Can Guided Decomposition Help End-Users Write Larger Block-Based Programs? A Mobile Robot E
summary
A block-based mobile-robot editor with two-canvas guided decomposition and a simulator is built and tested on 92 MTurk participants; the claim is that domain-specific enforced structure helps scale block programs. Compared to a conventional block editor, guided users reach higher success rates (task 2: 91% vs 71%, task 3: 83% vs 54%), finish no slower, but still often duplicate rather than reuse code and sometimes over-decompose into many tiny functions.
AI-written summary · about
chapters (8)
- 0:00Guided decomposition intro
- 10:33Block-based end-user decomposition
- 11:38Mobile robot domain
- 15:31Study design: guided vs traditional
- 19:01Traditional environment results
- 22:36Guided decomposition outcomes
- 25:04Conclusion on effectiveness
- 26:47Q&A
related talks
14mGang up on difficult problems with Mob Programming
18mGrounded Copilot: How Programmers Interact with Code-Generating Models
13mGetting Grammars into Shape for Block-based Editors
31mMob Programming: Team-Building Included
09mSituated Live Programming for Human-Robot Collaboration
18mRestructuring Structure Editing
48mMob Programming: A whole team approach
12mMOB Programming as Equilibrium