by condition
*
p < 0.05; **p < 0.01; ***p < 0.001. indicate that the highest perceptions of learning
a
If the nested factor is used as the error term in the learning
perception model, group collaboration is significant at 1% while
correspond to students in transfer-group courses
knowledge construction and the interaction of both factors are sig- (5.20). Fig. 2 presents graphically the interaction
nificant at 5%. between knowledge construction and group collabora-
b
If the nested factor is used as the error term in the model with tion on learning perception. Since the line segments
grades as dependent variable, knowledge construction is marginally
cross each other, there is a disordinal interaction. The
significant at 10%.
rank order of the effectiveness of the different
conditions on learning perception is: transfer-group,
transferred. The hierarchical MANOVA model result construction-group, construction-individual and trans-
for knowledge construction on learning perception is fer-individual. The combination of knowledge con-
significant (F = 4.25; p < 0.05), but the marginal means struction and group collaboration does not produce the
indicate that students in knowledge construction highest means. Therefore, H3 is not supported for
courses report on average lower learning perception learning perception.
than students in knowledge transfer courses (4.77 versus In terms of performance (grades), the lowest
5.01, respectively). Thus, H1 is not supported for averages are found in students enrolled in transfer-
learning perception. In the test of H1 using grades as a individual courses (means 3.38), whereas in the other
dependent variable, the results are opposite. Students in conditions the averages are in the 3.7 range. The
courses based on knowledge construction achieve significance of the interaction factor (F = 9.87;
higher grades than their counterparts in transfer courses p < 0.01) indicates that the presence of either knowl-
(means 3.74 versus 3.66; F = 14.84; p < 0.001). edge construction or group collaboration has a positive
Therefore, H1 is supported for grades. effect on student grades. However, the presence of both
The second hypothesis predicted better learning does not seem to have any positive synergistic effect, as
outcomes for collaborative courses when compared to hypothesized (H3). Fig. 3 shows graphically the
individualistic ones. For our first dependent variable, interaction effects of the two factors on final grades.
learning perception, the marginal means are higher for Knowledge construction is superior to knowledge
collaborative courses than for non-collaborative courses transfer but not by the same amount for the
(5.01 versus 4.49), and the group collaboration factor is individual-oriented and group-oriented courses. In fact,
significant (F = 32.80; p < 0.001). Likewise, for our the final grades for collaborative courses are very
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 789

instructor, because of his/her dual role as knowledge


expert who delivers the material and group facilitator
who guides the student teams. Because of this double
duty, the level of perceived learner–instructor commu-
nication is significantly higher in transfer-group than in
transfer-individual sections.
In terms of knowledge acquisition measured by final
grades, our results indicate that the presence of either
group collaboration or knowledge construction sig-
Fig. 3. Interaction effects on grades.
nificantly improves the students’ final grades. This
finding also supports the importance of collaborative
similar, regardless of the knowledge delivery approach. approaches documented by previous research, and
Since the combination of collaboration and construction suggests that knowledge construction is also beneficial
does not produce the highest scores, H3 is not supported for student performance. The combination of both,
for grades. however, does not produce positive synergistic effects
on grades. A possible explanation is that better learning
5. Discussion of results can be equally achieved by constructing knowledge or
by articulating knowledge to others, and that these
The findings for the learner–instructor and learner– two mechanisms substantially overlap when used in
learner communication manipulation checks suggest combination.
that the courses were delivered as intended, and as One common finding for both dependent variables is
reported by the instructors in the interviews. These that students in transfer-individual courses exhibited
student level perceptions provide additional support to the lowest perceptions of learning and the lowest grades
our course classification variables determined from the across all conditions. The number of sections classified
information provided by the instructors and the analysis in this category (only four) and the learning outcomes in
of course design indicators. In addition to serving as these courses indicate that this mode is neither the most
manipulation checks, these measures are also evidence preferred by instructors nor the most successful in terms
of the communication dynamics in each category of the of learning in web-based environments. Given that the
experimental framework. One noteworthy finding in perceptions of instructor–learner communication are
this regard is the low levels of perceived learner– the lowest for this category, it seems that instructors in
instructor communication in transfer-individual this mode used the technology as a replacement for their
courses. This suggests that the delivery of knowledge role as knowledge communicators. To achieve better
in these sections is mostly achieved via a set of learning outcomes, particularly for graduate manage-
packaged materials such as lecture notes and audio ment education, participants in online courses need to
lectures that is not complemented with live participation collaborate with their counterparts in virtual teams and/
or explanations from the instructor. or to be engaged in the construction of their own
In terms of learning perception, our results indicate knowledge [8,37].
