Tutor Meeting - LucasWolke/code-tutor GitHub Wiki

Feedback

AI

  • AI suggests concepts like Stack even when only Arrays should be used → limit to relevant concepts based on context.
  • AI should guide students toward official Java documentation instead of answering all Java-related questions directly.
  • Implement an AI layer that connects students to documentation to reduce hallucinations and promote better habits.
  • Ensure AI suggestions are appropriate for the student's level; professor-provided context should guide suggestions.
  • The AI should be able to answer "Am I done?" to help students know when they've completed a task.
  • AI feedback loop should end eventually; it should not keep suggesting endless improvements.
  • The AI can be misled into non-relevant topics (e.g., becoming a cooking tutor).

Usability

  • Input gets deleted when switching between tasks → should persist user input.
  • Students need clearer answers to task-specific questions to understand what to do.
  • AI responses should be markdown formatted for better readability.
  • Include test cases to automatically check the code instead of just running it.
  • Professors should be able to add metadata to tasks, such as hints or relevant links (e.g., course slides).
  • Add a button group with pre-defined prompts (like Gemini in Google Docs), e.g.:
    • "What is this tool about?"
    • "How is this different from ChatGPT/Gemini?"
    • "Any tips for me?"
    • "How do I break this problem down?"

EP1

  • Prevent 'Content Strategy' tutor from suggesting strategies above the student's level.
  • Context provided by the professor should influence AI suggestions.
  • Provide a dedicated section that explains the learning goals and expectations for EP1.

UI

  • Some UI elements (e.g. "Tutor", "Java Editor") have poor readability.
  • Text hierarchy is unclear — instructions/settings too large, logo and important labels too small.
  • Improve visual clarity so users know where to focus at first glance.

Tutor Meeting

Goal of the Session

  • Simulate the experience of a student using the Coding Tutor IDE.
  • Assess the AI assistant's usefulness, accuracy, and clarity.
  • Provide feedback on technical functionality, user experience, and overall educational value.

Acting Like a Student

Tutors should replicate typical student behavior and explore common use cases.

  • Ask Common Student Questions
    Think of questions you often hear from students and ask those to the AI. Examples:

    • Help in understanding the problem
    • Debugging help
    • Code explanations
    • Syntax or logic questions
  • Test Different Levels
    (The AI Tutor responds on different levels - based on the students needs)

    • Does the tutor respond with the correct level?
    • Are the levels meaningful and helpful?
  • Gaming the system

    • Does the tutor respond with too much help like code snippets?
  • Evaluate UI/UX

    • Is it intuitive?
    • Does it look good?
    • Does writing code/asking questions feel engaging and fun?
  • Comment on System Design Short feedback on my implementation, ideas on how to improve it.