2023 ASEE Annual Conference & Exposition

The effective introduction of the fundamentals of artificial intelligence (AI) to middle school students requires the novel integration of the existing science curriculum and AI concepts. This research focuses on leveraging 6th and 7th-grade science curricula related to state standards to introduce machine learning concepts by using fossil shark teeth. Researchers from engineering, education, and paleontology collaboratively developed learning modules to upskill Title I schoolteachers to meaningfully integrate AI fundamentals within their existing curriculum. With a special emphasis on machine learning (ML), five lesson plans were presented during a week-long teacher professional development. Teachers conceptualized and implemented ML models that distinguish fossil shark teeth by their taxonomy and primary functions to recognize ecological and evolutionary patterns. After introducing a lesson, each teacher curated the lesson plan content to directly relate to their specific context, in collaboration with each other and our research team.

We built the curriculum leveraging students’ existing conceptions and misconceptions about AI from prior work while testing the feasibility of addressing AI learning objectives, as well the AI4K12’s Five Big Ideas, in the broader context of middle school science, technology, engineering, mathematics, and computing (STEM+C) education. Our lessons were scaffolded using the machine learning development process: 1) data collection and preparation; 2) selecting and training the model; 3) evaluating the models’ accuracy; 4) tuning model parameters to improve performance. Each stage of the development process constituted a different lesson during a week-long summer professional development. Through these lessons, teachers were introduced to several open-source AI tools, including two platforms used to build/train ML models: Google’s Teachable Machine and Roboflow. The fifth and final day of the professional development gave teachers time to conceptualize how these lessons could be integrated with their existing curricula.

Initial feedback from the summer PD indicated we overestimated the teachers’ familiarity with technology. More time was necessary to orient teachers to each AI tool. Teachers readily adopted the use of Seek by iNaturalist and myFossil. However, the teachers’ use of AI tools in their classrooms highly favored Google’s Teachable Machine to Roboflow, which may relate to the affordances and constraints of each tool. Preliminary mixed-method data analyses show teachers' self-efficacy around teaching AI improved after engaging in the summer PD. Longitudinal data collection is underway and will inform future work related to improving teacher and student self-efficacy related to teaching and learning AI, respectively.