Pedido 07 - 1

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Five Design Recommendations for Teaching Teenagers’ about

Artificial Intelligence and Machine Learning


Marie-Monique Schaper Mariana Aki Tamashiro Rachel Charlotte Smith
mmschaper@cc.au.dk mariana@cc.au.dk rsmith@cc.au.dk
Aarhus University Aarhus University Aarhus University
Aarhus, Denmark Aarhus, Denmark Aarhus, Denmark

Maarten van Mechelen Ole Sejer Iversen


mvanmechelen@cc.au.dk oiversen@cc.au.dk
Aarhus University Aarhus University
Aarhus, Denmark Aarhus, Denmark

Figure 1: Workshop sessions: a) Intelligence Rating activity; b) Hands-on activity with a digital tool; c) Fictional Storyboard
activity
ABSTRACT for teaching young people about intelligent technologies, an ap-
Technologies based on AI/ML are playing an increasingly promi- proach that can be implemented in future research interventions
nent role in teenagers’ everyday lives. Mirroring this trend is a con- with similar objectives.
comitant interest in teaching young people about intelligent tech-
nologies. Whereas previous research in the field of Child–Computer CCS CONCEPTS
Interaction has proposed curriculum and learning activities that • Human-centered computing → Empirical studies in HCI.
describe what teenagers need to learn about AI/ML, there is still a
shortage of studies which specifically address teenager-centered KEYWORDS
perspectives in the teaching of AI/ML. This paper presents a study Machine Learning; Artificial Intelligence; Technology Education;
of teenagers’ everyday understanding of AI/ML technologies. Using Learning Activities; Learning Tools; Participatory Design; Teenagers;
a thematic analysis of the teenagers’ own explanations during a Computational Empowerment
series of workshops, we present a conceptual map of the teenagers’
ACM Reference Format:
understandings of these technologies. We go on to propose five Marie-Monique Schaper, Mariana Aki Tamashiro, Rachel Charlotte Smith,
general recommendations for the teaching of AI/ML to teenagers Maarten van Mechelen, and Ole Sejer Iversen. 2023. Five Design Recommen-
through the lens of Computational Empowerment. Taken together, dations for Teaching Teenagers’ about Artificial Intelligence and Machine
these recommendations serve as a teenage-centered starting point Learning. In Interaction Design and Children (IDC ’23), June 19–23, 2023,
Chicago, IL, USA. ACM, New York, NY, USA, 12 pages. https://doi.org/10.
Permission to make digital or hard copies of all or part of this work for personal or 1145/3585088.3589366
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for components of this work owned by others than the 1 INTRODUCTION
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission This paper describes what it means to be a teenager in a digi-
and/or a fee. Request permissions from permissions@acm.org. tized world. Taking a teenager-centered perspective, we present
IDC ’23, June 19–23, 2023, Chicago, IL, USA and illustrate teenagers’ understandings of and assumptions about
© 2023 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 979-8-4007-0131-3/23/06. . . $15.00 intelligent technologies. The intelligent technologies Artificial In-
https://doi.org/10.1145/3585088.3589366 telligence (AI) and Machine Learning (ML) are part of teenagers’

298
IDC ’23, June 19–23, 2023, Chicago, IL, USA Schaper et al.

everyday lives. Teenagers, in common with all users of these tech- specifically on teenagers (i.e. 13- to 19-year-olds) and their under-
nologies, are using algorithms that process sensor data from our standing of the technical aspects and ethical-societal considerations
devices and support advanced features of commonly used services. of intelligent technologies [16, 28]. Starting from this perspective,
It is of the greatest importance, however, that people understand our research aims to investigate how we can scaffold teenagers’
these technologies so that they can remain in control of the decision- learning about AI/ML by building on their everyday understanding
making processes related to their development, implementation, of these technologies.
and use. To achieve this, both adults and young people need to We present a series of workshops undertaken as part of a larger
become empowered through “being able to understand how the research project on intelligent technologies in education [5, 36, 41].
technology may affect them, trust it and feel in control” [2]. In This study engaged eleven Danish teenagers aged between 14 and
recent years, international organizations that focus on children’s 15 years in exploring their understanding of AI/ML technologies
rights, such as UNICEF [44] and the United Nations’ International through the lens of computational empowerment [24]. The goal of
Telecommunication Union, have started to examine the ethical con- the study was to use insights gained in the workshops to integrate
siderations affecting young users of these intelligent technologies. the teenagers’ perspectives and to address their knowledge gaps in
Policymakers have provided guidelines highlighting the impor- the development of learning activities and tools in formal education
tance of promoting children’s competencies in AI literacy, compe- contexts. We drew on our intention to scaffold Computational Em-
tencies capable of providing young people with meaningful skills powerment in practice [36] to support children and young people
and experience in understanding AI systems, user interactions, and in making informed decisions about the role of technology in their
outcomes. Further to this, the European Union’s ongoing Digital lives [24]. During the workshops, we employed several different
Action Plan (2021–2027) foresees integrating AI and data-related educational activities, namely (I) Semi-Structured Interviews; (II)
skills into the European Digital Competence Framework, with the Rapid-Prototyping Future Devices; (III) Intelligence Rating; (IV)
aim of empowering citizens to obtain a basic understanding of Fictional Storyboarding.
emerging technologies and to enable them “to engage confidently, Our analysis focused on the students’ utterances as expressed
critically and safely with them” [1]. during the workshop activities. This allowed us both to identify the
This understanding of the urgency of empowerment has sparked teenagers’ understanding of and to identify challenges in support-
a growing interest in the field of technology education, in par- ing their learning about intelligent technologies – their learning
ticular in teaching children and young people about intelligent not only about technological features, but also about the ethical
technologies [12, 32, 36, 47–49]. A new research field has opened and societal implications of AI/ML technologies. On the basis of our
up in this area, comprising two main strands. First, a large body thematic analysis, we developed five general recommendations for
of research is focused on supporting students’ understanding of teaching AI/ML to teenagers: (1) Diversity in learning paths about
technical aspects of AI/ML with the aim of preparing students for intelligent technologies; (2) Human intelligence as an entry point for
graduate studies and for work in the future computing industry artificial intelligence; (3) Accessible technology as an opportunity
[3]. Second, Child–Computer Interaction (CCI) scholars have called for sustainable engagement; (4) Empowerment of action-taking in a
for research, as part of the agenda in K–12 education, on how to digitalized society; (5) Reflection upon real-world dilemmas of intelli-
engage students with the ethical considerations and societal im- gent technologies as wicked problems. Based on these outcomes, we
pact of intelligent technologies [8]. In this regard, Van Mechelen provide two main contributions in this paper:
et al.’s comprehensive literature review [47] recently highlighted
• A conceptual map of the teenagers’ everyday understanding
the shortage of research on the societal and ethical implications of
of AI/ML technologies
emerging technologies as a learning objective.
• Five design recommendations for learning activities and
Starting from this research gap, we believe the first step to design-
tools of AI/ML technologies that can be implemented in
ing learning activities for students is to develop an understanding
future research interventions with similar objectives
of students’ everyday conceptualization of intelligent technolo-
gies. Such an understanding would allow researchers and educators
to explore a number of different entry paths for learning and un- 2 RELATED WORK
derstanding which contextual aspects impact students’ learning Our literature review in this section focuses on three research areas:
processes. Although some previous research has already explored (1) the role of AI/ML technologies in students’ lives and education;
children’s attitudes to and opinions of various types of intelligent (2) researchers’ design recommendations for AI/ML learning activi-
technologies such as voice assistants [13, 14, 46], drones [20], social ties and (3) exploring children’s understanding of AI/ML. Intelligent
robots [30, 33, 52] and educational robotics [28], most of these stud- technologies now permeate so many aspects of society that the need
ies have focused on children’s general perceptions of the technical to educate children about these technologies has become urgent
features of intelligent technologies (e.g. [14, 53] and their attitudes on the agenda of the K–12 school curriculum [43]. Technologies
and opinions during their interaction with them (e.g. [30, 52]). A such as AI and ML are predicted to have a profound impact on
few studies have limited their scope by focusing on specific con- society and our labor market, with close to half of society’s current
cerns around AI (e.g. children’s fears and hopes [28, 33]), while jobs in danger of radical transformation [25]. Numerous initiatives
some others have investigated children’s beliefs about the intelli- have attempted to teach students about intelligent technologies,
gent technologies they are exposed to in everyday situations (e.g. but most have focused on how the technology works [47], rather
[20]). Even fewer of the research studies in this field have focused than (e.g. [5, 11, 35, 40, 41, 50] discussing with students how tech-
nology impacts the individual and society. Despite the importance

