The (Honest) Truth About Dishonesty

HAVING ESTABLISHED ONE negative aspect of collaboration—that people are more dishonest when others, even strangers, can benefit from their cheating—we wanted to turn our experimental sights on a possible positive aspect of collaboration and see what would happen when team members watch each other. Imagine that you’re in a room with a few other participants, and you’re randomly paired up with someone you have never met before. As luck would have it, you’ve ended up with a friendly-looking young woman. Before you have a chance to talk to her, you have to complete the matrix task in complete silence. You are player 1, so you start first. You tear into the first matrix, then the second, and then the third. All the while, your partner watches your attempts, successes, and failures. When the five minutes are up, you silently put your pencil down and your partner picks hers up. She starts working on her matrix task while you observe her progress. When the time is up, you walk to the shredder together and shred your worksheets. Then you each write down your own score on the same slip of paper, combine the two numbers for your joint performance score, and walk over to the experimenter’s desk to collect your payment—all without saying a word to each other.

What level of cheating did we find? None at all. Despite the general inclination to cheat that we observe over and over, and despite the increase in the propensity to cheat when others can benefit from such actions, being closely supervised eliminated cheating altogether.

SO FAR, OUR experiments on cheating in groups showed two forces at play: altruistic tendencies get people to cheat more when their team members can benefit from their dishonesty, but direct supervision can reduce dishonesty and even eliminate it altogether. Given the coexistence of these two forces, the next question is: which force is more likely to overpower the other in more standard group interactions?

To answer this question, we needed to create an experimental setting that was more representative of how group members interact in a normal, day-to-day environment. You probably noticed that in the first two experiments, our participants didn’t really interact with each other, whereas in daily life, group discussion and friendly chatter are an essential and inherent part of group-based collaborations. Hoping to add this important social element to our experimental setup, we devised our next experiment. This time, participants were encouraged to talk to each other, get to know each other, and become friendly. We even gave them lists of questions that they could ask each other in order to break the ice. They then took turns monitoring each other while each of them solved the matrices.

Sadly, we found that cheating reared its ugly head when we added this social element to the mix. When both elements were in the mix, the participants reported that they correctly solved about four extra matrices. So whereas altruism can increase cheating and direct supervision can decrease it, altruistic cheating overpowers the supervisory effect when people are put together in a setting where they have a chance to socialize and be observed.

LONG-TERM RELATIONSHIPS