The Problem I Identified
Most feedback interventions focus on improving the quality of feedback. But that's not where the system breaks down. Students receive feedback — they just don't use it.
The Real Gap
Students check their score and move on. Feedback becomes a grade justification, not a growth tool. The research term for what students should be doing is "feed-forward" — using feedback to plan future improvement. Almost no one does this without structured support.
Nicol & Macfarlane-Dick (2006) identify this as a self-regulation failure: students lack the cognitive scaffolding to decode feedback, connect it to their goals, and translate it into action. The problem isn't motivation — it's structure.
How I Approached the Solution
Rather than trying to fix feedback itself, I designed a system that targets the moment feedback most often fails — the transition from receiving to acting. The insight was that this required two interlocking interventions, not one:
An AI Reflection Tool — for students
A structured dialogue tool that guides students through interpreting their feedback, connecting it to prior knowledge, and building a concrete action plan. The AI never gives answers — it scaffolds the reasoning process, pushing students from passive receivers into active planners.
A Learning Analytics Dashboard — for teachers
A dashboard that synthesizes behavioral engagement logs (did they access feedback? how quickly? how many AI turns?) with AI-scored cognitive data (quality of sense-making, specificity of goal-setting) into actionable profiles. Teachers get targeted intervention prompts — not raw data.
The Adaptive Logic: Three Learner Profiles
Crossing behavioral and cognitive data reveals three distinct learner types — each requiring a different response. This is where the system becomes genuinely adaptive rather than just descriptive:
Resourceful Learner
High engagement + high metacognition. Already doing it right.
"Surface" Checker
Low engagement + low metacognition. Checks score, exits fast.
Trial & Error Learner
Medium engagement + low metacognition. Stuck in a chat loop.
What This Demonstrates
This project shows that I can take a well-established theoretical problem — the feedback literacy gap — and translate it into a concrete, data-driven instructional system. The key design move was recognizing that fixing feedback quality was the wrong lever. The right lever was structuring what happens after feedback arrives.
By combining AI scaffolding with behavioral analytics, the system doesn't just support individual students — it gives teachers the evidence they need to intervene at scale.