CS 240 Spring 2026 AI Retrospective

Academic integrity and AI Toward the end of the Spring 2026 semester, I found myself at the center of a storm involving use of "AI" in CS 240: Programming in C. My approach to addressing the issue should have been better and is primarily why the individuals that ran afoul of the clearly stated course policy ultimately incurred little to no consequence. I am writing this post largely due to the number of students that continue to approach me and mention how much misinformation and inaccuracy surrounds the, apparently, ongoing discussions regarding the incident. Policies and Expectations To begin, I wish to be abundantly clear: the Spring 2026 offering of CS 240 had a clearly articulated prohibition with regard to using AI/LLMs to solve any assignments in the course. You can find it in the syllabus here. Specifically: You may discuss assignments in a general way with other students, but you may not consult anyone else's work. Among other ways to get an F, you are guilty of academic dishonesty if: ... - You utilize ChatGPT or other software to programmatically generate solutions to any part of an assignment, quiz, or exam Moreover, the syllabus clearly states that we may not immediately address violations of the academic integrity policy: If we find reason to believe that a student or team has cheated on any assignment, we may inform the student or team promptly, or we may decide to silently accumulate evidence against the student or team on later assignments. Finally, it was even clearly communicated in lecture that reading and understanding the syllabus was each student's responsibility, as mentioned on the slide from Lecture 1 found below: I have been told some people may be implying that the expectations above were somehow a surprise to those enrolled in the course. Such a claim is highly specious. In addition to the syllabus, I went back and checked: this expectation was addressed repeatedly in at least five lectures. I have included some of the slides from those lectures below. Note that while some may not explicitly mention AI, my slides are a starting point for what I discuss in lecture and do not encompass everything that is said. I will add that I am quite certain that I discussed these expectations at other times but without a slide to explicitly reference as well. Here are the slides from Lecture 1 on January 12, 2026: On January 14, 2026 I discussed the policy again in Lecture 2. While the slide itself does not mention LLMs, I did take the opportunity to mention the policy verbally again. Lecture 3 on January 21 included yet another reminder... Another discussion occurred in Lecture 7 on February 4, 2026. It was at this time we were beginning to focus on specific identifiers during the development of Argus, the tool that would later be deployed. So, it was always the case that it was made abundantly clear to students that they should not be using AI/LLMs to complete assignments and that doing so was a violation of the course academic integrity policy. Moreover, our usage of MOSS to identify cheating incidents proceeded as in semesters prior. Throughout the entire semester, students identified as having highly similar code were confronted and, when appropriate, incurred consequences resulting from their violation of the above policy. Argus Another mistaken impression some seem to have formed is that we trained our own LLM to assist with identifying students that violated this policy. We did not. The tool developed - named Argus - utilizes static analysis to identify indicators for which it is highly unlikely a reasonable explanation exists with regard to their presence in a student's source code. Over the summer, we actually authored a paper regarding this tool. You can find a preprint version of it on arXiv or locally here. The paper was submitted to SIGCSE 2027. Though ultimately not selected for publication, the peer reviews from that process are available here. As discussed in the paper, we have also run the tool against earlier semesters with very compelling and interesting results. In particular, the indicators that we use are virtually nonexistent prior to 2024 and their presence increases rapidly over the following years. All data gathered and analyzed point to high accuracy with regard to the H-score generated. I am certain many would like an exhaustive list of the indicators that we identify. But, providing such would almost certainly reduce the utility of the tool. If you are an educator or someone with a legitimate interest in this information, please feel free to contact me ([email protected]). I am unopposed to providing it in situations where I know it will be properly protected and used. It is important to point out that Argus served only as a starting point. Each potential case identified was looked at by at least one person prior to any further action. Not all cases were pursued. In fact, we took a very conservative look at each case, only pursuing ones for which there was clear evidence and a lack of any reasonable explanation for the existence of that evidence. Ultimately we flagged 267 out of 584 identified students (roughly 45.7%). The Lead Up Argus reached a usable state around mid-March of 2026. We began discussing it during our weekly TA meetings around that time, and on March 23 I ultimately decided to form an "AI Academic Integrity" (AIAI) team to address the large number of potential AI usage cases. We began efforts in earnest to set up the necessary processes. Discussions continued after the formation of the AIAI team and, as we considered division of labor and preparation for meeting with hundreds of students, I was reminded of the approach historically taken by CS 159. In this course, instructors would email students identified as having violated course policy and offer them an opportunity to simply admit responsibility. One of these emails from Fall 2017 is included below. During our April 13, 2026 meeting I decided to adopt a similar approach utilizing a form that I would subsequently create with the hope that a self-reporting mechanism would cut down on the number of cases that would require further discussion. Concerns of potential academic integrity violations have been raised based on your submission for the fifth homework. The solution you submitted has been measured and determined to be highly similar to that of at least one other student in the course. The work you submit must be your own original effort and not the result of unacceptable, even if unintentional, collaboration. Although it may not be intentional, extensive collaboration with others may result in highly similar work that resembles copying. Every student is also responsible for protecting his/her own work. If you inadvertently allow your work to be accessed, you may still be held responsible for facilitating an academic integrity violation. Please take a moment to review the policies of the course found in your syllabus related to academic integrity. In this particular case, the high similarity between your assignment and others was identified by computer software and upon review cannot be dismissed by mere coincidence. As a result you will be assigned a score of zero for the assignment according to the academic integrity policy outlined in the course syllabus. If you are willing to accept these findings, please reply to this email by the stated deadline (see below) containing the following response: "I acknowledge that my assignment submission is in violation of the academic integrity policy of the course and accept the score of zero for this assignment. I have no intentions of doing this again in the future and realize that a second offense would result in my failing the class." After the above statement, you must include a short summary of the actions and names of others involved that resulted in this incident. There would then be no further repercussions as a result of this incident and a favorable report will be sent to the Office of the Dean of Students stating that you were cooperative and accepting of the outcome. The case would then be closed as far as this class is concerned, but the Dean of Students reserves the right to discuss this incident with you in the future and take additional disciplinary action if warranted. If you wish to contest these findings, you must meet with me during office hours on or before Tuesday November 14. Please be aware that denial of the incident and/or fabrication of the events would result in a less favorable summary to the Office of the Dean of Students if you are found to be in violation, and additional penalties could then be applied. Failing to respond to this request will be reported to the Office of the Dean of Students as being non-cooperative. I hope that this matter can be resolved as pleasantly as possible for all involved. Thank you for your time and prompt attention to this matter. Based on this email (and my use of similar ones in past offerings of CS 240 as well), I created the form found below. I will also note that this is where my use of the favorable/unfavorable language originated. Although some students also seemed at the time to be hung up on the possibility of suspension or expulsion, that is not something within an instructor's power nor is it something that typically happens for a first time offense. I did not mention either possibility in my communications or on the form, though I believe I briefly mentioned them in lecture as something the Dean of Students could consider. The Process On April 16, 2026, I sent out the following email with the subject "[CS 240] Academic Integrity Violation - RESPONSE REQUIRED": Through careful analysis and manual review, we have identified what we consider to be clear and concrete indicators in one or more of your homework assignment solutions that it was partially or entirely generated by an AI/LLM tool like ChatGPT, Claude, Copilot, etc. This is