that the most successful combination for the design of Taken together, the results on both dependent
web-based courses is found in the transfer-group variables show an interesting paradox. Students who
category. Courses in which knowledge is directly are in knowledge construction courses do not perceive
transmitted from the instructor to the students and that they have learned as much as students in knowledge
collaborative approaches are used as the primary learning transfer courses. However, their final course grades
mode scored significantly higher in perceived learning show that they learn more or better than students in
than did courses that were conducted using other knowledge transfer sections. Because ‘‘constructive’’
combinations of knowledge delivery and collaboration. learning is less structured and requires students to make
This finding confirms previous conceptual and empirical sense of the material, students may feel they have
work supporting the importance of collaborative learning learned less, a perception that is at odds with their actual
in web-based learning [23,29]. It is interesting to note that knowledge acquisition evidenced in their final grades. It
transfer-group courses are where the higher perceptions is also possible that due to the different nature of student
of learner–instructor and learner–learner communication performance in graded activities in both types of
occur. These results suggest that students not only courses (for example, term papers or essays for
communicate more with each other, but also with the knowledge construction courses versus multiple choice
790 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

exams for knowledge transfer courses), performance Second, we did not explicitly take into account the
assessment is more subjective and results in higher particular characteristics of the instructors or the subject
grades when students construct their own knowledge matters taught in these courses [7]. We used the
than when students reproduce objective content. In professors’ opinions to classify the courses, without
either case, this finding underscores that learning imposing our own bias with respect to which courses
perception and performance are not always consistent, were more suitable for knowledge construction versus
as they are measuring different aspects of the learning knowledge transfer, or for collaborative versus indivi-
experience. dual activities. In fact, we classified several sections of
In sum, by separating the effects of knowledge the same course in different cells of the framework
construction from the effects of group collaboration, our based on the results of the instructors’ interviews. This
results indicate that collaborative approaches result in added flexibility to the comparative field study and
better learning perception, even in courses where allowed us to examine the effects of the two factors
knowledge construction is not the dominant or the most beyond the particular characteristics of each instructor
efficient approach. In terms of grades, we found that and subject matter. However, we control for the
either one of these factors has a positive effect on combined effect of subject matters and instructor
student performance. The evidence provided by this characteristics with the use of the nested factor in
research is stronger than previously reported results on our analyses.
this area because it includes multiple courses and Another limitation is that we developed our research
compares different approaches at once. To our knowl- framework by combining two dimensions (knowledge
edge, no other empirical study has attempted this multi- construction and group collaboration) that were
course comparison before. operationalized with categorical indicators. In our
Unlike prior multi-disciplinary or multi-course faculty interviews, we discovered that sometimes
empirical studies, our results are robust to the instructors combine both knowledge delivery models
systematic errors introduced by the use of different but subscribe predominantly to one approach over the
sections in each mode (group level effects). Using other. Similarly, our analysis of course design indicators
hierarchical analysis of variance, we separate the effects revealed that instructors sometimes use both group
of the main factors (knowledge construction and group collaboration and individual activities. In this study, the
collaboration), and their interaction from the nested perceptions of learner–learner and learner–instructor
factor (course–section). This technique has been communication reported by the students confirmed the
previously used by [4,27] to control for the effects of classification obtained from the instructors’ interviews.
working in small groups when collecting and analyzing Follow-up studies may choose to operationalize the
individual level data. However, to the best of our knowledge delivery approaches (direct transfer versus
knowledge, we are the first to use this technique in a constructive) as continuous variables, rather than
multi-course study of online environments where the categorical variables. This would, of course, require
nested factor is the course/section. an intensive and extensive validation process to ensure
the validity and reliability of these metrics when applied
5.1. Limitations to a multi-course, multi-instructor sample such as this
one.
