Learning AI in and Through the Arts: Insights From a Playful, Artful AI Curriculum Implementation

• In Event: Integrating AI Education in K–12 Humanities Curricula: Opportunities, Challenges, Design Considerations, and Promising Solutions

Abstract

Objectives or purposes: The myriad social factors constituting AI and its potential to reshape every domain of society mean teachers outside of STEM–those in the arts and humanities–are poised to significantly contribute to how and what students learn about AI. This paper reviews the implementation of artful, playful and unplugged AI activities in an 8th grade technology classroom; we examine unplugged activities in this context and explore why an art classroom could benefit students learning about AI from the same activities.

Theoretical framework: Based on our investigation (Stoiber et al., 2023) of teacher interaction with AI curricula, our approach is based on the understanding that 1) STEM teachers are not AI content experts, and 2) teachers in STEM classrooms may be reticent or uncomfortable engaging in explicit instruction around the social implications of AI. Our AI activities are designed with these considerations in mind, built on principles of learning playfully and artfully.
Halverson (2021) describes 3 principles/outcomes explicating why learning through the arts is powerful: 1) Creating, sharing, and critiquing representations foundational to success in schools; 2) Identity is developed and intertwined with processes of experimenting, creating, and enacting; and 3) Collaboration is embedded. Zosh and colleagues (2017) define 5 aspects of playful learning, that is joyful, hands-on, iterative, meaningful, and socially interactive.

Methods and Data Sources: AI activities were developed and implemented as part of the second year implementation of a Design-Based Research study. Activities were designed as one- or two-day lessons culminating in an interest-driven project. Data collected includes audio and visual recordings, memos, interviews with students and teachers, and student artifacts generated over two months during 9 classroom implementation and research visits. Cross-case analysis (Miles et al., 2020) is underway to understand the affordances and constraints of each activity in relation to teacher expertise.

Results: Students and our teacher reported positively to the spectrum of activities, with the most positive reactions to totally unplugged activities. What we call Teachable Machine Unplugged to introduce students to supervised and unsupervised machine learning (lessons 1 & 2) were particularly successful, engaging every student to participate by embodying either: 1) Data points–which can be sorted by 2) a Labeler–which can in turn train 3) the AI agent. For example, a student posing as a tennis player represents a piece of data to train a Summer Vs. Winter Olympics AI classifier. As activities scaffolded to invite and support more student creativity (e.g., “TikTok Unplugged”), our teacher struggled to direct their creativity and make connections to the AI.

Significance: Our study has shown unplugged, artful, and playful AI learning activities are engaging, accessible, and enjoyable. We argue art and humanities classrooms, in addition to STEM spaces, are an ideal fit for our unplugged activities. For example, art teachers are equipped to scaffolded learners from ephemeral embodiments in Teachable Machine Unplugged to enduring artistic representations that can serve to nuance the activity and support imperative discussions around the ethics of generative AI and art. In our implementation, students displayed significant concern for this topic, which our teacher was not prepared to unpack.

Authors

• Andy Stoiber, University of Wisconsin - Madison

• YJ Kim, University of Adelaide