The Surprising Way Crowd Size Affects Our Tendency to Cheat

That’s the lesson from new research by Maryam Kouchaki, a professor of management and organizations at the Kellogg School. Across several experiments, Kouchaki and her coauthors—Celia Chui of HEC Montreal and Francesca Gino of Harvard Business School—found that people cheat at higher rates in larger groups.

Why? The researchers found an intriguing self-fulfilling prophecy at work: people expect there to be higher numbers of cheaters within larger groups. This perception, in turn, increases the sense that cheating is common and therefore acceptable.

The study illustrates the importance of context and social norms in determining whether or not we behave ethically. After all, we don’t magically transform from saints in groups of five to sinners in groups of 100. Rather, we unconsciously cue off other people and what we expect their behavior to be.

“When you think your behavior is normative,” Kouchaki explains, “then it seems more defensible. Questionable behavior seems more justifiable when you think more people are doing it.”

Larger Groups, More Cheating

The researchers began their investigation by recruiting 88 participants to complete an in-person study. Participants were randomly assigned to a room with either a small group of 5 people or a large group of 25 people.

Participants were told they could earn money by unscrambling 10-word jumbles in three minutes, $1 for each one they solved. But there was a catch: they had to complete the puzzles in sequence to receive payment—that is, if they solved the first and third, but not the second, they would only be paid $1 for the first. They would also receive an additional $10 bonus if their performance was in the top 20 percent of the room.

Participants didn’t have to report solutions to the word jumbles—only which ones they solved, which presented a tempting opportunity to lie. And unbeknownst to them, the seventh word jumble was unsolvable, allowing the researchers to determine conclusively whose pants were on fire.

The results were clear: in the small groups, 27 percent of participants reported solving the unsolvable word jumble. In the large group, that figure rose to 54 percent.

The researchers found the same pattern when they did a similar experiment with a group of 187 online participants. Participants were only told the size of the group but couldn’t actually see their competitors. In keeping with the earlier results, they cheated at higher rates in a group of 100 than in groups of 10. In the large group, participants reported solving an average of 1.14 unsolvable word jumbles, compared with 0.747 in the small groups.

Everybody Else Is Doing It, So Why Can’t We?

While these first two experiments found strong evidence of more cheating within larger groups, they didn’t explain why. So, for the next experiment, the researchers investigated that question directly.

They recruited 296 online participants, assigned them to either a large group of 100 people or a small group of 10, and gave them instructions for the word-jumble task used in the earlier study. Instead of actually completing the task, however, participants answered a series of questions designed to probe the mechanism underlying large-group cheating behavior. Researchers asked participants, for example, how many people in their group they thought would cheat, whether they thought cheating was in line with the norms of the group, and how likely it was that cheaters would be caught.

The researchers found that, as group size increased, the expected number of cheaters increased too. Statistical analysis revealed that this larger number of expected cheaters caused people to view cheating as more normative. Their analysis did not show a clear correlation between expectations of getting caught and the likelihood of cheating.

It’s worth noting that only the expected number of cheaters—and not the expected percentage of cheaters—had the effect of making cheating feel normative. Large-group participants expected that, on average, 36.95 people would cheat; small group participants said 4.55. This translates to a higher percentage of expected cheating in the small group (45.55 percent) than the large group (36.95 percent).

The researchers suspect this is related to a common fallacy called ratio bias. Studies have found that people will choose a 7-in-100 chance of winning over a 1-in-10 chance of winning, even though the odds are worse. This is because our fallible brains focus on the numerator and think, essentially, “well, seven is more than one—better odds!” Similarly, if you believe 37 of the 100 people in your group will cheat, it makes lying feel more prevalent than if you believe 5 of 10 people will cheat—even though this perception is not accurate.

In other words, our collective, if faulty, belief that others are more likely to cheat in large groups becomes a self-fulfilling prophesy by spurring more of us to cheat when we’re in those large groups.

Why Bigger Isn’t Better

How can we avoid the moral perils of large groups? Perhaps by keeping the competition pool small. By minimizing the number of other competitors, leaders can decrease their employees’ expectations of facing numerous cheaters and, as a result, reduce the odds that they themselves will cheat, Kouchaki and her coauthors suggest.

More importantly, be explicit about how employees are expected to perform.

Communicating your expectations for honest, accurate, and transparent performance will lower employee expectations that others will cheat. This can help, even if competition-group size cannot be decreased, since you will be decreasing employee expectations that others will cheat.

And finally, don’t assume that people know that cheating is unacceptable. When people in larger competition pools expect that more people will cheat to win, they tend to believe that cheating is more acceptable, and they end up cheating more themselves. Leaders can counter these perceptions by clearly emphasizing which performance behaviors are tolerated and which are not.

And, of course, there’s no substitute for promoting an ethical culture.

To Kouchaki, one of the sobering lessons of this research is just how easily we justify unethical behavior. Something as simple as the size of our reference group can make the difference between cheating and truth-telling. So, a word of warning: left unchecked, she says, “these small psychological processes could create a path to a really corrupt culture.”