NSF grant ~$1.08M: feeling the math of machine learning on a weaving loom — undergraduate AI education (UC Irvine)

This award is a design study for undergraduate education that fuses learning with AI and learning about AI — and its stage is not a computer but a weaving loom.

1A national need for AI literacy

The backdrop is the national need for AI literacy. The project aims for all undergraduates, regardless of major, to understand, critique, and thoughtfully engage with AI, framing as nationally important the delivery of the conceptual foundations needed to use AI-enabled technologies, join public discourse, and make informed decisions.

2Feeling matrix factorization through weaving

The core idea stands out. It makes matrix factorization — the mathematical foundation of machine learning — tangible through a shaft loom used as a boundary object. A loom generates patterns within the physical constraints of how warp threads combine. Using those constraints as scaffolding, students explore core ML concepts — abstraction, pattern decomposition, compression, generalization, and overfitting — hands-on. The idea is to make abstract AI concepts something you can touch.

3The research questions

The research questions are concrete: how do loom-based physical constraints scaffold understanding of ML representations; which curriculum design features enable mastery of decomposition and abstraction; how does collaborative design between curriculum developers and instructors support disciplinary adaptation; and how can an AI conversational agent adjust scaffolding for students with backgrounds in math, programming, arts, and design.

These are evaluated with mixed methods — learning analytics, performance assessments, and qualitative analysis of student interactions — compared across multiple instructional contexts.