AI-Guided Circuit Reasoning System

Highly capable AI can accelerate task completion while reducing the need for people to construct their own explanations.

This project introduces a mobile reasoning support system that withholds direct answers and uses staged prompts. A pre/post study compared this interaction model with a control condition to evaluate effects on conceptual reasoning.

• Flutter mobile interaction environment
• Go backend
• LLM-based structured prompting

• Dialogic reasoning prompts instead of answer dumps.
• Commit-and-check workflow to surface mental models.
• Stepwise validation tied to claim quality.

The system refuses shortcut completion. Users move forward only after generating their own intermediate claims, which supports active cognition and reflection.

Participants. 10 senior high school students.

Setup. Two groups, control vs. AI-guided, with a pre-test / post-test design over an introductory circuits unit.

Interaction structure changes how people think, not only what they recall.

• Small sample size.
• Short study duration.
• Baseline imbalance between groups.

• Larger randomized studies.
• Multi-topic expansion.
• Integration with interactive systems (Circuit Quest).