There are several limitations that must be taken into Our dependent variables (learning perception and
account in the interpretation of results. First, since the final grades) may also be affected by methodological
hypotheses were tested in a field experiment in the limitations inherent to the mechanisms whereby these
context of an MBA program rather than in a controlled measures are collected (self-reports in the case of
laboratory experiment, some internal validity is lost at perceived learning) or produced (determined by the
the expense of a greater ability to generalize the results. instructors in the case of final grades). Nevertheless, the
Although these findings may not be exactly reproduced use of both dependent variables offers a more complete
in other contexts such as other kinds of programs or picture of the learning experience from the students’
university settings, the fact that we obtain significant perspective. It is possible that the discrepancy in the
effects attests to the importance of the factors examined findings of learning perception and final grades for
here. As Dennis et al. [17] p. 186 point out ‘‘If the knowledge construction and knowledge transfer courses
processes of interest are still observed in an uncon- may have resulted from a familiarity bias. Students were
trolled environment with all of its extraneous influences, probably more familiar with instructional methods than
this bespeaks of the power of those processes.’’ with constructivist approaches and this familiarity may
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 791

have produced higher levels of perceived learning in learning outcomes when their work is structured around
knowledge transfer courses, despite the fact that final collaborative activities and/or around knowledge con-
grades are in the opposite direction. struction exercises.
Another limitation is found in the types of measures There are also important implications for instructors
used in this project. Since this is a quantitative study, it of online environments, identified by prior research but
is likely that qualitative sources of data such as content presented in a new light in this study. Although their
analysis of web-based interactions or students inter- role may be different, instructors are principal actors in
views, may add more depth to our findings [9]. We any environment [14,42]. Regardless of the knowledge
welcome the efforts of future researchers to refine and delivery approach, successful web-based environments
improve on these methods and measures in future require active involvement from instructors. Our study
studies. shows that in the context of web-based MBA courses,
There are some advantages offsetting these limita- the technology cannot be used to replace instructors, but
tions. The main strength of this study is the validation of rather to complement their task. From the design of the
the course classification decisions (determined from the course to its delivery, instructors must create active and
instructors’ interviews) with two additional methods engaging learning environments for their students and
and data sources. First, we used course design indicators closely monitor the development of these environments
extracted from syllabi and course web sites to ascertain while the course is in progress.
whether the courses had different structures. Second, we The research implications of this study are manifold.
also used the students’ perception of learner–instructor Future multi-disciplinary studies should expand upon
and learner–learner communication from the post-test the predictors of outcomes of web-based courses in two
survey to confirm whether the courses were delivered as directions – individual level variables and course level
described in the syllabi and reported by the instructors. variables – and use multilevel analysis techniques to
This validation is equivalent to manipulation checks assess the significance of these predictors. In addition,
usually performed in controlled experiments. the knowledge construction/group collaboration frame-
Another strong point of this study, which sets it apart work developed here can be used to examine the
from previous empirical research in multi-course effectiveness of web-based courses at the undergraduate
samples, is the use of hierarchical data analysis methods level and/or in other institutional contexts. Finally,
to control for course–section effects. researchers should focus their efforts on the develop-
ment of alternative measures to capture the underlying
5.2. Implications learning model implemented in each course. These are
intriguing areas for future research endeavors.
