Download - Unplugged Assignments for K-12 AI Education
AI MATTERS, VOLUME 7, ISSUE 1 MARCH 2021
Unplugged Assignments for K-12 AI EducationDuri Long (Expressive Machinery Lab, Georgia Institute of Technology; [email protected])Jonathan Moon (Expressive Machinery Lab, Georgia Institute of Technology; [email protected])Brian Magerko (Expressive Machinery Lab, Georgia Institute of Technology;[email protected])DOI: 10.1145/3465074.3465078
Introduction
In this column, we introduce our two “un-plugged” (i.e. no technology needed) ModelAI Assignments: Introducing AI and Seman-tic Networks and Knowledge Representations.We also reflect on the potential benefits of un-plugged activities for broadening access to AI-related learning experiences.
Why “Unplugged” Activities to Learnabout AI?
Resources for computer science (CS) educa-tion that do not require technology have be-come valuable in computing education for avariety of reasons, including their low cost,ease of implementation, incorporation of phys-ical/embodied interaction, and often playfulnature (Nishida et al., 2009). Inspired by these“CS Unplugged” materials (Bell, Rosamond,& Casey, 2012), there have been a few ex-isting online resources for AI education devel-oped in the past year or two that do not re-quire technology. Ali et al. have developedan unplugged middle-school curriculum for AIethics (Ali, Payne, Williams, Park, & Breazeal,2019) and Lindner et al. have developed asix-lesson unplugged curriculum for teachingabout concepts like decision trees and re-inforcement learning (Lindner, Seegerer, &Romeike, 2019). A few other resources forunplugged AI resources that have not beenformally published have been recently madeavailable as lesson plans online (Microsoft,n.d.; Group, n.d.; Krueger, n.d.; Seegerer &Lindner, n.d.).
There remains a lot of space in the field for thedevelopment of additional “unplugged” AI ac-tivities. These activities have the potential tobroaden access to AI-related learning experi-ences at a low cost to educators. They also
Copyright c© 2021 by the author(s).
have the potential to be more engaging fornovice audiences, since they involve hands-on paper-based activities that typically do notrequire prerequisite coding knowledge.
The Model AI Assignments
In this column, we present two unplugged as-signments that aim to teach introductory AIconcepts to young learners with no prior ex-perience in AI or computer science.
Introducing AI
The first assignment, Introducing AI, is in-tended as a high-level introduction to AI andcan be used to kick off an AI-related class,unit, or workshop. Students engage in an in-teractive worksheet activity and explore ques-tions such as: What is artificial intelligence?;Where have you used it before?; How do youfeel about it?; How does it work?. This assign-ment can be completed as a worksheet activ-ity or the worksheet can be used as a guide tolead an in-class activity.
The activity requires the use of a printabledeck containing cards with examples of AItechnologies and possible inputs, algorithms,and outputs for AI devices. The worksheetactivity begins by prompting students to con-sider where they have seen AI before and howthey feel about AI. Then, they are asked tolook at the examples of AI in the card deckand select cards with technologies they haveinteracted with previously. Finally, studentsare walked through a high-level explanation ofhow AI works and are guided by the worksheetto create an imaginary AI using the input, al-gorithm, and output cards. Our assignmentalso includes additional questions for learnersto reflect on after the activity, such as prompt-ing learners to reflect on the strengths andweaknesses of their imaginary AI device or
10
AI MATTERS, VOLUME 7, ISSUE 1 MARCH 2021
asking learners to discuss whether they thinktheir feelings about AI might change in the fu-ture.
In our experience, this activity has been en-gaging for family groups with children ages 6and up. Younger children (6-9) needed sup-port from adults or older siblings to read theworksheet instructions/card descriptions butwere able to actively engage in identifying AIexamples they had previously interacted with,discussing their feelings about AI, and find-ing sensor/dataset/algorithm cards to createan imaginary AI (mostly using the pictures onthe front as a guide).
This activity aims to equip learners to be ableto a) identify AI technologies that they haveused before; b) distinguish between technol-ogy that uses AI and technology that does notuse AI; c) identify their preconceptions anddiscuss their feelings about AI; d) define theterms sensor, dataset, and algorithm and rec-ognize several examples of each; and e) ex-plain that AI takes an input, processes that in-put using an algorithm, and produces an out-put.
