The Problem I Identified
U.S. high school students have the second-lowest GPA in science among all subjects (NCES data). But the deeper issue isn't performance — it's identity. Students who struggle with abstract concepts and complex language start believing they're simply "not a science person." Once that belief sets in, STEM career paths close off before students have a real chance to explore them.
The Root Cause
Science misconceptions don't get corrected through lecture. They get corrected through dialogue — when students are forced to articulate their ideas, hear alternatives, and revise their thinking. Most classrooms don't create the right conditions for that to happen consistently.
How I Approached the Solution
I designed a two-loop adaptive system grounded in social constructivism — each loop targeting a different level of the learning problem:
Peer Discussion
Heterogeneous group discussions using a Driving Question Board (DQB) to surface and organize misconceptions. Lower-understanding students get accessible peer support; higher-understanding students deepen their knowledge by explaining.
LLM Feedback
Targeted, real-time AI feedback on individual student explanations. Rather than broad corrections, the LLM responds to the specific misconception in each student's own words — making the feedback feel relevant and actionable.
The key design insight was that peer discussion and AI feedback work at different levels — discussion reveals what students believe, AI feedback corrects it precisely. Neither works as well alone.
What This Demonstrates
This project shows that I can map a well-documented educational problem — science disengagement and misconception persistence — to a two-level adaptive architecture. The Task Loop / Step Loop framework is borrowed from adaptive learning systems design, applied here to a collaborative science context.
The result is a module that addresses both the cognitive problem (misconceptions) and the motivational one (identity) — by giving students a structured space to think out loud, get corrected gently, and experience science as something they can actually do.