299
Five Design Recommendations for Teaching Teenagers’ about Artificial Intelligence and Machine Learning IDC ’23, June 19–23, 2023, Chicago, IL, USA

of these matters, van Mechelen et al. [47] stress that most of the (K–2, 3–5, 6–8, and 9–12). Lin and Van Brummelen [26] organized
research studies in the field of technology education seem to add co-design workshops with teachers (6th to 12th grade) to identify
humanistic reflections upon emerging technologies as merely an opportunities to integrate AI education into the K-12 curriculum.
“afterthought”, rather than making them an integral part of the Based on participatory practices conducted with teachers and re-
teaching activity. searchers, Schaper et al. [36] presented four educational principles
There are a number of reasons for the paucity of initiatives (Closeness, Embodied Learning, Design Process, Decoding) as a means
that attempt to address the ethical and societal implications of to scaffold teenagers’ development of Computational Empower-
intelligent technologies. First, most of the existing resources and ment. However, although some general recommendations have been
curricula focus on computational-driven learning goals [47], which presented for teaching AI in K–12 education, most of these recom-
tends to influence educators to replicate those types of content and mendations do not take the age-specific needs and understanding of
goals. Second, the discussion of humanistic aspects of intelligent teenagers into account. At the same time, it seems that exemplary
technologies in educational contexts seems to be fairly new [47], educational activities and tools for teenagers rarely include learning
which means that not only teachers, but also researchers tend to content that focuses on the ethical and societal aspects of intelligent
be unfamiliar with the strategies, themes, or vocabulary to sup- technologies (e.g. [39, 49]). We argue that exploring teenagers’ under-
port students’ understanding of emerging technologies [36]. Third, standings, misconceptions, and imaginaries can address these gaps
increasing technological complexity and the non-transparency of and can inform the design of education activities to address these
intelligent technologies make it harder for young people to grasp important themes.
their inner workings and their wider impact. This complexity and
difficulty may prevent young people from becoming empowered Exploring children’s understanding of AI/ML Previous re-
agents in a data-driven society [10, 24]. These challenges highlight search in the field of CCI has focused on understanding children’s
the need for design recommendations for AI/ML learning activities conceptualizations, with the goal of proposing design recommen-
in K-12 education that are capable of integrating both technical and dations for intelligent interfaces [53], voice assistants[53], voice-
humanistic aspects of intelligent technologies. assistants [13, 14, 46], smart toys [15], drones [20] and social robots
[33]. Many of these recommendations have focused on the design
Design Recommendations for AI/ML learning activities characteristics of intelligent technologies for young users. For in-
Several scholars have recently presented design recommendations stance, Woodward et al. [53] proposed design recommendations for
and guidelines for learning activities for AI/ML. Several of these error detection and correction in speech, gesture, and writing, with
recommendations are distilled from literature reviews. For instance, the aim of improving children’s (7–12y) user experience. Druga et
Long and Magerko’s [27] suggested several design considerations to al. [14] analyzed children’s understanding (aged 3–10y) of voice as-
support AI developers and educators in generating learner-centered sistants, and suggested design considerations for future child–agent
AI. These design considerations were organized in a conceptual interactions centering on around voice and prosody, interactive
framework based on five key concepts, namely What is AI?; What engagement, and facilitating understanding. Other research studies
can AI do?; How does AI work?; How should AI be used?; and How focused on the impact of social situations and relationships related
do people perceive AI? Tedre et al. [42] highlighted potentials and to intelligent technologies. In this regard, Gamboa [20] presented
pitfalls in teaching about ML. Sanusi and Oyelere [34] proposed design opportunities and took a critical stance on child-drone inter-
strategies for teaching about ML based on active, participatory, and action based on her personal experience living with her family and
design-oriented learning. several intelligent devices during one year. Further, Rubegni et al.
Other studies have presented exemplary learning activities and [33] investigated how children’s (8–14y) hopes and fears with refer-
tools that target teenagers. For instance, based on several years of ence to social robots depended on social interactions and contextual
research, Clarke [9] proposed an AI curriculum for high-school stu- configurations.
dents to demystify the topic of AI, support students’ development of Alternatively, some scholars have compared age-dependent dif-
topic-specific vocabulary, and feature discussions of moral and eth- ferences in conceptualizing intelligent technologies. For instance,
ical impacts of AI systems and usage. Vachovsky et al. [45] carried Druga et al. [15] investigated the differences between children’s
out an AI Summer Program with tenth-grade female high-school (4–10y) and parents’ interaction with smart devices, based on asking
students, which focused on project-based learning. The program them to make assumptions about the intelligence of mice, robots,
combined the application of AI technical skills with humanitarian and their own intelligence in a maze-solving activity. In particular,
purposes (e.g. health sector, natural disaster prevention). Wan et the children focused more often on sensory abilities and social-
al. [49] developed the SmileyCluster tool, a hands-on and collabo- emotional characteristics for the mouse and robot, referenced cog-
rative learning environment to support learning opportunities for nitive abilities less often than their parents did, and tended to over-
ML for high-school students (15–17y) with widely differing STEM estimate the capacity of the robot. Nguyen [29] observed that differ-
skills. Sperling and Lickerman [39] adapted academic material into ent age groups have different perceptions of conversational agents.
a software engineering curriculum specialized in ML/AI to meet Younger participants (12–13y) mostly perceived the agent as “more
the needs of twelfth-grade high-school students. competent, trustworthy, and sociable,” while older students (14–15y)
A few scholars have worked with participatory approaches to based their expectations on prior experience with these devices and
define learning activities for AI/ML. For instance, Touretzky et al. expected them to be knowledgeable about many diverse topics and
[43] collaborated with AI experts to develop guidelines for learning skills that went beyond the activity.
about AI in formal education, organizing them into four grade spans

