Tuesday, March 29, 2005

Undergraduate cheating

It's been a long time since I linked to a MeFi post. Oh, here's one.

Actually, the real reason I'm posting this is that I want to mention one salient fact that, oddly, none of the 200-odd comments mention: at most universities, even when a student is caught cheating in the most blatant and silly fashion, it is pretty hard for the instructor to convince the administration to punish the student harshly. Usually, the penalty for cheating is a slap on the wrist --- a zero grade on the one cheating assignment (not even the whole course!), perhaps a mark in the departmental file (which future employers usually do not see). Cheating is so pervasive in most places that when I hear about undergrads cheating, I'm not surprised that they're cheating; I'm surprised that they were caught. The net result is that it is quite rational for a student to cheat, because the expected utility of cheating is high. Given this environment, almost any escalation of negative consequences for cheaters is probably an improvement.

In fact, let's analyze the cheating problem from a Law and Economics point of view. Under a L&E analysis, the aim of law enforcement should be to maximize total social welfare, which means, roughly, minimizing enforcement costs while maximizing compliance. Now, model the student as a rational actor that tries to get the maximum expected grade with the minimum effort (yes, this is a pretty insulting view of students, but there definitely exists a minority of students who take exactly this attitude), and calculates the expected grade by multiplying the probability of getting caught by the grade demerit for cheating, yielding the expected grade demerit, and subtracting that from a B+. It's clear, then, that there are exactly two ways of getting students not to cheat:

  1. Increase the probability of getting caught.
  2. Increase the penalty for cheating.

Now, the ideal would be to increase the probability of getting caught to 1, in which case the penalty for cheating could be relatively small --- the student would see negative expected benefit from cheating, even if the penalty were trivial (like getting a zero on the cheated-upon assignment). Of course, that's impossible. The next best thing would be to increase the probability of getting caught to some figure "sufficiently high" as to create a perception of "reliable" enforcement. But --- and this is the key point --- reliably catching the form of cheating that Pahl engaged in would incur extremely high enforcement costs.

The dumb student who copies-and-pastes a few incongruously out-of-place paragraphs off a Google search can be caught with relatively little effort (although if you multiply that effort by 100 papers, it still adds up to a large amount of time that the instructor and teaching assistants can't spend on improving the quality of instruction). But how do you catch somebody who hires somebody else to write a paper? Pahl chose a particularly stupid way to go about hiring a surrogate writer, but online paper mills have institutionalized the service. Presumably, over time, the relative quality of these services will get sorted out, and that knowledge disseminated through undergrad folklore. The countermeasures that occur to me fall into two categories. First, one could co-opt the paper mills somehow: pay them off, either to send a copy of each paper they generate to the university, or to turn in students directly. Second, one could compute grades based solely on work that can be done while the students are under surveillance: grade based only on in-class exams, or move to an apprenticeship system where the master assesses student mastery through extensive one-on-one interaction. All such measures exact high costs from the university, in either cash, labor, or educational quality.

So the rational thing, from the university's point of view, is to choose the other option: increase the cheating penalty. This keeps enforcement costs low, and, in theory, has the same deterrent effect as increasing enforcement. Of course, one wouldn't want to institutionalize Kushner's methods exactly, but it seems to me that public shame might be an important element of an effective deterrent system.

On the other hand, if the probability of getting caught is too low, and the penalty too severe, then the whole business starts to look like an inverse lottery, at which point human psychology outweighs rational calculation, and students will probably cheat willy-nilly even though the expected return is negative. Basically, human beings have almost no intuitive grasp of very large numbers or very small fractions, so they can't rationally calculate in such circumstances. So maybe the entire analysis above is bunk. However, in that case, I think the only reasonable alternatives vastly increase the cost or reduce the quality of undergraduate education.

No comments:

Post a Comment