Semantic Networks and KnowledgeRepresentations
The second assignment, Semantic Networksand Knowledge Representations, is focusedon communicating concepts related to knowl-edge representations and reasoning. AIagents store and organize information in theirmemory using structures known as knowledgerepresentations. One type of knowledge rep-resentation is a semantic network. Seman-tic networks are a way of representing rela-tionships between objects and ideas. For ex-ample, a network might tell a computer therelationship between different animals (e.g. acat IS-A mammal; a cat HAS whiskers). Inthis assignment, learners can create their ownsemantic networks (Figure 1) by gluing downprintable cards containing concepts (e.g. cat,mom, friend) and arrows containing relation-ships (is, has, likes, dislikes). Provided carddecks contain concepts related to animals,family, and musical instruments. Blank cardsare also provided to allow learners to makenetworks on custom topics.
Students can simulate an AI-user interactionusing their semantic networks. Two students
Figure 1: Example of a completed semantic net-work activity
can trade completed semantic networks andask their partner questions about the networkthey created (e.g. “What is a cat?”). The stu-dent’s partner should answer the questionsusing the semantic network as their only guide(simulating an AI agent whose only knowledgeis based on the semantic network). Learn-ers are then encouraged to reflect on the net-works they create and consider the strengthsand limitations of the knowledge representa-tion using a provided list of questions that canbe used to foster discussion or as a written ac-tivity.
In our experience, this activity has been en-gaging for learners ages 6 and up, althoughyounger learners many need some adult sup-port during the latter half of the activity whenthey are asked to simulate an AI agent usingtheir network. We have also observed thatthe activity is engaging and fosters learning forlearners with little to no prior knowledge aboutAI. This assignment could be adapted as ei-ther a take-home written activity or an in-classgroup project, and cards/arrows could be cus-tomized to foster interdisciplinary connections.
This activity aims to help learners a) under-stand that one way computers store common-sense knowledge is using networks of con-
11
AI MATTERS, VOLUME 7, ISSUE 1 MARCH 2021
nected concepts and relationships; b) explainat a high-level how a computer would use thenetwork they built to answer questions (e.g.follow the HAS arrows connected to “cat” toanswer “What does a cat have?”); and c) re-flect on the strengths and limitations of se-mantic networks as a way of representingknowledge.
References
Ali, S., Payne, B. H., Williams, R., Park, H. W.,& Breazeal, C. (2019). Construction-ism, ethics, and creativity: Developingprimary and middle school artificial intel-ligence education. In International work-shop on education in artificial intelligencek-12 (eduai’19).
Bell, T., Rosamond, F., & Casey, N. (2012).Computer science unplugged and re-lated projects in math and computer sci-ence popularization. In The multivariatealgorithmic revolution and beyond (pp.398–456). Springer.
Group, C. E. R. (n.d.). Artificial Intelli-gence. Retrieved from https://classic.csunplugged.org/artificial-intelligence/
Krueger, N. (n.d.). 3 unplugged activities forteaching about AI.
Lindner, A., Seegerer, S., & Romeike, R.(2019). Unplugged Activities in the Con-text of AI. In International Conference onInformatics in Schools: Situation, Evo-lution, and Perspectives (pp. 123–135).Springer.
Microsoft. (n.d.). Unplugged activity: Pa-per AI. Retrieved from https://minecraft.makecode.com/courses/csintro/ai/unplugged
Nishida, T., Kanemune, S., Idosaka, Y.,Namiki, M., Bell, T., & Kuno, Y. (2009).A CS unplugged design pattern. ACMSIGCSE Bulletin, 41(1), 231–235. (Pub-lisher: ACM New York, NY, USA)
Seegerer, S., & Lindner, A. (n.d.). AI Un-plugged. Retrieved from https://www.aiunplugged.org/
Duri Long is a Ph.D.candidate in Human Cen-tered Computing at Geor-gia Institute of Technol-ogy. Her research fo-cuses on issues relatedto human-centered AI, in-cluding fostering public AIliteracy and human-AI in-
teraction through co-creative, embodied expe-riences.
Jonathan Moon is anIndustrial Designer witha background in userexperience research andtraditional product design.Jonathan was the leaddesigner of the carddecks for both activitiespresented in this column.
Brian Magerko is a Pro-fessor of Digital Media,Director of GraduateStudies in Digital Me-dia, and head of theExpressive MachineryLab at Georgia Tech. Hisresearch explores how
studying human and machine cognition caninform the creation of new human/computercreative experiences.
12