300
IDC ’23, June 19–23, 2023, Chicago, IL, USA Schaper et al.

A further strand of research has focused on exploring children’s chosen as a means to research not only what to teach teenagers
conceptualizations of intelligent technologies in order to guide the about intelligent technologies, but also to define how they might
design of learning experiences that relate to their understanding. learn about these technologies. Thus, we used the activities as “a
For instance, Druga et al. [13] investigated children’s imaginaries different way of knowing” [19], expanding the research space by
(7–12y) across four different countries (the United States, Germany, engaging specifically with the teenagers’ own understanding of
Denmark, and Sweden) around smart devices and toys of the future these technologies. In this study, we conducted a series of work-
and their understanding of current AI technologies. Further, Druga shops with eleven eighth-grade students aged 14–15 years (f=6;
et al. [12] in their 4As framework, promoted a critical understand- m=5) from a Danish secondary school located in an urban area of
ing of AI by proposing a new approach to AI literacy for designers the city of Aarhus with a majority of high socio-economic families.
and educators, with the goal of helping families to develop a critical Each workshop lasted for two sessions and involved three to four
understanding of AI systems and their use. The framework con- students. The students came from three different classes and were
sidered how differently children (5–11y) perceived AI bias from recruited in collaboration with their teachers. The participating
adults, and addressed the need for future tools and curricula related students signed up for the workshops on a voluntary basis. Half
to algorithmic justice to take this into account. Evangelista [17] of the students had a special interest in learning about technology,
conducted a short survey with high-school students (14–17y) to and all of them felt comfortable communicating in English with
explore their conceptions of ML. The insights gained were used the two primary researchers. Both the students and their parents
to formulate a proposal for an introductory workshop to learn were provided with a consent form informing them about and ask-
about ML and to discuss the ethical issues involved, such as data ing for their agreement on data collection and dissemination. We
collection and privacy, transparency, manipulation, or failures. In acknowledge that in this study we engaged only a small number
this regard, the authors emphasized the importance of addressing of participants in our workshop series. However, our goal was to
the similarities and differences between ML and human learning. use the findings of the study as stepping stones to define design
Schaper and Ruiz [36] suggested design recommendations for in- recommendations for developing learning activities and tools in
volving at-risk students (11-12y) in learning about AI in remote future interventions within our over-arching research project. The
contexts. Malinverni et al. [28] used the creation of stories and the outcomes will lead to iterative refinements of the design recom-
realization of audiovisual narratives to highlight differences in nu- mendations.
ances between the children’s (10–11y) and the teenagers’ (13–14y)
ways of constructing meaning for the notion “what a robot is and Procedure in Workshop Series The workshop series explored
what it can do”. Malinverni et al. also focused on the role of an- the students’ everyday experiences and understanding of AI/ML
thropomorphization of robots, on gender stereotypes, and on the technologies by introducing them to basic technical features and
relationship between robots and violence. These studies provide providing them with different ways of expression, e.g. verbal ex-
interesting insights that can help to guide children’s learning pro- pression through writing, drawings, and the creation of physical
cess about AI/ML technologies. But there is still very little research artifacts. The workshops took place in the school library and were
that investigates understandings of these intelligent technologies facilitated by the first two authors (f=2). The activities took place
among teenagers [17, 28, 29]. To address these challenges, our re- in two sessions and were organized as follows.
search focused on participatory approaches to developing novel ways The goals of session 1 were (I) to understand the imaginaries that
of teaching and learning about intelligent technologies through the the students attributed to intelligent technologies; (II) to co-explore
lens of Computational Empowerment. Our study engaged a group of students’ understanding through teaching them about the basic
Danish teenagers (14–15y) in exploring their everyday understanding technology fundamentals of intelligent technologies; (III) to under-
of AI/ML technologies. Our principal goal was to use the insights stand how they imagined the role of technology in their future lives.
gained as a starting point to develop the students’ understandings The session lasted 90 minutes in total. Session 1 was structured into
of AI/ML technologies as they expressed them during the learning three main parts, beginning with a semi-structured interview with
activities. the students (10 minutes) about their general ideas related to the
concept of “intelligent technologies.” Here, we prompted students
with questions like Which kinds of technologies do you use? What
3 METHOD do you have at home or at school? We then showed the students
This study forms part of a larger research project called Computa- pictures of various different technologies and asked them to guess
tional Empowerment in Emerging Technology Education (CEED), if they represented an intelligent device/application, encouraging
which focuses on investigating how to develop teaching activities them to explain the reasoning behind their answers. In the second
about emerging technologies for K–12 education. Using a partici- part of the session, we instructed the students in the activity AI
patory design approach, working with teachers and students, we Bingo [31] (40 min) to start a conversation about what kind of AI
have developed novel ways of teaching about emerging technolo- systems they use in their everyday lives. We aimed to provide them
gies. The project is a five-year ongoing collaboration involving with basic technical vocabulary about intelligent technologies, with
32 pioneer teachers and 52 schools in Aarhus municipalities in the aim of supporting their ability to describe their technical fea-
Denmark. The smaller study presented in this paper feeds into the tures. To do this, we explained to them the difference between the
larger research project by working with teenagers through the use concepts of a data set, a learning algorithm, and prediction. We
of participatory design techniques in the context of workshops. To then shared with them a sheet of A3 paper that showed a bingo
this end, the participatory techniques employed in this study were table with various concrete examples of data sets, algorithms, and

301
Five Design Recommendations for Teaching Teenagers’ about Artificial Intelligence and Machine Learning IDC ’23, June 19–23, 2023, Chicago, IL, USA

predictions. The students were asked to identify the prediction the through drawing, through sorting images). In this regard, we as-
system was trying to make, along with the dataset it might use sumed that each activity might prompt a different range of aspects
to make that prediction. The final activity was a hands-on design of their understanding of intelligent technologies. Thematic anal-
process about how the students might imagine future technologies ysis was used to interpret the qualitative data, as described by
(40 minutes), with the aim of understanding which potential oppor- Braun and Clarke [7] . First, to familiarize themselves with the
tunities and risks the students associated with AI/ML. We, therefore, data, the authors transcribed the video recordings, read through
provided them with low-tech crafting materials. The students were the transcripts, and noted down initial ideas. Second, the first and
instructed to first brainstorm and document their ideas through second authors discussed the initial ideas and then coded all the
drawings, then to create a rapid prototype. The students worked in transcripts by applying one or more inductive conceptual categories
pairs. At the end of the session, each group briefly presented their (e.g. student knowledge, misconceptions, fears and hopes) to the
idea to the researchers. data. To distinguish parts of the data, one or a few sentences were
The goals of session 2 were (I) to investigate how students con- coded in relation to the inductive categories. The coding process
ceptualize the notion of “intelligent technologies”; (II) to deepen was iterative and in ongoing dialog to the point where a common
the students’ understanding of intelligent technologies by explor- agreement was defined on the categories and the corresponding
ing a web-based ML tool; (III) to investigate their fears and hopes examples. Third, the first and second authors collated categories
about the impact of intelligent technologies in their everyday lives. into potential themes. Based on the research focus of how AI/ML
The session lasted for 80 minutes. First, we showed the students needs to be taught from the teenagers’ own perspectives, we induc-
printouts of pictures of various different technologies. The pur- tively identified ten individual codes, which we combined into four
pose here was to establish which of these technologies (including core themes as follows: (1) Teenagers’ everyday interaction with
social robots, smart toys, and voice assistants) they had been in intelligent technologies; (2) Teenagers’ conceptualization of the
contact with in their everyday lives. We asked them to arrange technical features of intelligent technologies; (3) Teenagers’ future
the pictures on a horizontal scale from less (left) to most (right) imaginaries about the use of intelligent technologies; (4) Teenagers’
intelligent (Figure 1, left). and when they finished, we asked them awareness of ethical considerations and societal implications. These
to explain the reasoning behind their choices. The activity took themes were documented in a shared codebook. We synchronized
20 minutes. Second, we introduced the students to the Teachable the extracted verbal expressions into one document and elaborated
Machine tool [21], a web application that allows the creation of a summary. Subsequently, we developed a conceptual map (Figure
ML models, specifically image classifiers, without coding (40 min- 2) which represents an overview of our findings.
utes). We encouraged the students, in pairs, to explore the different
functionalities of the tool, using objects, drawings, or gestures as
input for the learning algorithms (Figure 1, middle). Each pair had 4 RESULTS
a laptop which they could use to run the software. At the end of The thematic analysis identified four core themes in the teenagers’
the activity, each pair explained what strategies they had used and everyday experiences with intelligent technologies. In this section,
how the tool worked in relation to deciding on the input, training we first present an overview of the conceptual map that we devel-
the machine learning model, and checking the model’s prediction. oped based on our analysis. We then discuss each individual theme
Third, we asked the students to imagine a future case scenario in and how it is supported by our data. The conceptual map illustrates
which birthday parties would be mediated by technology. We then the details that follow. The boxes are the core themes and the sub-
asked each of the pairs to produce one fictional storyboard (20 lists in each box are the individual themes. We used these concepts,
minutes) to represent their idea (Figure 1, right). The first pair was based on the teenagers’ verbal expressions, to identify links and
instructed to describe a best-case scenario of the use of intelligent relationships between the themes. In our data, the students talked
technologies in everyday life, and the second pair focused on an about their everyday interaction (T1) with AI/ML technologies.
alternative worst-case scenario. Finally, each pair was encouraged We observed that topics close to teenagers’ everyday experiences
to explain their drawing. shaped their future imaginaries (T3) towards positive technology
views, and that these topics seem to support their learning about
Data Collection and Analysis Process We employed a the- ethical and societal impacts (T4). Specifically, the teenagers cre-
matic analysis to identify how the teenagers conceptualized the ated future imaginaries about AI/ML technologies that influenced
AI/ML technologies. Specifically, we wanted to capture their un- their understanding of societal risks. The teenagers’ conceptualiza-
derstanding of their everyday interaction with these technologies, tion (T2) of AI/ML technologies was based on two inputs – their
their awareness of related societal and ethical implications, and everyday interactions with these technologies, and the hands-on
their future imaginaries. To this end, we analyzed the students’ ver- activities that we employed during the study, which helped them
bal expressions during the following activities: (I) Semi-Structured to better understand the nuances of technical features. The data
Interviews; (II) Rapid-Prototyping Future Devices; (III) Intelligence also showed that a lack of knowledge about the technical features
Rating; (IV) Fictional Storyboards carried out during three differ- of these technologies influenced the teenagers’ understanding of
ent workshops with eleven (age=14–15y; f=6; m=5) teenagers. We negative views of AI/ML technologies and their notion of intelli-
decided to focus on this selection of activities because each one gence. In what follows, we discuss each core theme and subtheme
allowed the students to engage with the topics by expressing their of the conceptual map in detail, with examples from our workshop
thoughts in various different ways (e.g. through verbal expressions, sessions.

302
IDC ’23, June 19–23, 2023, Chicago, IL, USA Schaper et al.

Figure 2: A conceptual map about students’ understanding of AI/ML technologies


T1) Everyday interaction with intelligent technologies: set up by educational collaborators (e.g. FabLabs, universities, com-
During the workshops, we aimed to understand teenagers’ everyday panies). These technologies are too expensive to buy for private use.
interactions with intelligent technologies. This implied considering At the beginning of our workshop, one student mentioned her con-
the role of intelligent technologies in teenagers’ daily lives, their cern that using these technologies would require specific training.
motivation to use these technologies, and their accessibility. Role However, after the Teachable Machine activity, a web-based tool to
in daily lives: The findings indicated that all students had their create machine learning models, most of the students reported pos-
own smartphone and computer at home, which they used to do itively on their experience. One of the girls stressed: “I don’t know
their homework and to play games. One female student explained super much about computers, and I was very glad that you didn’t
that she had an educational robot at home, but the family barely make it really hard to understand because I was a little afraid that
used it. Most of the teenagers frequently used Google services and you would just talk about how to program a game or something.”
social media platforms (e.g. Facebook, Twitter, Instagram, Snapchat, Another boy said: “I personally now discovered this new program
TikTok). Further, they mentioned that they used social media plat- (...) I found that it was easy to set up, and how fast it was. I thought
forms in different ways. One student explained: “I think Instagram it would be something that would take ages, but it was not.” One
is more serious. It’s more like selfies and coffees and stuff like that. of the male students asked for further information to build on the
Well, TikTok, it’s for jokes. We just post like dances and random learning activities featured in the workshop. Specifically, he was in-
stuff.” Two male students reported that they were “not on social terested in exploring the software Adacraft, a program that would
media” because they were conscious of the risk of misuse of their recognize gestures and allow for combining them with Scratch-
private data. Some students were not familiar with the use of voice based programming blocks. We had briefly mentioned this program
assistants (e.g. Amazon Alexa) and smart toys (e.g. Hello Barbie). during our discussions with the students. Our findings suggest that
Motivation to use: The teenagers’ motivation for using social intelligent technologies, particularly social media platforms, play an
media platforms was driven by the desire to obtain information important role in the teenagers’ everyday lives. However, there are
about others (e.g. in Spotify, another peer’s taste in music). One stu- technologies that they use less frequently (e.g. voice assistants, Hello
dent stated that he would take a part in an online survey and provide Barbie). The students reported differing preferences for engaging in
his private data if he received a financial reward. Two other female learning about these technologies, which spanned a wide range from
students highlighted that they particularly enjoyed creative activi- creative activities to in-depth coding activities. Interestingly, two of
ties and tools that allowed them to create their own ideas. Three the male students who were engaged in coding practices had made a
male students stressed that they were very interested in coding, conscious decision not to use social media, because they did not want
and frequently explored different programming languages/tools other people to access their private data or have personal information
such as Scratch, Unity and Python. about them. Our attempt to engage the less experienced students with
Accessibility for learning: The students highlighted that some an easily accessible “off-the-shelf” tool such as the Teachable Machine
technologies are often only accessible in school or in installations triggered their interest and gave them self-confidence in their ability

303
Five Design Recommendations for Teaching Teenagers’ about Artificial Intelligence and Machine Learning IDC ’23, June 19–23, 2023, Chicago, IL, USA

to learn more about intelligent technologies. “decode” than those that they used in their everyday lives. Specifi-
cally, social media platforms have many hidden functions that were
T2) Conceptualization of technical features: One of our not obvious to the students.
goals for the workshops was to teach the teenagers about the tech- Knowledge about technical features: Three students had a
nical features of intelligent technologies so as to convey an under- very deep understanding of the technical features of the technolo-
standing of “what these technologies can do.” We focused in this gies presented. Once again, these were those with a special interest
regard on the notion of the intelligence of technological devices and in programming. They were aware of the existence of algorithms in
applications, on the students’ knowledge about technical features, social media platforms. In this regard, one student stated: “YouTube
and on their misconceptions about the functionality of intelligent and Google (...) have an algorithm that knows the things you search
technologies. Notion of Intelligence: During the rating activity, for, like if you search for (...) or watch something on YouTube, it
we observed three major trends related to the students’ notion of in- chooses some videos to pop up on your front page that are recom-
telligence. First, the students rated technologies as more intelligent mended for you.” Another student in the same group described the
according to the number of their components, because “we need functioning of intelligent AR technologies as follows: “Snapchat
more AI to be able to control all the components. (...) The hospital and Pokemon (...) both use GPS and (...) put an overlay on your
robot (...) has a lot of these features, it has the screen, it has voice screen/your camera.” They also used terms that showed some ad-
recognition and I think it has face recognition and it can recognize ditional knowledge: for instance, one student stressed that “the
walls.” In comparison, they regarded as less intelligent those tech- micro:bit is not as powerful as a Raspberry Pie.” Before engaging
nologies that were designed for small children with simple features, in the hands-on activities, the other students showed a very basic
as well as devices with simple built-in input and output systems understanding of what intelligent technologies can do (e.g. robots
(e.g. a 3D printer, Microbits, robotic vacuum cleaners). For instance, can measure blood pressure). After the hands-on activities, they
one student commented about a 3D printer: “I think it’s probably expressed a more nuanced understanding of the functionalities of
not as smart as you think, because when you 3D print something technical features. For instance, after the Teachable Machine ac-
(...) you already made a 3D model. So it’s actually just like a printer, tivity, several students were able to troubleshoot technical issues
but more advanced.” and improve their dataset for the model. One student said: “It could
Second, a further criterion for intelligence seemed to be if the recognize my hand, my face, or both. But it didn’t really work, be-
technology had human-like capabilities, e.g. if it used voice recogni- cause I had two different backgrounds.” Similarly, three girls looked
tion and was able to control other devices. For instance, one group for a smooth surface in the room in order to reproduce the same
considered Siri and Alexa as the most advanced. In this regard, background to train their data set.
one student mentioned: “They can both recognize your voice, and Misconceptions about technical features: During the Case
they can turn on lights in your house. I also think they can turn on Scenario Drawing, several students showed misconceptions related
the oven. (...) And they have access to the entire internet, you can to the understanding of which technologies use ML. One student
say nearly anything to them and they will have an answer.” At the assumed that every robot would use ML. He specifically described
same time, some students were not familiar with voice assistants it as “something that has mechanics in it.” One girl concluded that
(e.g. Amazon Alexa) or smart toys (e.g. Hello Barbie). Therefore, devices connected to a computer (e.g. a surveillance camera) also
in the first round of the Intelligent Rating Activity, the students use the internet, and consequently ML algorithms. Three students
placed these two technologies in the category of less intelligent proposed a “CO2 converter” that could separate atoms by AI. They
technologies. After we had explained their technical features, they explained that this would take away the need for humans “to do
corrected their choices and placed “Hello Barbie” in the middle anything”. In other words, they reduced the intelligent function
of the rating meter and Amazon Alexa on the side of the highly of the technology to automation processes. Our findings suggest
intelligent technologies. that the students ranked the intelligent technologies according to the
Third, some students ranked Google and YouTube as less intelli- number of their components and the existence of human-like features.
gent than Siri, because they thought that these platforms could only Moreover, students with prior technology interests seemed to have a
offer searching for content by text input, allowing communication deeper knowledge of the technical features of the technologies that we
with others, and playing video content. Other students described discussed with them. The other students were quickly able to grasp
Pokemon Go, Snapchat and YouTube as “not so complicated” since specific aspects of AI through the proposed hands-on activity, but they
they did not use an algorithm to process data. Other students even presented general misconceptions about the difference between intelli-
perceived microcontrollers as more intelligent than social media gent and non-intelligent technologies. Intelligent technologies were
platforms. For instance, one student stated: “We have a micro:bit associated with the existence of mechanical components, computer
and this Mindstorm (...) They’re both pretty similar (...) they can software, and automatization processes.
both be coded and the reason why it’s higher than Google, I would
say is because it needs to recognize what input you’re using, and it T3) Future Imaginaries: We also investigated the teenagers’ fu-
can.” This example shows that the students’ notion of intelligence ture imaginaries about the use of intelligent technologies. To do
related to technologies can be determined both by their everyday this we focused specifically on the utopian and dystopian views of
experiences and by their knowledge of their functioning. Particu- technology that the students expressed, and the hopes and fears
larly, technologies that they had used in maker activities in school that were embedded in these scenarios.
(e.g. microcontrollers, 3D printers) seemed to be easier for them to Utopian View: Two groups presented a utopian view of tech-
nology when they explained their best-case scenario using the

