A dashboard that tells you how AI is changing the way you learn.

Context

Concerns about AI's effect on student learning have been mostly framed in terms of cheating. The more interesting concern, in our view, is metacognitive: students lose track of the line between independent thought and AI-assisted thought, and existing tools don't help them see it. No existing system combines AI usage tracking with structured metacognitive reflection.

We set out to build that system and study what students actually wanted from it.

The question we asked

The dominant framing — "how do we stop students from using AI?" — wasn't interesting to us. Students are going to use it. The question that mattered was different.

Not "how do we stop students from using AI." But: "what would let students stay metacognitive inside AI use?"

What we did

• Literature review across cognitive science, HCI, and learning sciences (Winne, Fan, Bastani, Lehmann, Singh).
• First prototype: a standalone reflection dashboard built on Bloom's Taxonomy.
• Needfinding interviews with 15 students (5M, 10F, mean age 24.7), analyzed via thematic + grounded theory coding.
• Pivoted: rebuilt the prototype as a ChatGPT side-panel extension that surfaces reflection at the point of use.
• Usability testing with 6 additional students on the second prototype.

The finding that mattered

Ten of fifteen students liked the dashboard. They also said it was too passive. Stats showed activity, but they didn't help students understand the quality of their collaboration with AI, whether they were becoming too dependent, or what to do differently.

This killed our first prototype. Students didn't want a dashboard to look at later. They wanted reflection delivered in the moment, attached to the work they were already doing. That insight reshaped the entire intervention.

The second prototype put reflection inline, in the side panel, where the student was already typing. Usability testing confirmed it: users wanted fewer, sharper, more actionable reflection moments — not more data.

What this points to

The dominant frame for AI literacy in education is detection ("did you use AI?") and policy ("you shouldn't have"). Neither will hold. What might work: tools that respect that students will use AI and help them stay aware of how it's shaping their thinking. I'm developing this into a research framework I'm calling cognitive delegation — the idea that AI use isn't cheating, it's delegation, and the question is which delegations build skill versus atrophy it.

What I'd do next

Run a longer-term study. Most existing work on AI-in-learning is point-in-time — a survey, an interview, a usability test. What we need is longitudinal: tracking the same students across a semester to see whether reflective AI use actually changes outcomes.