Cheating Allegations Lead to Chaos at Purdue University

At Purdue University, the CS 240 is a notoriously difficult class. It requires students to write a great deal of code, and it’s a course that many students struggle with.

However, this semester’s class has created even more consternation than usual.

On or about April 15, the class’s professor, Jeffrey Turkstra, sent an email to more than 200 students accusing them of using AI to cheat on their assignments, as reported by the Journal & Courier.

The email, to put it mildly, was blunt. It said that there were “clear and concrete indicators” of AI usage, and students were given five days to fill out an online form to explain what assignments they had used AI on.

The timing of the email also raised eyebrows. Turkstra sent the email just before the deadline for students to drop the class with a “W” on their transcript. Given that Turkstra threatened to fail students who didn’t respond or failed to disclose what assignments they used AI on, over half of the accused opted to drop the class.

However, theoretically, even those students wouldn’t have been exempt from punishment. He said that those who failed to respond and dropped the class would “result in an unfavorable letter being sent to the Dean of Students.”

The story exploded, in particular on the Purdue subreddit, which quickly turned the story into memes and jokes. Ultimately, cooler heads prevailed. The allegations were dropped, and students were allowed to re-enroll in the course.

However, all of that came after a very tense few days and a class session that was overflowing with students eager to learn about the situation. During that class, Turkstra said, “I want you guys to graduate from Purdue. I want you to leave this course as good programmers and good computer scientists. If you’re using ChatGPT, that’s not going to happen.”

As Purdue’s independent student newspaper, The Purdue Exponent, reported, this has left many of the honest students with questions. Turkstra said, “We have students who have invested a lot of effort, have not cheated, and now get to watch some of your classmates leave with no consequences. I don’t have an answer to that.”

However, neither Turkstra nor the school had much choice. The situation had gotten so out of hand that it was virtually impossible for them to hand out punishments in a fair and equitable manner.

Where Things Went Wrong

Setting aside the issues of AI detection and the inherent challenge of trying to determine if a student used AI on an assignment, a lot was done very poorly in this case.

The biggest was that the new detection method Turkstra used was only applied after all but two of the assignments had been graded.

While testing a new AI detection method on previously graded assignments is a common practice and a good idea, using it to retroactively punish students is not. The students neither consented to the use of the technology nor were they given a chance to see how their work would be processed by it.

If you are going to punish students for using AI, the goal should be to detect any issues as early as possible and minimize the harm to the students. Springing a tool like this on students late in the semester feels more like a “gotcha” than a fair and equitable process.

Think about how much different this conversation would have been if we were only talking about the first assignment and not an entire semester’s worth of work. It would still be a serious issue and a difficult conversation, but it wouldn’t have been as life-or-death as it became later in the semester.

As frustrating as it is, I think the school and Turkstra made the right decision to back down. The findings, as presented, were unverifiable and inevitably would have punished honest students alongside cheaters.

To put it simply, it’s almost certain that some of the students accused did cheat. However, it’s equally certain that not all of them did. There’s simply no way to separate those two groups this late in the semester and with so many works at stake.

Some Buried Ledes

Amidst the chaos, confusion and controversy, there were some good things that came out of this story.

Particularly interesting is Turkstra’s own data.

According to an “ad hoc analysis” he performed, he found that AI performed significantly worse than human students on the assignments. The difference was between 10 and 15 percentage points.

To be clear, we don’t know much about how Turkstra conducted this analysis and concluded which assignments were AI-generated. So what this essentially says is that the papers his system flagged as AI-generated performed worse than the ones that weren’t.

Still, even if his detection method is only somewhat accurate, it still may bode well for honest students in the classroom.

The second lede is the dynamic between Purdue, a school that has largely embraced AI, and this story. Turkstra repeatedly stated that the use of AI in the course was problematic because it is a foundational course. While he embraces the use of AI more broadly, it’s important to build foundational skills.

This highlights a tension that universities are facing all over the world. On the one hand, they recognize that AI is here to stay and will likely be a skill students need to learn. On the other, there are still times and places where using AI is inappropriate.

Drawing those lines and then enforcing them will be difficult. But the story does recognize that AI can be useful, it just isn’t a replacement for human knowledge and understanding.

Bottom Line

My advice to Turkstra would be to simply try again next semester. Take the summer and use that time to test the effectiveness of your AI-detection tools, create processes to ensure honest students are not punished, and figure out how to implement them with the resources you have.

I would also say that it might be time to reevaluate how you handle assessment. It’s a chance to ask tough questions about not just the assignments you give, but how they are administered. Are there ways you could mitigate or reduce AI usage by making structural changes to the class?

All of this is easier said than done. But I know that Turkstra is not alone in this.

Ultimately, I think his goal of keeping AI out of a foundational class is a noble one. He’s right to be worried and even right to be angry on behalf of the honest students.

However, the implementation and timing were simply a mess. This was not the right way to approach this issue, and it was far too late in the semester.

Still, a more thoughtful approach to this could help ensure that future students get the skills that they need to succeed and keep the integrity of the class intact.

If any good comes of this, it will be the conversation around it. To that end, people are definitely talking about it. I just hope that the conversation steers into a more productive direction.