Abstract
Through a NSF funded ITEST program, labeled ImageSTEAM, a summer workshop was conducted at UGA by the team to empower middle school teachers to Artificial Intelligence (AI) topics. For the summer workshop, teachers were introduced to AI topics including computer vision, machine learning and computational cameras as part of their professional development. After the introduction, lesson modules were co-created with teachers to introduce them to middle school students in their classroom. The lesson modules were related to math/science/art topics and align with state and national standards of learning in middle school curriculum. After developing the lesson module, each teacher presented the lesson to workshop students to gauge their understanding and interest in AI topics. This paper discusses the variety of lesson modules developed by the teachers and how the lesson modules were adopted into middle school curriculum.
Introduction: Bridging the AI Literacy Gap
While Artificial Intelligence (AI) permeates modern life, K-12 education has struggled to keep pace. A significant "training gap" exists: most teachers lack a formal background in AI, and existing resources often relegate AI to computer science electives. This creates a bottleneck that prevents a broad, diverse group of students from understanding the technology that will define their future.
The ImageSTEAM project addresses this gap by empowering middle school teachers to integrate computer vision and machine learning into core subjects like science, math, and art. The challenge is twofold: providing teachers with technical fluency and finding creative ways to embed AI concepts into "core" classes so that AI education is accessible to all students, not just those in technical electives.
Methodology: The Learn-Create-Teach Model
The researchers conducted a two-week summer workshop at the University of Georgia (UGA) following a three-phase model:
- Professional Development: Teachers were introduced to fundamentals like computer vision and machine learning using "low-code/no-code" tools such as Google’s Teachable Machine, NVIDIA’s GauGAN, and Tinkercad.
- Co-Creation: Teachers worked with the project team to develop lesson modules aligned with Georgia’s state standards.
- Pilot Implementation: Teachers delivered their lessons to students, providing an immediate feedback loop for engagement and technical hurdles.
Datasets and Practical Application
Rather than using large-scale academic datasets, the program focused on user-generated data created during the learning process. This approach makes the technology tangible. For example, in a physics lesson on light waves (Translucent, Transparent, and Opaque), students created a custom dataset to train a model to recognize how different substances affect visibility.
In this specific module, teachers and students generated 104 image samples to train the AI to determine if a letter could be seen through a glass. This hands-on data collection reinforces the concept that AI models are only as good as the data provided to them.
Key Insights
- AI as a Tool, Not Just a Topic: Successful modules used AI to explore existing curriculum—such as classifying rocks in geology or analyzing Spanish art—making it a functional tool for inquiry.
- The Expertise Threshold: Feedback revealed a "knowledge-to-confidence" lag. Even after training, some teachers desired more rigid, pre-set curricula before feeling fully equipped to innovate independently.
- Inclusivity through Integration: By targeting Title I schools and core subjects, the program ensures diverse exposure to AI, regardless of a student's eventual career path.
Results and Future Work
Six teachers successfully developed and piloted modules covering geology, geometry, CAD modeling, and art. Qualitative feedback was overwhelmingly positive, with teachers noting the structure was "user-friendly for kids." Finalized resources are now publicly available at imagesteam.org.
Future work will focus on scaling the program beyond the initial cohort and addressing the "elective temptation"—the pressure to move AI out of core subjects into specialized classes. Long-term evaluation of student retention and learning outcomes is currently underway.
Summary updated