304
IDC ’23, June 19–23, 2023, Chicago, IL, USA Schaper et al.

Figure 3: Case Scenario Drawing activity: Storyboard portraying the best-and worst-case scenarios of the use of intelligent
technologies.

storyboard. The first example substituted a human interviewer expressed several fears about how intelligent technologies could
with a robot. The student made this choice with the idea that robots impact their own or others’ lives. These fears were mainly related
should be installed in as many places as possible and that the num- to the lack of privacy and the misuse of data. For instance, one stu-
ber of survey participants should be increased accordingly. The dent explained: “I think it’s kind of scary that Snapchat and other
second group described how intelligent technologies could be used devices can see where you are at all times if you have them on you.
to prevent climate change. As previously described, the group pro- Sometimes you think, okay I’m alone now, but you really aren’t. ”
posed a CO2 converter into oxygen. The students explained: “All Another student said: “It’s very scary that Siri is listening to you
the polar bears are now saved and they don’t die of drowning (...) when you’re not saying Siri. And I heard that the government is
because it can split the particles from each other by AI, so we don’t recording from Siri. Like if in a situation that happens, and they
need to do anything.” Both examples illustrate how intelligent tech- have to record what you’re saying, they’re using Siri to record.”
nologies can be used to undertake tasks that usually imply human These comments illustrated that the students had the feeling of
labor. Dystopian View: During the case scenario drawing activity, being surveilled all the time. It also caused them anxiety that they
two female students proposed a story about hackers who got ac- had no control over who could “listen to them” and access their pri-
cess to the security cameras of a house with the intent to kidnap vate data. Other students mentioned specific technical deficiencies
someone. They said: “The drawing shows a security camera which of intelligent technologies that would make their use unsecured.
was hacked by some guy, it could be like a kidnapper (...) then he For instance, one student explained: “If I had facial recognition
can track this girl and find out where she is (...) her exact position on my phone, and I had glasses then if another person took them
through the security cameras (...) and then kidnap her.” Another on, they would probably not come into my phone, but it would
group described a story inspired by the case of Edward Snowden put the percentage way higher than if he didn’t have my glasses
(Figure 3, right): “I saw a movie about a CIA agent who discov- on.” Another fear among the students was that someone might be
ered that the people in the CIA could technically watch people accidentally identified as a murderer because of misinterpretations
through the cameras. And here we have an example of a random of facial-recognition systems. Our findings show that the teenagers
girl watching YouTube and a boy watching TV. Then we have this expressed utopian views and hopes about how intelligent technologies
guy watching through the webcam to see what they’re doing and could help to detect criminals, connect people, and improve the effi-
he tells it to somebody else, and gives some personal information ciency of technical processes. In contrast, the students’ dystopian views
to other people.” Both examples point to the students’ dystopian and fears of intelligent technologies were related to public surveillance,
views of how intelligent technologies could be misused to obtain lack of privacy, and misuse of the technology.
sensitive data and information about others.
Hopes and Fears: Students’ hopes of intelligent technologies T4) Awareness of ethical and societal implications: Another
were connected to the potential opportunities that they saw in them. goal of the workshops was to promote teenagers’ awareness of
For instance, they hoped that authorities such as the police could the ethical considerations and societal implications of intelligent
use people’s private data to track down criminals (Figure 3, left). technologies. We focused specifically on the risks and opportunities
At the same time, they believed that robots and intelligent tech- of how intelligent technologies can impact society. We also doc-
nologies could substitute and perform human tasks, which would umented whether the students discussed conflicts of interest and
make processes faster and more efficient. In contrast, the students ethical dilemmas provoked by one specific intelligent technology.

305
Five Design Recommendations for Teaching Teenagers’ about Artificial Intelligence and Machine Learning IDC ’23, June 19–23, 2023, Chicago, IL, USA

Risks of intelligent technologies: The students mentioned sev- the use of private data from smart speakers – and on the teenagers’
eral negative views of how intelligent technologies might negatively beliefs on how their own data should be used. In another group,
impact society. One theme was the collection of private data. For the students described a story in which users received financial
instance, one student highlighted: “Siri and Alexa are collecting the remuneration for giving away their private data. The student ex-
voice and everything you say. Record everything, and when you’re plained: “It is about a guy and he just walks somewhere. And he
carrying your phone it has Siri on it.” Another student pointed out sees this ATM where it says free cash. And then an intelligent robot
risks in the social media tool Snapchat: “If the company wants to, it appears which says ‘Hello, I am RG1O2 and I am going to give you
can get most of the people on Snapchat, they have their birthdays free cash. First, you need to answer my questions.’ So it is like a
on there, their personal emails. And some people even use the same robot asking you a question and then you have to answer it. For
password on most of their things, most of their accounts. So if they example, a company that made a quiz or something doing research
wanted to, they could try to get into maybe a Google account, an and then they give out free cash.” This story shows the potential
email. And then maybe they could see a lot of personal information conflict of interest between technology users and other stakehold-
on that.” One student stated that he decided not to use any social ers. Further discussion of both examples revealed that the students
media platforms because he was aware of the implications for his were not conscious of the dilemmas and that deeper engagement
private data. In contrast, another student stated that he was aware would be required to support their capacity to critically reflect upon
of the possible loss of privacy in relation to social media, but that the conflict of interests between stakeholders. Our findings show
he would not go so far as to change his behavior. He explained: “I that the students were capable of articulating both potential risks and
don’t like being recorded and all of that. But I won’t give up social opportunities for intelligent technologies. They also showed signs of
media because of it. (...) I already think that they know so much ability to critically reflect upon the ethical and societal implications of
about me, so it doesn’t make any sense to stop.” This example may those intelligent technologies they were familiar with. In contrast, we
point toward teenagers’ feelings of “being powerless” in a digitized observed that the students were not able to detect ethical dilemmas.
society and their lack of tools and abilities to protect their own Some students also seemed to lack strategies or tools to protect their
private data from misuse. In addition, two students changed their private data against technology companies.
perspective on the use of “Hello Barbie” when they were informed
about its features. After the explanation, they stated that they found
it “scary” that their private data was recorded and that they would
never consider playing with it. These examples highlight that the 5 DISCUSSION
students were conscious of the ethical and societal implications This study aimed to investigate how we can scaffold teenagers’
related to social media platforms and voice assistants, but to a lesser learning about AI/ML by building on their everyday understanding
extent with regard to devices they were not familiar with (e.g. smart of these technologies. Consistent with prior research, our study
toys). Several students also showed critical reflections on the im- confirms that teenagers build their understanding of AI/ML on their
plications for themselves, as well as reluctance to use devices that existing knowledge of technical features and the implications for
would record their personal data. people [36]. We also observed a tendency to understand intelligent
Opportunities for intelligent technologies: Several students technologies through an anthropomorphic perspective [28, 33, 51],
highlighted positive opportunities or ways in which intelligent and that the students’ fears and hopes related were dependent on
technologies could impact society. During the prototyping activity, situated contexts [33]. Our study goes beyond prior work by iden-
several students presented ideas for intelligent vehicles and devices tifying relationships in the conceptual map and illustrating how an
(e.g. drones, flying cars) and stressed how products and people interplay of these aspects influenced the teenagers’ understanding
could be transported faster with those. Further, in the fictional sto- of AI/ML technologies (Figure 2). In addition, we propose five de-
ryboard, one group proposed an AI system that could contribute sign recommendations for learning activities and tools for AI/ML
to preventing climate change. Another group stressed that voice technologies. These recommendations were developed based on
recordings from Alexa could be used to prevent domestic violence. the themes described in the results. These recommendations are
Other students mentioned the potential of intelligent technologies intended to support teenagers’ critical reflection about technical
to connect people. During the prototyping activity, they designed features and their societal implications, and to empower them in
an intelligent pin that could be used at parties to socialize with becoming informed and responsible users in a digitized society [24].
other teenagers with the same interests. In this regard, a student Our findings point towards the need for educational strategies that
explained: “It really depends on your situation, or what you want embrace (1) Diversity in learning paths about intelligent technologies
to do with this information. Because when I date the person then (2) Human intelligence as an entry point for artificial intelligence; (3)
sure, it’s good to know what their favorite music is, yeah, but if you Accessible technology as an opportunity for sustainable engagement;
like to tag them, it could also be useful.” This example highlights (4) Empowerment of action-taking in a digitalized society; and (5)
how the provision of personal information and the scope for con- Reflection upon real-world dilemmas of intelligent technologies as
necting people offered by social media platforms are very relevant wicked problems. We focus our discussion on these five design
to teenagers’ everyday lives. recommendations for education activities and possible future direc-
Awareness of Dilemmas: During the Fictional Storyboard activity, tions. To illustrate how the recommendations unfolded through the
one group proposed a story in which Alexa “witnessed” a crime thematic analysis, we also refer in the descriptions to the theme
and the police could use the recording as evidence. This example numbers (T1–4) and to the subtheme titles.
touched on a real-world dilemma – the legal and ethical debate on 1) Diversity in learning paths about intelligent technologies:

