An Interactive Learning Environment for Program Design
summary
An interactive tool called Rad guides students through the design recipe with step‑by‑step natural‑language prompts backed by an LLM and a custom GPT flow. It auto‑generates Scala data definitions and English variable names, skips some followups when students give compound answers, and leaves the testing step unimplemented. In a 42‑student study 22 complete designs succeed, 10 sessions break (4 user inputs, 6 system faults) and 10 are incomplete; 16 students report frustration with fine dialog granularity.
AI-written summary · about
chapters (14)
- 0:00Rad interactive learning env
- 1:50Motivation for Rad
- 3:16Rad learning challenges
- 5:27Rad implementation
- 6:33User study results
- 8:41Student feedback on dialogue
- 10:24Rad input flexibility
- 11:51Questionnaire results
- 12:36Future work for Rad
- 14:00Rad vs related systems
- 15:08Summary
- 15:30Q&A: Granularity adaptation
- 18:40Q&A: Class size guidance
- 20:11Q&A: Input extraction teaching
related talks
16mUser Interaction Models for Disambiguation in Programming by Example
58mAI, Determinism, Code Generation and Refactoring
19mFrom brief to live prototype using a design-to-code tool
25mQ&A #1
52mBuild a Prompt Learning Loop
01h 02mAI Effects on Research and Education: A Programming and Software Language Perspective
01h 15m10x Development: LLMs For the working Programmer
49mPrompt Darwinism: Evolutionary Engineering in the Age of LLMs