The (Honest) Truth About Dishonesty

What do you think happened? In the question condition, our participants claimed to have solved an average of ten matrices—about three more matrices than in the control condition (which means they did cheat) but by about two fewer matrices than in the shredder condition and by five fewer than in the Madoff condition. After observing the experimenter telling David that he could do what he wanted, cheating actually decreased. That was the opposite of what would have happened if our participants had engaged solely in a rational cost-benefit analysis. Moreover, this result suggests that when we become aware of the possibility of immoral behavior, we reflect on our own morality (similar to the Ten Commandments and the honor code experiments in chapter 2, “Fun with the Fudge Factor”). And as a consequence, we behave more honestly.

A Fashion Statement

Although those results were promising, we still wanted to get more direct support and evidence for the idea that cheating might be socially contagious. So we decided to go into the fashion business. Well, sort of.

The structure of our next experiment was the same as in the Madoff condition: our actor stood up a few seconds into the experiment and announced that he had solved everything and so forth. But this time there was one fashion-related difference: the actor wore a University of Pittsburgh sweatshirt.

Let me explain. Pittsburgh has two world-class universities, the University of Pittsburgh (UPitt) and Carnegie Mellon University (CMU). Like many institutions of higher learning that are in close proximity, these two have a long-standing rivalry. This competitive spirit was just what we needed to further test our cheating-as-a-social-contagion hypothesis.

We conducted all of these experiments at Carnegie Mellon University, and all our participants were Carnegie Mellon students. In the basic Madoff condition, David had worn just a plain T-shirt and jeans and had therefore been assumed to be a Carnegie Mellon student, just like all the other participants. But in our new condition, which we named the “outsider-Madoff condition,” David wore a blue-and-gold UPitt sweatshirt. This signaled to the other students that he was an outsider—a UPitt student—and not part of their social group; in fact, he belonged to a rival group.

The logic of this condition was similar to the logic of the question condition. We reasoned that if the increased cheating we observed in the Madoff condition was due to the realization that if David could cheat and get away with it, so could the other participants, and it would not matter if David was dressed as a CMU or a UPitt student. After all, the information that there were no negative consequences to egregious cheating was the same regardless of his outfit. On the other hand, if the increase in cheating in the Madoff condition was due to an emerging social norm that revealed to our participants that cheating was acceptable in their social group, this influence would operate only when our actor was part of their in-group (a Carnegie Mellon student) and not when he was a member of another, rival group (a UPitt student). The crucial element in this design, therefore, was the social link connecting David to the other participants: when he was dressed in a UPitt sweatshirt, would the CMU students continue to play copycat, or would they resist his influence?

To recap the results so far, here’s what we saw: When cheating was possible in the shredder condition but not publicized by David, students claimed to have solved, on average, twelve matrices—five more than they did in the control condition. When David stood up wearing regular CMU attire in the Madoff condition, the participants claimed to have solved about fifteen matrices. When David asked a question about the possibility of cheating and he was assured that it was possible, participants claimed to have solved only ten matrices. And finally, in the outsider-Madoff condition (when David wore a UPitt sweatshirt), the students observing him cheat, claimed to have solved only nine matrices. They still cheated relative to the control condition (by about two matrices), but they cheated by about six fewer matrices than when David was assumed to be a part of their CMU social group.

Here’s how our results looked:

Together, these results show not only that cheating is common but that it is infectious and can be increased by observing the bad behavior of others around us. Specifically, it seems that the social forces around us work in two different ways: When the cheater is part of our social group, we identify with that person and, as a consequence, feel that cheating is more socially acceptable. But when the person cheating is an outsider, it is harder to justify our misbehavior, and we become more ethical out of a desire to distance ourselves from that immoral person and from that other (much less moral) out-group.