306
IDC ’23, June 19–23, 2023, Chicago, IL, USA Schaper et al.

The first design recommendation focuses on the need to provide stu- 3) Accessible technology as an opportunity for sustain-
dents with a spectrum of different types of activities, ranging from able engagement: The third design recommendation concerns
creative exploration to concrete exercises intended to foster learn- the need to provide learning materials that are sustainable over
ing about the technical features of AI/ML. Our findings revealed time and that can promote a long-term engagement of the students
that whereas several students were keen on coding and exploring with these technologies. In our study, the students reported that
different programming languages and digital, other teenagers par- the technologies used in school, such as projects with collaborat-
ticularly enjoyed creative activities and tools that allowed them ing partners (e.g. FabLab maker spaces, the university, companies,
to create their own ideas (T1: Motivation to use). In this regard, etc.), were not accessible to them outside the school context (T1:
Wan et al. [49] called for more learning opportunities about ML for Accessibilityfor learning ). Much previous work has discussed the
students with diverse STEM skills. Similarly, Smith at al. [38] recom- need for sustainable practices to maintain and scale [18, 23, 37]
mended that scholars and educational practitioners should explore teaching approaches through the creation of suitable resources and
new educational strategies to engage students who initially seem infrastructures. We believe that the sustainability of student engage-
less motivated or interested in learning about digital technology ment needs to go beyond the classroom setting. Learning activities
and its implications, suggesting that such students might quickly need to be engaging for students just as they are, i.e. they need to
become engaged. Interestingly, prior work often seems to focus motivate them to explore AI/ML technologies. One possible way
on exploring new educational strategies in diverse contexts, for of addressing this challenge could be to use “off-the-shelf,” open-
example as part of after-school activities or involving students who ended technologies, i.e. tools that are easily available to students
are interested in technology per se (e.g. [45, 54]). We would argue, in out-of-the-classroom contexts and that support the exploration
however, that rather than concentrating on technology-interested of different technological features in multiple ways. For instance,
students, future research needs to explore inclusive approaches in during our study, we introduced the Teachable Machine tool and
education that embrace a broader spectrum of students’ learning gave students the freedom to use any kind of material to test out
paths in formal education. image-recognition features (T2: Knowledge about technical features).
After some time, the students started exploring the tool with the
2) Human intelligence as an entry point for artificial intel- materials they had in hand. The students’ exploration turned into
ligence: The second design recommendation focuses on the need to a lively music performance, using candies, faces, and gestures as
introduce the notion of intelligence from a humanistic perspective, inputs and sounds as outputs. At the end of the workshop, several
e.g. to discuss different types of intelligence and their differences, students reported that using the digital tool was easier than they
before introducing the technical features. In our study, we observed had expected and asked for additional information about using
that the teenagers rated as highly intelligent those technologies “off-the-shelf” technologies in their spare time. This example shows
with human-like abilities of speech, voice recognition, and the abil- how open-ended activity with free accessible tools (e.g. Teachable
ity to control other devices (T2: Notion of intelligence). However, Machine, Dance with AI, Adacraft, Scratch Lab) may allow students
even though human-like aspects can sometimes be a proxy for to engage with intelligent technologies in their own ways and mo-
intelligence (e.g.: humanoid robots like Nadine and Sophia), the tivate them for further use.
correlation is not reliable in all cases, and this may be confusing to
some students. Therefore, using a more familiar concept (human 4) Empowerment of action-taking in a digitalized world:
intelligence) and comparing it to a more unfamiliar one (artificial The fourth design recommendation focuses on the need to support
intelligence) may be a promising strategy for avoiding student teenagers’ ability to take action against the misuse of their personal
misconceptions regarding intelligent technologies. For some other data. In our study, most of the students were aware that social
students, the notion of intelligence was also associated with a larger media platforms collected and stored their personal data. In this
number of features and higher complexity of the devices. These regard, they expressed fears about privacy issues, misuse of data,
results indicate that the teenagers focused on technical features and the technical deficiencies of intelligent technologies (T4: hopes
when ranking the degree of intelligence of the various technologies. and fears). One student in particular stressed being “powerless”
This finding is consistent with approaches in previous work aiming against those companies which hold on to his data (T3: risks of in-
to investigate children’s understanding of intelligent technologies telligent technologies). In this regard, the agenda of Computational
[28, 33, 51]. Further, Wan et al. [49] stressed how a basic under- Empowerment foresees the development of students’ digital design
standing of how ML works may make teenagers more attentive to literacy to promote active citizenship for all children [24]. This goal
the ethical aspects of AI. Other scholars have stressed that students not only addresses providing students with knowledge about intel-
particularly need to learn about the limitations of ML algorithms ligent technologies, but also addresses providing them with tools to
in comparison to human learning [16]. Despite the relevance of take responsible action in a digitized society. For instance, Yu et al.
such prior work, many research projects still primarily focus on the [54] proposed civic programs to engage young people in cases of
technical features of intelligent technologies [47]. Future directions socio-political injustice. In addition, several scholars have reported
need to have a stronger focus on human-centered approaches to on legal and conceptual approaches to protect privacy in a data-
intelligent technologies. Inspirational examples can be found in driven society [22]. These, however, mainly target adults, which
learning activities and tools related to ethical dilemmas and tech- points to a research gap in age-appropriate tools for teenagers and
nologies [6, 11, 35]. in enabling them to protect their individual rights as users and
consumers of intelligent technologies.

307
Five Design Recommendations for Teaching Teenagers’ about Artificial Intelligence and Machine Learning IDC ’23, June 19–23, 2023, Chicago, IL, USA

