On Cheating

People become more likely to lie or cheat when they see others lie or cheat, explains Dan Ariely:

…we gave participants 5 minutes to solve as many mathematical problems as possible (where they were instructed to find which two numbers out of 12 add up to 10).

In the control, where no cheating was allowed, the average student solved 7 problems, which gave them a pay off of $3.50 out of a maximum of $10 (if they solved all 20 problems). To see how witnessing and act of dishonesty would affect participants, we had one student—a confederate named David—stand up after only a minute and claim he’d solved all 20 matrices. The experimenter merely responded that in that case he could take his earnings and go. So how did the participants respond to this display when asked to self-report the number of matrices they solved? By cheating a whole lot: they claimed an average of 15 correct answers, more than twice the average score when cheating was not allowed.

But, when Dan ran the same experiment with one slight tweak, dishonesty decreased dramatically:

This time, instead of looking like all the other participants, who were students at Carnegie Mellon University, we had our confederate wear a sweatshirt that located him within a different social group. This time he was wearing a University of Pittsburgh sweatshirt (Carnegie Mellon’s neighboring and rival university). When the dishonest act was committed by a person from an out-group, we found that cheating decreased dramatically to the lowest level in all the experiments (participants claimed “only” 9 correct problems).

So if you want people to stop being dishonest, stop rewarding dishonesty. If this is impossible, segregate people into tranches, so that the effects of any dishonesty in one tranche will be least likely to spread to another.