These findings have important implications for the
design of online courses. Some practical implications 6. Conclusions
include the importance of incorporating collaborative
learning techniques to create virtual teams of students This research contributes to the IS literature in
and foster the growth of a learning community. Group- technology-mediated learning by analyzing the effects
communication facilities, such as discussion boards and of knowledge construction and group collaboration in a
chat rooms should be brought to the forefront of the multi-course sample, and showing how each dimension
design and delivery of online courses, rather than being contributes to learning perception and learning perfor-
an optional or an after-thought activity. This research mance. With the use of a multi-course sampling frame,
shows the importance of these group communication we assess the effects of each factor separately and in
facilities to improve the level of learning outcomes combination across a range of subject matters and
achieved by students in online courses. validate the commonly held assumption that collabora-
This study also shows the importance of developing tive approaches are better than individual approaches
challenging learning environments where students can for online courses. Our findings also provide new
discover and create knowledge on their own [6]. These evidence of the importance of knowledge construction
settings are particularly beneficial for knowledge approaches in web-based MBA courses, particularly in
acquisition (measured by final grades). In light of the terms of final grades.
increasing popularity of web-based courses in graduate Our results suggest that the one-way transmission of
and undergraduate programs and their increasing use in knowledge, where students work individually with a set
corporate training, these findings suggest that online of packaged materials is not the most successful
communities for learning or training may obtain better approach for web-based courses. Such courses scored
792 R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793

lowest on both the dependent variables included in a. Did you provide/present course material and students
our study (perceived learning and objective learning). subsequently demonstrated how well they acquired
Learners in this kind of environments do not receive that material?
the stimulation, the reinforcement and the social b. Did you provide course material from multiple
support usually found in group activities. The social sources and students put together and made sense of
dimension of learning appears to be essential for the the material to create knowledge for themselves?
success of these courses, where the sense of the
classroom may be lost in the technology-mediated A.2. Regarding learning
environment. Likewise, the construction of knowl-
edge seems to be essential to engage and challenge Based on the proportion of individual versus group
learners that are dispersed in time and distributed in activities in your course, which of the following best
place, particularly when they are expected to work represents the PREDOMINANT learning mode in your
individually. course:
This study opens new possibilities for future
research, specifically in the area of multi-disciplinary a. While there may have been some group and or class
studies. Given the amount of research already con- interactions, a student primarily learned individually.
ducted in the field, the comparison between traditional b. The students predominantly learned by interacting
classrooms and online environments in the context of online with other course participants.
single courses appears to be no longer productive. The
new research challenge is to understand how and why References
some online environments are more successful than
others and derive meaningful lessons, which are [1] M. Alavi, Computer-mediated collaborative learning: an empiri-
applicable across several disciplines and for a wide cal evaluation, MIS Quarterly 18, 1994, pp. 159–174.
[2] M. Alavi, D.E. Leidner, Research commentary: technology-
range of courses. This challenge echoes the call for
mediated learning—a call for greater depth and breadth of
more breadth and depth in technology-mediated research, Information Systems Research 12(1), 2001, pp. 1–10.
learning [2], and further suggests that by examining a [3] M. Alavi, D.E. Leidner, Review: knowledge management and
broader sample of courses with a deep understanding of knowledge management systems: conceptual foundations and
the different implementations of virtual environments, research issues, MIS Quarterly 25(1), 2001, pp. 107–136.
these two objectives can be achieved. [4] M. Alavi, G. Marakas, Y. Yoo, A comparative study of dis-
tributed learning environments on learning outcomes, Informa-
tion Systems Research 13(4), 2002, pp. 404–416.
Acknowledgments [5] M. Alavi, B.C. Wheeler, J.S. Valacich, Using IT to re-engineer
business education: an exploratory investigation of collaborative
We are very grateful for the useful suggestions telelearning, MIS Quarterly 19, 1995, pp. 293–312.
provided by the anonymous reviewers and the editor of [6] J.B. Arbaugh, Managing the on-line classroom: a study of
technological and behavioral characteristics of web-based
the SOS section of Information & Management, and for MBA courses, Journal of High Technology Management
the feedback offered by various colleagues at our Research 13, 2002, pp. 203–223.
institutions. [7] J.B. Arbaugh, How much does ‘‘subject matter’’ matter? a study
of disciplinary effects in on-line MBA courses Academy of
Management Learning & Education 4, 2005, pp. 57–73.
