Learnability-Focused AI Programming Mentor with Desirable Friction
👨💻 Author & Lead Developer: Miftahul Islam Tashfin
🎓 Course: CSE428: Human-Computer Interaction (Group C)
CodeMentor is a pedagogical VS Code extension engineered to solve the cognitive atrophy caused by modern AI code generation. Instead of allowing passive copy-pasting or unearned AI solutions, CodeMentor enforces Desirable Friction and Active Cognitive Engagement based on cognitive science learning theories (Chi's ICAP Framework & Vygotsky's Scaffolding).
Intercepts pasted code snippets and AI agent buffer insertions.
Temporarily locks the code with visual red highlighting and generates a context-aware Multiple-Choice Verification Challenge (MCQ) targeting the core logic, time complexity, or edge case.
Answering correctly dynamically unlocks the editor with glowing emerald confirmation.
2. 🗣️ "Explain Back" Mode (Self-Explanation)
After unlocking code, prompts students to provide a 1-sentence plain-English explanation of the logic.
Employs Gemini Flash to provide real-time accuracy scoring (0–100%) and conceptual feedback.
3. 📊 Learning Independence Meter
A live, real-time cognitive metric in the sidebar tracking your autonomous problem-solving performance.
Rewards active explanations and correct predictions while penalizing excessive hint-taking.
Features progressive mastery ranks: Level 1: Explorer ➔ Level 2: Architect ➔ Level 3: Master.
4. 🧠 "Lead-and-Reveal" Progressive Scaffolding
Prevents raw code dumps by breaking complex algorithmic problems into multi-tier progressive hints.
Provides high-level pseudocode first, with interactive buttons to reveal Step 1, Step 2, and Full Solution code progressively.
5. 🔮 Predict Output Before Run (Mental Simulation)
Challenges students to mentally trace execution and predict return values or terminal outputs under specific variable test cases before executing.
