Friday, March 18, 2011


Having tackled lying here, here, and here, the obvious next target is cheating, with stealing to come.

By cheating, I mean breaking a rule to one's advantage, be it peeking at someone else's exam or hand of cards. Cheating is supposed to be fundamentally wrong, whether in the classroom or the card game.

As an example, suppose you're taking an exam that will be strictly curved to 10 A's, 10 B's, 10 C's. Then the sole purpose of the exam is to rank the students relative to each other. Why is cheating a no-no? Because it could potentially disrupt the true ranking, subverting the entire point of the exam. Fair enough.

Of course, if everyone else is cheating...

If you've been following along previously, you probably know where this is going. To make the point cleanly, suppose you have been told in advance that for question #3, you will have to write out a randomly chosen line of some particular long poem, from memory. It's more trouble than it's worth, and you know that everyone else is just sneaking a copy of the entire poem into the test. The teacher is unobservant and certainly won't catch anyone. Do you follow suit, or forfeit the question? Obviously sneaking in the poem is cheating...but if everyone else is doing it, then actually you preserve the ranking by cheating too. Which is a good thing, right?

Following the rules instead of the objective that generated them does not an angel make, although in this case it might make an unfortunate martyr. In this case, the real point of the exam is to accurately rank students, not to have students diligently follow rules governing their conduct.

Now let me be clear that while most cheating situations are far more complicated than the above, they nevertheless often contain the effect I've just identified as a component, which means you cannot disappear this problem by adding on layers of complexity. Sure, cheating is typically not a binary decision -- how much to cheat? And when only some people are cheating, or cheating by different amounts, then cheating yourself corrects for the imbalance between you and the cheaters, while potentially creating an imbalance between you and the non-cheaters. What to do then? OMG moral ambiguity! It can get messy, but the answer is not that we get to ignore the above effect just because it is no longer clearly dominates. Sometimes we have to take a stand; the answer will depend a lot on the particulars of the situation, and the objective chosen -- yes, a judgment call.

Take your "never cheat" if you want. If you're after simple rules instead of judgment calls, you might also consider only flinching when the brick is definitely positively going to hit you in the face.


"That's all fine and well as a point of theory, but as a practical matter, when have I ever actually been in a situation where cheating on an exam was the right thing to do?"

Yeah, I haven't gone through life righteously cheating on exams either. But maybe it has more to do with the situations I've been in, and not so much a fundamental drive to do what society says is the "honorable" thing. Maybe if I was in the bottom decile in high school, competing with the people around me to not be the one who fails, and everyone else in my decile was cheating, and the risk of getting caught wasn't too high...well it could easily be not just selfishly optimal, but in fact even socially optimal for me to cheat. Is academic honesty about fairness? What could possibly feel less fair than me failing because everyone around me cheated?

Therefore do not be so quick to assume that in practice cheating is pretty much always wrong just because it has been for you, and do not be so eager to judge the cheating of people who have found themselves in very different situations from yours.

Do you feel differently? Do you feel like you would never cheat? Should never cheat? Something to chew on.


NB: I can think of two important objections to this and the Rules versus objectives post more generally. First, as a practical matter, simple rules can perhaps be ingrained into behavior more readily than more complicated ones, and may actually be a more effective channel to align individual behavior with socially optimal behavior. The other objection has to do with equilibrium selection in what is really a repeated game. These are topics for another time though, and are more caveats than contradictions to the above content.


  1. I have yet to read your posts on lying, but I imagine you present a similar scenario. Is it always wrong to lie? No. There are plenty of situations were lying is in fact the right thing to do. For some strange reason, the cheating issue seems to be different, although I'm not sure why. I'll have to ruminate over this one for a while.

    I think the real problem with an exam that has a sole purpose of ranking students relatively is that it creates the "smart for one, dumb for all" problem. There is an incentive for one person to cheat to improve their relative position, but if everyone cheats then what's the point (no one's relative position changes). I believe Robert Frank described a version of this scenario in his book Luxury Fever. I'm often amazed at the "smart for one, dumb for all" phenomenon in the workplace. It's smart for one person to work late into the evening to stand out relative to their peers, but when everyone does it no one is better off relatively and everyone is worse off in absolute terms.

  2. Greg,

    See second cheating post for more on this.

    You're imagining the (often realistic) scenario where everyone cares about their own test score, so they cheat to improve it, but collectively it gets them nowhere. In your situation, each individual has a dominant strategy (cheat) which drives them unequivocally to the equilibrium where everyone cheats. That's fine, and very important -- as rows of fans standing uncomfortably at the football game will testify to -- but it's actually not what's driving my argument. In my argument, we are able to get these results even *assuming* people care about social welfare. In this world, there are actually 2 equilibria, "all cheat" and "no cheat," and they are both socially optimal given everyone else's behavior.