Developing Middle School Students' Artificial Intelligence Literacy Through Interactive Approaches

• In Event: Teaching Artificial Intelligence in K–12: Innovative Learning Designs in an Evolving Landscape

Abstract

OBJECTIVE
While AI education becomes increasingly relevant, teaching K-12 students AI, however, is not easy because the curricula must be engaging and age-appropriate to young learners. In this paper we report a curriculum that aims to foster AI literacy through interactive activities that integrate the technical, ethical, and career aspects of AI. Our approach was established based on previous work (e.g., authors, 2021; Payne, 2020) and the notion that AI is not only a technical field but one that has wide-ranging societal impacts.

THEORETICAL FRAMEWORK
The work was mainly informed by research on interactive instructional approaches in AI/CS education. Marques et al. (2020), in their systematic mapping study, highlighted the importance of using interactive approaches to teach machine learning at the K-12 level. Activities that first expose students to AI concepts through interactions with AI tools as an end-user and then engage them in knowledge building and reasoning about how AI works are beneficial. They also suggested using hands-on activities to teach complex AI processes to prevent students from being cognitively overwhelmed.

THE CURRICULUM
The curriculum features kinesthetic learning (Sivilotti & Pike, 2007), participatory simulation (Klopfer et al., 2005), and unplugged activities (Lindner et al., 2019). Students make sense of the technical aspects of AI through experimenting with AI-inspired tools and link their technical understanding to ethical issues by considering how datasets and models contribute to bias in AI. For instance, in the Supervised Learning (SL) module, students first learn to train SL models using Google’s Teachable Machine. They then experiment and compare the models trained with datasets of unbalanced and balanced data, identify the biased model, brainstorm ways to mitigate the bias, and discuss societal implications of using biased models (e.g., facial recognition system).

METHODS
We implemented the 30-hour curriculum in five online summer camps with 49 middle school students (56% male, 43% female, 84% racial and ethnic minorities). Students met online for three hours every day for two weeks. All the activities were implemented synchronously via Zoom and all curricular materials were accessible through Google Classroom. The camps were taught by a team of teachers who were trained in a professional development program on AI and the curriculum (2 to 3 teachers per camp).

FINDINGS & SIGNIFICANCE
Comparing students’ pre/post-test performances shows that students significantly improved their understanding of key AI concepts after the camp (pretest: Mean (SD) =25.06 (3.34), total score: 40; posttest: Mean (SD)=27.37 (3.31), t(48)=3.42, p<.01, Cohen’s d=.70). At exit, students were able to identify bias of AI tools and describe ways to mitigate bias based on their understanding of machine learning. Classroom observations revealed that the interactive activities and positioning AI as closely related to them engaged all students, particularly female students of color. Overall, our work informs the AI education field by providing an innovative approach of engaging all learners and a working definition of AI literacy for middle school students.

Authors

• Irene A. Lee, Massachusetts Institute of Technology
  Presenting Author

• Helen Zhang, Boston College
  Non-Presenting Author

• Kate Moore, Massachusetts Institute of Technology
  Non-Presenting Author

• Beatriz Perret, Massachusetts Institute of Technology
  Non-Presenting Author