Appendix A. Interview protocol for instructors
[8] J.B. Arbaugh, Is there an optimal design for on-line MBA
courses? Academy of Management Learning & Education 4,
I’d like you to think about the course _ you taught in 2005, pp. 135–149.
the semester _. As you know, your course is part of [9] J.B. Arbaugh, R. Benbunan-Fich, In defense of quantitative
larger sample of online MBA courses that we are methods to research networked management learning: a reply
examining in a research project. We are interested in to Hodgson and Watland, Management Learning 35, 2004, pp.
117–124.
comparing the predominant teaching and learning [10] R. Benbunan-Fich, Improving education and training with infor-
modes in those courses. mation technology, Communications of the ACM 45(6), 2002,
pp. 94–99.
A.1. Regarding teaching [11] R. Benbunan-Fich, S.R. Hiltz, Mediators of the effectiveness of
online courses, IEEE Transactions on Professional Communica-
tion 46(4), 2003, pp. 298–312.
Which of the following would you say best [12] J.M. Carroll, R.L. Mack, C.H. Lewis, N.L. Grischkowsky, S.R.
represents the PRIMARY mode of teaching in that Robertson, Exploring exploring a word processor, Human–Com-
course? puter Interaction 1, 1985, pp. 283–307.
R. Benbunan-Fich, J.B. Arbaugh / Information & Management 43 (2006) 778–793 793

[13] J.F. Chizmar, M.S. Walbert, Web-based learning environments [32] J. Lu, C.-S. Yu, C. Liu, Learning style, learning patterns and
guided by principles of good teaching practice, Journal of learning performance in a WebCT-based MIS course, Informa-
Economic Education 30(3), 1999, pp. 248–264. tion & Management 40(6), 2003, pp. 497–507.
[14] N.W. Coppola, S.R. Hiltz, N.G. Rotter, Becoming a virtual [33] J. Nunnally, Psychometric Theory, McGraw Hill, New York,
professor: pedagogical roles and asynchronous learning net- NY, 1978.
works, Journal of Management Information Systems 18(4), [34] G. Piccoli, R. Ahmad, B. Ives, Web-based virtual learning
2002, pp. 169–189. environments: a research framework and a preliminary assess-
[15] S.A. Davis, R.P. Bostrom, Training end users: an experimental ment of effectiveness in basic IT skills training, MIS Quarterly
investigation of the roles of the computer interface and training 25, 2001, pp. 401–426.
methods, MIS Quarterly 17(1), 1993, pp. 61–81. [35] S. Raudenbusch, A. Bryk, Hierarchical Linear Models: Applica-
[16] S.A. Davis, F.D. Davis, The effect of training techniques and tions and Data Analysis Methods, Sage Publications, Thousand
personal characteristics on training end users of Information Oaks, CA, 2002.
Systems, Journal of Management Information Systems 7, 1990, [36] A.P. Rovai, Sense of community, perceived cognitive learning,
pp. 93–110. and persistence in asynchronous learning networks, The Internet
[17] A.R. Dennis, B.H. Wixom, R.J. Vandenberg, Understanding fit and Higher Education 5, 2002, pp. 319–332.
and appropriation effects in Group Support Systems via meta- [37] M. Rungtusanatham, L.M. Ellram, S.P. Siferd, S. Salik, Toward a
analysis, MIS Quarterly 25(2), 2001, pp. 167–193. typology of business education in the internet age, Decision
[18] M. Gallivan, R. Benbunan-Fich, A framework for analyzing Sciences Journal of Innovative Education 2, 2004, pp. 101–
levels of analysis issues in studies of E-collaboration, IEEE 120.
Transactions on Professional Communication 48(1), 2005 [38] A.C. Sherry, C.P. Fulford, S. Zhang, Assessing distance learners’
March, pp. 87–104. satisfaction with instruction: a quantitative and a qualitative
[19] J.K. Gallini, A framework for the design of research in technol- measure, The American Journal of Distance Education 12(3),
ogy-mediated learning environments: a sociocultural perspec- 1998, pp. 4–28.
tive, Educational Technology 41(2), 2001, pp. 15–21. [39] S.J. Simon, J.M. Werner, Computer training through behavior
[20] V. Grover, T. Davenport, General perspectives on knowledge modeling, self-paced, and instructional approaches: a field
management: fostering a research agenda, Journal of Manage- experiment, Journal of Applied Psychology 81(6), 1996, pp.
ment Information Systems 18(1), 2001, pp. 5–23. 648–659.
