Personalized Probing Puzzles (p³) test if students understand their own AI-generated code
New method p³ automatically generates puzzles to verify code authorship and understanding in minutes.
Researchers developed Personalized Probing Puzzles (p³) to assess whether students truly understand code they submit, especially given AI coding agents. In a pilot study with a graduate cloud computing course, p³ auto-generated, asynchronous puzzles that took minutes to complete. The method identified potential gaps in understanding, addressing academic integrity and professional accountability risks from AI-generated code. Future work is needed to correlate puzzle results with comprehension and embed p³ into professional code reviews.
- Method named Personalized Probing Puzzles (p³) auto-generates assessments from a student's own code artifacts.
- Pilot study in a graduate cloud computing course showed p³ can identify gaps in code understanding in minutes.
- Future plans include correlating puzzle results with comprehension metrics and integrating into professional code review workflows.
Why It Matters
Ensures accountability in software engineering as AI coding agents blur the line between human and machine authorship.