A professor teaching CS 240 (Programming in C) at Purdue University has reversed the decision to retroactively apply AI-based detection findings to earlier homework submissions. The decision follows a period of controversy surrounding the use of the Argus tool to identify students suspected of using Large Language Models (LLMs) to solve assignments.
Purdue Professor Reverses Retroactive AI Cheating Penalties Amid Controversy
The instructor stated that the decision to discard retroactive findings was made after discussions with university leadership, including the Dean. Key concerns raised during the process included the potential for innocent students to feel coerced due to the timing of the investigations, which occurred near the course drop deadline. Additionally, concerns were raised regarding the format of self-reporting forms and the challenges of ensuring due process as the semester approached its end.
The instructor noted that data from ad-hoc analysis and subsequent research showed a significant performance gap between students identified as using LLMs and those who were not. In one instance, the difference in midterm exam scores was 10.75% on the first exam and 14.5% on the second.
While retroactive penalties were rescinded, the Argus tool will continue to be used for subsequent assignments. Out of 144 students who initially dropped the course following the incident, approximately half (74 students) have since re-enrolled.
Sources
- CS240 AI Cheating Retrospective （Hacker News Frontpage, 2026-09-30）
- on arXiv