[21] L. Harasim, On-line education: an environment for collaboration [40] R. Walczuch, R. Watson, Analyzing group data in MIS research:
and intellectual amplification, in: L. Harasim (Ed.), On-line including the effect of the group, Group Decision and Negotia-
Education: Perspectives on a New Environment, Praeger, New tion 10(1), 2001, pp. 83–94.
York, 1990, pp. 133–169. [41] P. Webb, Student interaction and learning in small groups,
[22] S.R. Hiltz, The Virtual Classroom: Learning without Limits Via Review of Educational Research 52, 1982, pp. 421–445.
Computer Networks, Ablex Publishing Corporation, Norwood, [42] J. Webster, P. Hackley, Teaching effectiveness in technology-
NJ, 1994. mediated distance learning, Academy of Management Journal
[23] S.R. Hiltz, M. Turoff, What makes learning networks effective? 40, 1997, pp. 1282–1309.
Communications of the ACM 45(4), 2002, pp. 56–59.
[24] W.L.D. Hung, D.-T. Chen, Technologies for implementing social
Raquel Benbunan-Fich is an associate professor at the SCIS Depart-
constructive approaches in instructional settings, Journal of
ment in the Zicklin School of Business, Baruch College, City Uni-
Technology and Teacher Education 7(3), 1999, pp. 235–256.
versity of New York. She received her PhD in Management
[25] D. Jonassen, Thinking technology, Instructional Technology
Information Systems from Rutgers University, Graduate School of
1993, pp. 35–37.
Management. Her research interests include educational applications
[26] R.C.W. Kwok, M. Khalifa, Effects of GSS on knowledge
of computer-mediated communication systems, asynchronous learn-
acquisition, Information & Management 34(6), 1998, pp.
ing networks, evaluation of Web-based systems and e-commerce. She
307–315.
has published articles on related topics in Communications of the
[27] R.C.W. Kwok, J. Ma, D.R. Vogel, Effects of Group Support
ACM, Decision Support Systems, Group Decision and Negotiation,
Systems and content facilitation on knowledge acquisition,
IEEE Transactions on Professional Communication, Information &
Journal of Management Information Systems 19(3), 2002, pp.
Management, International Journal of Electronic Commerce, Journal
185–229.
of Computer Information Systems and other journals.
[28] D.E. Leidner, M. Fuller, Improving student learning of concep-
tual information: GSS-supported collaborative learning vs. indi-
vidual constructive learning, Decision Support Systems 20, J.B. Arbaugh is the Curwood Endowed professor and a professor of
1997, pp. 149–163. Strategy and Project Management at the University of Wisconsin
[29] D.E. Leidner, S.L. Jarvenpaa, The use of information technology Oshkosh. He is an associate editor of Academy of Management
to enhance management school education: a theoretical view, Learning & Education. Ben’s research interests are in online manage-
MIS Quarterly 19, 1995, pp. 265–291. ment education, international entrepreneurship, the management of
[30] S.-S. Liaw, Designing the hypermedia-based learning environ- rapidly growing firms, and the intersection between spirituality and
ment, International Journal of Instructional Media 28, 2001, pp. strategic management research. Some of his recent publications
43–56. include articles in Academy of Management Learning & Education,
[31] K.H. Lim, L.M. Ward, I. Benbasat, An empirical investigation of Decision Sciences Journal of Innovative Education, Management
computer system learning: comparison of co-discovery and self- Learning, the Journal of Management, Spirituality, and Religion,
discovery methods, Information Systems Research 8(3), 1997, the Journal of Enterprising Culture, and the Journal of Management
pp. 254–273. Education.

You might also like