Game Theory and Cheating

A controversial blog post from an NYU Stern School professor has been circulating. The original has been removed but it is available here. From reading it, I can’t understand why the original was taken down. It is a solid tale and, in the end, the cheaters were caught.

Let me recap the story. Professor offers assignment to class that is the same as previous years. This is the first year the Professor has had easy access to Turnitin; a copying detection tool which is extremely effective. This is also the first year the Professor has tenure (but I think that is secondary). The Professor first notices lots of students just coping and pasting from the Internet (direct plagiarism). He decides to admonish rather than punish. Then he discovers copying of past assignments. Much hilarity ensues as the offenders try to talk themselves out of it (using the old “mixed up files from roommate trick”) and then — and this is the good part — copies the same assignment again when correcting the “error.”

I want to pause here to reflect on what is going on. First, this kind of cheating is not new. I suspect there are long-standing institutions within most universities that trade in past assignments. What is new is the ease of that process coming at the same time as a step-up in easy of detection. The problem is that the detection and enforcement is still not often used so students, playing out an equilibrium, cheat. So solid is this equilibrium that students do not appear to revise their beliefs when cheating is detected — forecasting (correct for some equilibrium) that the Professor won’t bother checking again. In this particular case, the students were wrong about the Professor’s behaviour.

But they weren’t too far off. The post is about the Professor telling the world that this was personally too costly and he wouldn’t do it again. And it was personally costly — right down to reduced teaching evaluations and the consequences of that. Detecting any crime is a public good and we expect too little of it to be done. The problem is that we end up in an equilibrium with too much cheating and too little enforcement.

The Professor actually did do the right thing and that was change the next assignment. He didn’t do it in a major way, however, which would have killed his problem dead. Instead, he did it in a sneaky way that — on my reading — was almost designed to catch up more cheaters. And it did. And hilarity ensued. And the Professor met with the costs of that including those flowing from a clever “turn yourself in” mechanism.

After all this, the Professor reflected on what he could do and decided that he could modify how he assessed students. But as he pointed out, it is not always a good idea to change assignments year on year and some assignments work and are valuable learning tools. My point here is that the Professor shouldn’t have to change his behaviour to minimise cheating. It is a public good for the School and the School should solve it.

So what should the administration do to change the equilibrium? First, take the enforcement of cheating out of the Professor’s hands. Employ a policing person who is the one who receives the Turnitin results and decides what to do about it. Make it their job and you solve the free-riding problem right away. You also allow cross-course, repeat offender detection. And you change the cheating equilibrium to one where enforcement is expected. Finally, the Professor has their hands clean and so suffers no conflict over teaching evaluations.

Actually, I wrote ‘first’ here, but that is The Solution.

Now, for those academics whose School does not employ The Solution, here is how I do it. You have to learn to love the whole forensic grading process. I love it the same way I love dreaming up imaginative and ironic punishments for my children. Nothing pleases me more than to detect cheating and to find ways of “dealing” with it.

Let me give you an example from, now 25 years ago. I was tutoring first year undergrads at the University of Queensland on the History of Economic Thought. I set an assignment for the students on Karl Marx — something I had enjoyed reading about myself when I was in High School. Two instances of cheating turned up. The first was identical assignments handed in by two students. Examining them, it was clear who had cheated — the one who had corrected the typos. Anyhow, the assignments had two different tutors — myself and another. We deployed the “confront individually” strategy and weeded out a confession. It turned out that the non-cheater’s was completely oblivious to what had happened as it was lifted from a common computer in their college. I recall that I tortured the cheater — my usual approach is a grade of 0 and you have to still write and pass the assignment to pass the course.

The second one was better. I read an assignment that had a ring of familiarity. It turned out that the reason for that is that it was mine. The very assignment I had written in High School. The student had gone to the same High School and, somehow, had a copy of my assignment. I have no idea how. Anyhow, they were not sensible enough to actually have checked the names of the assignment’s author and their tutor. They were punished but I had to admit, at least they had taste. It is one thing to copy a bad assignment. Copying a clearly excellent one demonstrates some feel for the material of the course.

Anyhow, let that be a message to all my future students. I enjoy detecting cheating and punishing cheaters. I let all my students know it. But, sadly, for some reason, I don’t get that many cases to play with.

4 thoughts on “Game Theory and Cheating”

  1. For an honours level econometrics course, the first assignment was practically identical to that of last year. Everyone had the code and solutions. Good marks for everyone.

    Second assignment,  everyone tries the same approach. code works but sampling algorithm gets stuck and people leave their computers running over night, all to no avail. When approached with this problem, wily old prof smiles and says “maybe there is something wrong with *your* code”

    For the rest of the semester, no one looked at last years answers. 


  2. The main issue i’ve found with Turnitin is that if I submit the assignement first to check that I’m not going to be wrongly charged with plagerism and then submit it to my lecturer It gets assessed as being 98% copied. I’ve conferred with lecturers and if them get a 98% they just ignore it encouraging complete theft, or submission of the assignment with plagerism to turnitin prior to submission


  3. W is unto an important issue: Cheating can easily be avoided, or at least be made quite difficult, by lecturers that do not repeat last semester’s/year’s assignments. Of course, the reality of what goes for teaching in most institutions of higher learning gives lecturers all the incentives to go the path of least resistance. In other words, this is an institutional problem.


  4. Andreas,
    “the reality of what goes for teaching in most institutions of higher learning”
    You sound like a man who didnt enjoy his first semester in the trenches of Australian university education!


Comments are closed.

%d bloggers like this: