Visible Adoption, Untracked Contribution: GitHub Evidence of the Accountability Gap Across Three Cohorts of an HCI Prototyping Course

Disclosure alone doesn't ensure accountability: While AI tool disclosure increased dramatically from 0% to 66% of repositories across three cohorts, explicit attribution of what specific tasks AI actually performed remained rare. This reveals a significant "accountability gap" where students name that they used AI tools but don't explain what those tools contributed to their work.

Current accountability frameworks are outdated for modern AI use: The paper argues that disclosure-based accountability systems designed for "episodic, identifiable acts" fail to capture how students actually use AI today—as continuous, ambient infrastructure embedded in their workflows. Students aren't deliberately hiding AI use; rather, the tools have become so integrated that specifying their contributions is difficult.

AI is becoming invisible infrastructure in educational settings: By 2025, AI tools were embedded directly into course templates and student-built devices, making it harder to track where human work ends and AI contribution begins. This shift from occasional tool use to ambient infrastructure represents a fundamental change in how accountability needs to be approached.

The problem requires rethinking, not just better tracking: The authors argue the real challenge isn't that students conceal AI use, but that we need new frameworks for "genuine co-thinking accountability" that can handle continuous human-AI collaboration rather than discrete, easily-attributable contributions.