Alex Tabarrok writes over at Marginal Revolution:

I sometimes find evidence of cheating on exams but I rarely take action, I don’t have to. Almost invariably the cheaters get abysmally low grades even without penalty. Some people I know get annoyed when students without evident handicap ask for and receive special treatment such as extra time on exams. I comply without rancor as the extra time never seems to help. Over the years I have had a number of students ask for incompletes. None have ever become completes.

I call this the law of below averages.

Unfortunately, those same people asking for incompletes go on to get jobs in the corporate world where they ask for exceptions to security policies with a promise that they’ll “remediate after go-live.”

The comments are also priceless and detail one example after another of people attempting to exploit operational weaknesses in academic processes. I’d like to look at a couple of different threat models for analyzing the risk of academic cheating.

For the cheater, I assume they have chosen to accept the risk of getting caught, whatever that may be. When I was at university, the penalty for cheating was receiving no credit on the activity you were caught cheating on. That meant that there was a direct correlation between the potential benefit of cheating (getting, say, an A on the final and thereby improving their letter grade from failing to passing) and the impact of getting caught (getting no credit for the final, probably producing a failing grade in the class). Of course, if you were already going to fail, that effectively meant that there was no down side, thus creating an incentive to cheat.

Even in that case, I thought that relying on the opportunity to cheat even existing was a poor risk decision on the part of the cheater.

I’ll consider a couple of different scenarios for cheating in an exam. Cheating on papers is a whole different issue and has much more to do with the amount of effort the grader is willing to put into it than anything else, thanks to the ease with which Google and a halfway diligent grader can catch all but the most careful plagarists.

Start with the good ol’ Cheat Sheet, whereby the cheater sneaks information into the exam for reference.

First, some sort of mechanism has to be developed for getting the knowledge into the exam without detection, then the course material must be analyzed to determine what subset of it would be most useful.

Secondly, by the time you’ve performed that selection process and prepared the material, you’ve probably spent about as much time as it would have taken to just learn the material.

Third (in my case at least), the reason that I might have needed to cheat was because I had been ditching class, meaning that I wouldn’t have the material to produce the cheat sheet anyway.

So much for Cheat Sheets.

Next is copying answers from someone else.

Consider that in nearly every exam I ever took, the seating was designed to prevent copying, either by spacing out the test takers and/or by using multiple test forms. The academic testing environment definitely is a long-running cat & mouse game between cheaters and proctors. This makes depending on the opportunity to copy a bad risk from the start.

Then, as some of Alex’s commenters relate, the person cheating may choose to copy off someone even less knowledgable than they are:

some years ago, when I was a math TA, another math TA came to me to ask how to handle this situation: one guy had copied from another on a calculus exam. Thing is, the cheater had picked his target poorly: he was the 2nd-worst student in the class. And, of course, the cheater didn’t have enough time to copy everything.

Basically, ignoring the cheating and grading the exam like all others, the cheater’s exam got an F and the target of the cheater got a D. I told my friend that ist seemed the problem took care of itself.

Finally, it was always my experience that the people one would most want to copy off of were aware of their desirability and, expecially if the class was graded On The Curve, fiercely protected their work to maximize their outlier status.

As the number of variables mount, the odds of success start to get very long very quickly. Once again, not a good risk.

Finally, the would-be cheater should take a long, hard look in the mirror and consider that if they can’t be bothered to do a half-decent job of keeping up with their studies, they probably aren’t going to do a very good job of cheating, either.

Of course, since I studied political science there wasn’t much point in trying to cheat on exams. They were all essays and you could either produce a coherent argument or you couldn’t.

No amount of cheating in the world is going to help you explain what you think is the US Supreme Court’s worst decision ever and why if you don’t know the details of one you think was crap. (I chose Katzenbach v. McClung, because I thought it was a horrible cop-out to hide behind the commerce clause when the 14th Amendment existed specifically to address the right to equal protection under the law. If I had to answer that question today, I think I’d have to go with Bush v. Gore, but that’s a whole different ball of yarn.)

To address some of your points, I say: All exams should be open book and material. If it helps, bring in an encyclopedia. All written material should be acceptable EXCEPT the material of the people sitting next to you.

1. dispense with the tiniest writing on the tiniest piece of paper competitions (for legit or illegit cheat sheets.)
2. profs will have to make more of an effort thinking up _good_ exam questions instead of dumbing them down.
3. minimal reliance on memory, with maximal use of comprehension of the material… the exam questions, undoubtedly will be made such that this is true.
4. i wished i could bring a dictionary to every exam I took… (I was a non native English speaker in a US university.)

I always did better on open-material exams because I wasn’t stressed and was confident in my understanding, rather than cramming to memorize equations or dates the night before. I spent a good time planning which books to bring.
Also, I would not limit the time for exams (within reason); for non-science disciplines I would only limit the AMOUNT of writing to avoid a core dump of all-inclusive crap.