5) Reflection on real-world dilemmas of intelligent tech- agreement on data collection and dissemination. The participating
nologies as wicked problems: The fifth design recommendation students signed up for the workshops on a voluntary basis.
focuses on the need to support teenagers’ reflections upon the real-
world dilemmas of intelligent technologies as conflicts of interest ACKNOWLEDGMENTS
between stakeholders. In this regard, we specifically focus on the This work is part of a five-year research project called CEED (Com-
concept of wicked problems, i.e. conflicts of interest with many putational Empowerment for Emerging Technologies in Education),
interdependent factors which probably make them unsolvable. Our exploring and building new practices of computational empower-
findings showed that the students were not conscious of the poten- ment for emerging technologies in Danish secondary education
tial dilemmas inherent in what they presented in their storyboards through a cross-disciplinary approach between computer science,
– for instance, of the legal and individual rights of data collected humanities, and engineering. The CEED project is funded by the
by voice assistants (T3: awareness of dilemmas). Previous research Villum Foundation under grant number 28831.
has presented the use of activities and tools to support a critical
reflection upon conflict of interests between different stakeholders, REFERENCES
e.g. cards [4, 6, 36], cultural traditions as prompts [35] and matrix [1] 2021. Digital Education Action Plan – Action 8 | European Education Area. https:
exercises [11] to support a critical reflection upon conflict of inter- //education.ec.europa.eu/focus-topics/digital-education/action-plan/action-8
[2] Ashraf Abdul, Jo Vermeulen, Danding Wang, Brian Y Lim, and Mohan Kankan-
ests between different stakeholders. However, previous work has halli. 2018. Trends and trajectories for explainable, accountable and intelligible
shown challenges in addressing teenagers’ knowledge gaps about systems: An hci research agenda. In Proceedings of the 2018 CHI conference on
which are the potential technology stakeholders [36] and nuances human factors in computing systems. 1–18.
[3] Saira Anwar, Nicholas Alexander Bascou, Muhsin Menekse, and Asefeh Kardgar.
in ethical decision-making (i.e. who will benefit or be harmed by 2019. A systematic review of studies on educational robotics. Journal of Pre-
the system) [11]. Future research may investigate suitable examples College Engineering Education Research (J-PEER) 9, 2 (2019), 2.
[4] Stephanie Ballard, Karen M. Chappell, and Kristen Kennedy. 2019. Judgment
of dilemmas that teenagers find easy to understand and are close Call the Game: Using Value Sensitive Design and Design Fiction to Surface
to their everyday experiences with intelligent technologies. Ethical Concerns Related to Technology. In Proceedings of the 2019 on Designing
Interactive Systems Conference (San Diego, CA, USA) (DIS ’19). Association for
Computing Machinery, New York, NY, USA, 421–433. https://doi.org/10.1145/
Summarizing, we argue that implementing these five design 3322276.3323697
recommendations in conjunction can lead to the unfolding of the [5] Karl-Emil Kjær Bilstrup, Magnus Høholt Kaspersen, Mille Skovhus Lund-
full potential of educational activities and tools that can support ing, Marie-Monique Schaper, Maarten van Mechelen, Mariana Aki Tamashiro,
Rachel Charlotte Smith, Ole Sejer Iversen, Information Studies, and Digital Design.
teenagers’ learning experiences of AI/ML technologies in formal 2022. Supporting Critical Data Literacy in K-9 Education : Three Principles for
education. Specifically, we believe that the design recommenda- Enriching Pupils ’ Relationship to Data. In Proceedings of the 2022 ACM Interaction
Design and Children Conference. Association for Computing Machinery.
tions can open the path for Computational Empowerment [24] of [6] Karl-Emil Kjær Bilstrup, Magnus H. Kaspersen, and Marianne Graves Petersen.
teenagers by building up their understanding of intelligent tech- 2020. Staging Reflections on Ethical Dilemmas in Machine Learning. In Proceed-
nologies and of their own agency in a digitized society. Together, ings of the 2020 ACM Designing Interactive Systems Conference. ACM, New York,
NY, USA, 1211–1222. https://doi.org/10.1145/3357236.3395558
the recommendations serve as a teenage-centered starting point [7] Virginia Braun and Victoria Clarke. 2006. Using thematic analysis in psychology.
for engaging young people in learning about intelligent technolo- Qualitative Research in Psychology 3, 2 (jan 2006), 77–101. https://doi.org/10.
gies. However, it is important to acknowledge that our study was 1191/1478088706qp063oa
[8] Vicky Charisi, Laura Malinverni, Elisa Rubegni, and Marie-Monique Schaper.
conducted in a school context involving families from high socio- 2020. Empowering Children’s Critical Reflections on AI, Robotics and Other
economic backgrounds. Further, we looked only at the specific age Intelligent Technologies. In Proceedings of the 11th Nordic Conference on Human-
Computer Interaction: Shaping Experiences, Shaping Society. ACM, New York, NY,
range of teenagers aged between 14 and 15. In future work, we will USA, 1–4. https://doi.org/10.1145/3419249.3420090
employ, iterate and refine these design recommendations in the con- [9] Beverly Clarke. 2020. Exploring CS: Alternate AI Curriculum. Technical Report.
text of multiple research interventions. We encourage researchers [10] Jan Cuny. 2012. Transforming High School Computing: A Call to Action. ACM
Inroads 3, 2 (June 2012), 32–36. https://doi.org/10.1145/2189835.2189848
in the IDC community to explore these design recommendations [11] Daniella DiPaola, Blakeley H Payne, and Cynthia Breazeal. 2020. Decoding design
to guide the development of learning activities for teenagers in agendas: an ethical design activity for middle school students. In Proceedings of
diverse socio-economic contexts, with different age ranges, and the Interaction Design and Children Conference. 1–10.
[12] Stefania Druga, Fee Lia Christoph, and Amy J. Ko. 2022. Family as a Third
across cultures. Space for AI Literacies: How do children and parents learn about AI together?
Conference on Human Factors in Computing Systems - Proceedings (2022). https:
//doi.org/10.1145/3491102.3502031
6 SELECTION AND PARTICIPATION OF [13] Stefania Druga, Sarah T Vu, Eesh Likhith, and Tammy Qiu. 2019. Inclusive AI
literacy for kids around the world. In Proceedings of FabLearn 2019. 104–111.
CHILDREN [14] Stefania Druga, Randi Williams, Cynthia Breazeal, and Mitchel Resnick. 2017. "
Hey Google is it ok if I eat you?" Initial explorations in child-agent interaction.
This study forms part of a larger research project which is a part of In Proceedings of the 2017 conference on interaction design and children. 595–600.
a five-year ongoing collaboration involving 32 pioneer teachers and [15] Stefania Druga, Randi Williams, Hae Won Park, and Cynthia Breazeal. 2018. How
52 schools in Denmark. Through workshop sessions with school smart are the smart toys? Children and parents’ agent interaction and intelligence
attribution. In Proceedings of the 17th ACM Conference on Interaction Design and
teachers and students, we developed teaching activities with the Children. 231–240.
aim of educating teenagers about emerging technologies from a [16] Ignacio Evangelista, German Blesio, and Emanuel Benatti. 2019. Why are we not
teaching machine learning at high school? A proposal. 2018 World Engineering
teenage-centered perspective as well as from the perspectives of Education Forum - Global Engineering Deans Council, WEEF-GEDC 2018 (2019).
their teachers. The students in this study came from three different https://doi.org/10.1109/WEEF-GEDC.2018.8629750
classes of Danish public schools and were recruited in collaboration [17] Ignacio Evangelista, German Blesio, and Emanuel Benatti. 2019. Why are we not
teaching machine learning at high school? A proposal. 2018 World Engineering
with their teachers. Both students and their parents were provided Education Forum - Global Engineering Deans Council, WEEF-GEDC 2018 (2019).
with a consent form informing them about and asking for their https://doi.org/10.1109/WEEF-GEDC.2018.8629750

308
IDC ’23, June 19–23, 2023, Chicago, IL, USA Schaper et al.

[18] Barry J Fishman, William R. Penuel, Stephen Hegedus, and Jeremy Roschelle. [41] Mariana A Tamashiro, Maarten Van Mechelen, Marie-Monique Schaper, and
2011. What happens when the research ends? Factors related to the sustainabil- Ole Sejer Iversen. 2021. Introducing Teenagers to Machine Learning through
ity of a technology-infused mathematics curriculum. Journal of Computers in Design Fiction: An Exploratory Case Study. In Proceedings of the 20th ACM
Mathematics and Science Teaching 30 (2011), 329–353. Conference on Interaction Design and Children (IDC 2021). ACM, New York, NY,
[19] Christopher Frauenberger, Judith Good, Geraldine Fitzpatrick, and Ole Sejer USA.
Iversen. 2015. In pursuit of rigour and accountability in participatory design. [42] Matti Tedre, Tapani Toivonen, Juho Kahila, Henriikka Vartiainen, Teemu Val-
International journal of human-computer studies 74 (2015), 93–106. tonen, Ilkka Jormanainen, and Arnold Pears. 2021. Teaching machine learn-
[20] Mafalda Gamboa. 2022. Living with Drones, Robots, and Young Children: In- ing in K-12 Classroom: Pedagogical and technological trajectories for artifi-
forming Research through Design with Autoethnography. ACM International cial intelligence education. IEEE Access 9, July (2021), 110558–110572. https:
Conference Proceeding Series (2022). https://doi.org/10.1145/3546155.3546658 //doi.org/10.1109/ACCESS.2021.3097962
[21] Google. [n. d.]. Teachable Machine. https://teachablemachine.withgoogle.com/ [43] David Touretzky, Christina Gardner-McCune, Fred Martin, and Deborah Seehorn.
[22] Oskar J Gstrein and Anne Beaulieu. 2022. How to protect privacy in a datafied 2019. Envisioning AI for K-12: What should every child know about AI?. In
society? A presentation of multiple legal and conceptual approaches. Philosophy Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 33. 9795–9799.
& Technology 35, 1 (2022), 1–38. [44] UNICEF. 2019. Workshop Towards Global Guidance on AI and Child Rights. Tech-
[23] Ole Sejer Iversen and Christian Dindler. 2014. Sustaining participatory design nical Report June.
initiatives. CoDesign 10 (2014), 153–170. https://doi.org/10.1080/15710882.2014. [45] Marie E. Vachovsky, Grace Wu, Sorathan Chaturapruek, Olga Russakovsky,
963124 Richard Sommer, and Li Fei-Fei. 2016. Towards more gender diversity in CS
[24] Ole Sejer Iversen, Rachel Charlotte Smith, and Christian Dindler. 2018. From through an artificial intelligence summer program for high school girls. SIGCSE
computational thinking to computational empowerment. (2018), 1–11. https: 2016 - Proceedings of the 47th ACM Technical Symposium on Computing Science
//doi.org/10.1145/3210586.3210592 Education (2016), 303–308. https://doi.org/10.1145/2839509.2844620
[25] Marguerita Lane and Anne Saint-Martin. 2021. The impact of Artificial Intelli- [46] Jessica Van Brummelen, Viktoriya Tabunshchyk, and Tommy Heng. 2021. Alexa,
gence on the labour market: What do we know so far? (2021). Can i Program You?": Student Perceptions of Conversational Artificial Intelligence
[26] Phoebe Lin and Jessica Van Brummelen. 2021. Engaging Teachers to Co-Design before and after Programming Alexa. Proceedings of Interaction Design and
Integrated AI Curriculum for K-12 Classrooms. In Proceedings of the 2021 CHI Children, IDC 2021 February (2021), 305–313. https://doi.org/10.1145/3459990.
Conference on Human Factors in Computing Systems. 1–12. 3460730 arXiv:2102.01367
[27] Duri Long and Brian Magerko. 2020. What is AI literacy? Competencies and [47] Maarten Van Mechelen, Rachel Charlotte Smith, Marie-Monique Schaper, Mar-
design considerations. In Proceedings of the 2020 CHI Conference on Human Factors iana Aki Tamashiro, Karl-Emil Kjær Bilstrup, Mille Skovhus Lunding, Mari-
in Computing Systems. 1–16. anne Graves Petersen, and Ole Sejer Iversen. 2022. Emerging technologies in
[28] Laura Malinverni, Cristina Valero, Marie Monique Schaper, and Isabel Garcia K–12 education: A future HCI research agenda. ACM Transactions on Computer-
de la Cruz. 2021. Educational Robotics as a boundary object: Towards a research Human Interaction (2022).
agenda. International Journal of Child-Computer Interaction 29 (2021), 100305. [48] Henriikka Vartiainen, Tapani Toivonen, Ilkka Jormanainen, Juho Kahila, Matti
[29] Ha Nguyen. 2022. Examining Teenagers’ Perceptions of Conversational Agents Tedre, and Teemu Valtonen. 2020. Machine learning for middle-schoolers: Chil-
in Learning Settings. Proceedings of Interaction Design and Children, IDC 2022 dren as designers of machine-learning apps. In 2020 IEEE Frontiers in Education
(2022), 374–381. https://doi.org/10.1145/3501712.3529740 Conference (FIE). IEEE, 1–9. https://doi.org/10.1109/FIE44824.2020.9273981
[30] Leire Ozaeta, Manuel Graña, Maya Dimitrova, and Alexander Krastev. 2018. [49] Xiaoyu Wan, Xiaofei Zhou, Zaiqiao Ye, Chase K. Mortensen, and Zhen Bai. 2020.
Child oriented storytelling with NAO robot in hospital environment: preliminary SmileyCluster. In Proceedings of the Interaction Design and Children Conference.
application results. Problems of Engineering Cybernetics and Robotics 69 (2018), ACM, New York, NY, USA, 23–35. https://doi.org/10.1145/3392063.3394440
21–29. [50] Randi Williams and Cynthia Breazeal. 2020. How to train your robot: A middle
[31] Blakeley H Payne. 2019. An ethics of artificial intelligence curriculum for middle school AI and ethics curriculum. IJCAI.
school students. MIT Media Lab Personal Robots Group. Retrieved Oct 10 (2019), [51] Randi Williams, Hae Won Park, and Cynthia Breazeal. 2019. A is for artificial
2019. intelligence: the impact of artificial intelligence activities on young children’s
[32] Juan David Rodríguez-García, Jesús Moreno-León, Marcos Román-González, and perceptions of robots. In Proceedings of the 2019 CHI Conference on Human Factors
Gregorio Robles. 2020. Introducing Artificial Intelligence Fundamentals with in Computing Systems. 1–11.
LearningML. In Eighth International Conference on Technological Ecosystems for [52] Sarah Woods. 2006. Exploring the design space of robots: Children’s perspectives.
Enhancing Multiculturality. ACM, New York, NY, USA, 18–20. https://doi.org/ Interacting with Computers 18, 6 (dec 2006), 1390–1418. https://doi.org/10.1016/j.
10.1145/3434780.3436705 intcom.2006.05.001
[33] Elisa Rubegni, Laura Malinverni, and Jason Yip. 2022. "Don’t let the robots walk [53] Julia Woodward, Zari McFadden, Nicole Shiver, Amir Ben-hayon, Jason C Yip,
our dogs, but it’s ok for them to do our homework": children’s perceptions, fears, and Lisa Anthony. 2018. Using co-design to examine how children conceptualize
and hopes in social robots. Proceedings of Interaction Design and Children, IDC intelligent interfaces. In Proceedings of the 2018 CHI Conference on Human Factors
2022 (2022), 352–361. https://doi.org/10.1145/3501712.3529726 in Computing Systems. 1–14.
[34] Ismaila Temitayo Sanusi and Solomon Sunday Oyelere. 2020. Pedagogies of [54] Junnan Yu, Janet Ruppert, Ricarose Roque, and Ben Kirshner. 2020. Youth civic
Machine Learning in K-12 Context. Proceedings - Frontiers in Education Conference, engagement through computing: cases and implications. ACM Inroads 11, 4
FIE 2020-Octob (2020). https://doi.org/10.1109/FIE44824.2020.9274129 (2020), 42–51. https://doi.org/10.1145/3432727
[35] Marie-Monique Schaper, Laura Malinverni, and Cristina Valero. 2020. Robot
Presidents: Who should rule the world? Teaching Critical Thinking in AI through
Reflections upon Food Traditions. In Proceedings of the 11th Nordic Conference
on Human-Computer Interaction: Shaping Experiences, Shaping Society. https:
//doi.org/doi.org/10.1145/3419249.3420085
[36] Marie-Monique Schaper, Rachel Charlotte Smith, Mariana Aki Tamashiro,
Maarten Van Mechelen, Mille Skovhus Lunding, Karl-Emil Kjæer Bilstrup, Mag-
nus Høholt Kaspersen, Kasper Løvborg Jensen, Marianne Graves Petersen, and
Ole Sejer Iversen. 2022. Computational empowerment in practice: Scaffolding
teenagers’ learning about emerging technologies and their ethical and societal
impact. International Journal of Child-Computer Interaction (2022), 100537.
[37] Laura Scheepmaker, Kay Kender, Christopher Frauenberger, and Geraldine Fitz-
patrick. 2021. Leaving the field: Designing a socio-material toolkit for teachers
to continue to design technology with children. Conference on Human Factors in
Computing Systems - Proceedings (2021). https://doi.org/10.1145/3411764.3445462
[38] Rachel Charlotte Smith and Ole Sejer Iversen. 2018. Participatory design for
sustainable social change. Design Studies 59 (nov 2018), 9–36. https://doi.org/10.
1016/j.destud.2018.05.005
[39] Ahuva Sperling and Dorit Lickerman. 2012. Integrating AI and machine learning
in software engineering course for high school students. Annual Conference on
Innovation and Technology in Computer Science Education, ITiCSE (2012), 244–249.
https://doi.org/10.1145/2325296.2325354
[40] Shashank Srikant and Varun Aggarwal. 2017. Introducing data science to school
kids. In Proceedings of the 2017 ACM SIGCSE technical symposium on computer
science education. 561–566.

309

You might also like