With the results for both of these conditions, we could compare the performance in the control condition, in which cheating was impossible, to the reported performance in the shredder condition, in which cheating was possible. If the scores were the same, we would conclude that no cheating had occurred. But if we saw that, statistically speaking, people performed “better” in the shredder condition, then we could conclude that our participants overreported their performance (cheated) when they had the opportunity to shred the evidence. And the degree of this group’s cheating would be the difference in the number of matrices they claimed to have solved correctly above and beyond the number of matrices participants actually solved correctly in the control condition.
Perhaps somewhat unsurprisingly, we found that given the opportunity, many people did fudge their score. In the control condition, participants solved on average four out of the twenty matrices. Participants in the shredder condition claimed to have solved an average of six—two more than in the control condition. And this overall increase did not result from a few individuals who claimed to solve a lot more matrices, but from lots of people who cheated by just a little bit.
More Money, More Cheating?
With this basic quantification of dishonesty under our belts, Nina, On, and I were ready to investigate what forces motivate people to cheat more and less. The SMORC tells us that people
Before I divulge the answer, I want to tell you about a related experiment. This time, rather than taking the matrix test themselves, we asked another group of participants to guess how many answers those in the shredder condition would claim to solve correctly at each level of payment. Their predictions were that the claims of correctly solved matrices would increase as the amount of money went up. Essentially, their intuitive theory was the same as the premise of the SMORC. But they were wrong. It turned out that when we looked at the magnitude of cheating, our participants added two questions to their scores on average, regardless of the amount of money they could make per question. In fact, the amount of cheating was slightly
Why wouldn’t the level of cheating increase with the amount of money offered? Why was cheating slightly lower at the highest level of payment? This insensitivity to the amount of reward suggests that dishonesty is most likely not an outcome of a cost-benefit analysis. If it were, the increase in the benefit (the amount of money offered) would lead to more cheating. And why was the level of cheating lowest when the payment was greatest? I suspect that when the amount of money that the participants could make per question was $10, it was harder for them to cheat and still feel good about their own sense of integrity (we will come back to this later). At $10 per matrix, we’re not talking about cheating on the level of, say, taking a pencil from the office. It’s more akin to taking several boxes of pens, a stapler, and a ream of printer paper, which is much more difficult to ignore or rationalize.
To Catch a Thief
Our next experiment looked at what might happen if participants felt that there was a higher probability of getting caught cheating. Basically, we inserted the mental equivalent of a partially operating security camera into the experiment.
We asked one group of participants to shred one half of their worksheet—which meant that if they were dishonest, we might find some evidence of it. We asked a second group to shred the whole work sheet, meaning that they could get off scot-free. Finally, we asked a third group to shred the whole worksheet, leave the testing room, and pay themselves from a sizable bowl of money filled with more than $100 in small bills and coins. In this self-paying condition, participants could not only cheat and get away with it, but they could also help themselves to a lot of extra cash.
Again, we asked a different group to predict how many questions, on average, participants would claim to solve correctly in each condition. Once again, they predicted that the human tendency for dishonesty would follow the SMORC and that participants would claim to solve more matrices as the probability of getting caught decreased.
What did we find? Once again, lots of people cheated, but just by a bit, and the level of cheating was the same across all three conditions (shredding half, shredding all, shredding all and self-paying).
NOW, YOU MIGHT wonder if the participants in our experiments really believed that in our experimental setting, they could cheat and not get caught. To make it clear that this was indeed the case, Racheli Barkan (a professor at Ben-Gurion University of the Negev), Eynav Maharabani (a master’s candidate working with Racheli), and I carried out another study where either Eynav or a different research assistant, Tali, proctored the experiment. Eynav and Tali were similar in many ways—but Eynav is noticeably blind, which meant that it was easier to cheat when she was in charge. When it was time to pay themselves from the pile of money that was placed on the table in front of the experimenter, participants could grab as much of the cash as they wanted and Eynav would not be able to see them do so.
So did they cheat Eynav to a greater degree? They still took a bit more money than they deserved, but they cheated just as much when Tali supervised the experiments as they did when Eynav was in charge.
These results suggest that the probability of getting caught doesn’t have a substantial influence on the amount of cheating. Of course, I am not arguing that people are entirely uninfluenced by the likelihood of being caught— after all, no one is going to steal a car when a policeman is standing nearby—but the results show that getting caught does not have as great an influence as we tend to expect, and it certainly did not play a role in our experiments.
YOU MIGHT BE wondering whether the participants in our experiments were using the following logic: “If I cheat by only a few questions, no one will suspect me. But if I cheat by more than a small amount, it may raise suspicion and someone might question me about it.”
We tested this idea in our next experiment. This time, we told half of the participants that the average student in this experiment solves about four matrices (which was true). We told the other half that the average student solves about eight matrices. Why did we do this? Because if the level of cheating is based on the desire to avoid standing out, then our participants would cheat in both conditions by a few matrices beyond what they believed was the average performance (meaning that they would claim to solve around six matrices when they thought the average was four and about ten matrices when they thought the average was eight).
So how did our participants behave when they expected others to solve more matrices? They were not influenced even to a small degree by this knowledge. They cheated by about two extra answers (they solved four and reported that they had solved six) regardless of whether they thought that others solved on average four or eight matrices.
This result suggests that cheating is not driven by concerns about standing out. Rather, it shows that our sense
