Project Story

What Inspired Us

University policies—like strict attendance rules, grading curves, or frequent assessments—are usually created with good intentions. But from our own campus experiences, we noticed that these policies often have second‑order effects: increased stress, surface learning, or unintended inequities. These consequences are rarely discussed upfront.
This inspired us to build a tool that helps people explore trade‑offs, rather than predict outcomes or claim what is “best.”

What We Learned

• AI is most valuable as a thinking partner, not a predictor.
• Clear structure and constraints reduce hallucinations more effectively than complex models.
• Separating roles—backend/system plumbing vs AI reasoning design—keeps projects focused.
• Multi‑agent systems work well when each agent has a single, narrow responsibility.

How We Built the Project

We built a lightweight web application with a JavaScript frontend and a Python backend.

1. Users adjust a few campus policy parameters.
2. The backend passes these inputs through a fixed sequence of AI agents.
3. Each agent analyzes the policy from a different perspective and returns structured bullet points.
4. A final synthesis highlights trade‑offs and feedback loops.

Conceptually, we treated the system as an exploration across perspectives, not a predictive model:

• Policy Inputs
• Multiple Agents
• Structured Trade-offs

There is no data training, no statistics, and no accuracy claims.

Challenges We Faced

• Resisting the urge to over‑engineer.
• Preventing AI outputs from sounding overly confident or predictive.
• Deciding what not to build within a hackathon timeframe.
• Aligning “good engineering” with responsible AI framing.

Outcome

The final result is a decision‑support tool that makes reasoning explicit and transparent. It doesn’t tell users what to do—it helps them think more clearly about the possible consequences of policy decisions.

Built With

• flask
• json
• postgresql
• python
• react
• redis