In Defense of "Little Code"
summary
Study of the classic rainfall sentinel study finds students taught a design-first, functional approach favor clean-first solutions and show much lower error rates compared with historical data where 70% single-loop solutions dominate; automated AST classifiers fail on this noisy variety so the team uses think-aloud lab studies to uncover misconceptions (students don’t understand sentinels, don’t see two accumulators as okay). They build a class-sourcing tool QuizIast and pair crowdsourced questions with machine learning to classify mental models and scale insights.
AI-written summary · about
chapters (23)
- 0:00Intro
- 0:13Kathy Fiedler
- 2:13Big Code Big Data
- 3:56Tuning a Product
- 5:12Computing Education
- 6:08Case Studies
- 6:35Rainfall
- 11:17Historical Performance
- 16:51Similarities
- 18:25First Takeaway
- 18:52Ongoing Project
- 20:10Problem
- 22:58Results
- 24:55Lesson
- 33:18Misconceptions
- 34:57How Learning Works
- 36:12Teaching Doesnt Mean You Understand How Education Works
- 39:27Data Thinking
- 41:43Data Centric Computing
- 46:31Conclusion
- 47:23They ask different questions
- 49:47Programming language design
- 52:45Whats at stake
related talks
39mBack to Basics: Cutting Through the AI-Generated Code Noise
44mClean Machine Learning Code: Practical Software Engineering...
35mDebugging the Mind: A Programmer's Guide to Critical Thinking
01h 01mDeclarative Programming and Education
43mMost Influential Paper - Don’t Underestimate the Importance of Experime...
30mCognitive Shortcuts: Models, Visualizations, Metaphors, and Other Lies
37mCreating correct and capable classifiers
33mThe Art of Teaching