🐻 TEACH THE BEAR

A Gamified Clinical Reasoning Tool for T2DM Education

Can you teach a clueless bear how to treat diabetes?

SHNackathon 2026 · GRIZZLY · February 2026

MEET THE Creator

Jeea

Frontend / UI Design

• first time vibe coder YAY

first year undergraduate student at McMaster University �

🐻

"Umm...

what is

metformin?"

THE PROBLEM

📚

Medical students are drowning in passive information — guidelines, PDFs, lectures

🧠

Passive reading creates the illusion of competence without real understanding

❌

Existing tools (UWorld, AMBOSS, Anki) test recall — not clinical reasoning

⚡

T2DM drug selection is complex: comorbidities, eGFR, CV risk all change the answer

The problem is not access to information. It is the failure to engineer learning.

OUR SOLUTION

🐻 Teach the Bear

You are a medical student. Dr. Teddy is clueless. Your job: write him a treatment plan on his clipboard (make a cheat sheet for him)

🖊️

Whiteboard Canvas

Draw, type, highlight & diagram a treatment plan from memory

📖

Peek-able Notes

Slide-out synthesized guideline notes — but the bear is watching

🎙️

Voice Input

Speak your reasoning directly onto the whiteboard

🧑‍⚖️

Supervising Bear

Strict AI grader gives pass/fail + detailed clinical feedback

LEARNING THEORY

THEORY

Desirable Difficulties

+ Protégé Effect

DESIGN FEATURE

Whiteboard-first creation before peeking at notes

LEARNER BEHAVIOUR

Student generates a treatment plan from memory as the 'teacher'

COGNITIVE PROCESS

Forced retrieval + self-explanation reduces illusion of competence

EXPECTED OUTCOME

Better long-term retention & transfer to real clinical scenarios

Grounded in: Retrieval Practice · Desirable Difficulties · The Protégé Effect · Metacognitive Calibration

TECHNOLOGY STACK

⚛️

React

UI framework — single-page app

🎨

HTML5 Canvas

Freehand drawing & whiteboard

🤖

Claude API

Supervising Bear grading + notes synthesis

🎙️

Web Speech API

Voice-to-whiteboard transcription

💅

Tailwind CSS

Pixel-art gamified styling

🚀

Vercel

Free instant deployment & live link

🔑 All AI grading is powered by Anthropic's Claude API — reading whiteboard content against clinical T2DM guidelines from the dataset

HOW IT WORKS

1

📋 Receive Case

Student sees a pixel-art patient card + clinical scenario (e.g. Margaret, 67F, HFrEF, eGFR 35, A1C 9.2%)

2

🖊️ Create Cheat Sheet

Student writes, draws, and diagrams a treatment plan on Dr. Teddy's whiteboard — from memory

3

👀 Peek if Needed

Student can slide out synthesized guideline notes — peek counter increments each time

4

📤 Submit

Whiteboard content (canvas image + text) is sent to Claude API for evaluation

5

🐻 Supervising Bear Grades

Strict bear gives pass/fail, star rating, specific clinical feedback, and options to retry or advance

Prototype

🐻

• https://www.loom.com/share/56a0a39317654b7bae58a416af9d37a9
• https://youtu.be/lERFQn4st24?si=YWWxlGbcrVDqUM_s

LIVE DEMO

https://teach-the-bear.lovable.app

IMPACT & FUTURE DIRECTIONS

🎯 IMPACT NOW

▸ Forces active retrieval over passive reading

▸ Teaches T2DM drug selection through doing, not memorizing

▸ Immediate, specific clinical feedback from AI

▸ Fun enough that students actually want to use it

▸ Works on any device — no install needed

🚀 FUTURE DIRECTIONS

▸ Expand to other clinical domains (HTN, CKD, heart failure)

▸ Spaced repetition: resurface weak cases automatically

▸ Multiplayer mode — two students debate, bear decides

▸ Integrate real EHR-style patient data for fidelity

▸ RCT to measure retention vs. traditional Q-banks

🐻 The bear is learning. Are you?