Weekend Reads | The Honesty of Groups

This weekend's read is a recent research paper looking at the "honesty of groups." The researchers, from the University of Regensburg in Germany, investigated specifically how the size and makeup of groups affect their collective decision-making. And they found some fascinating patterns.

Much of this kind of research has traditionally been done in psychology departments, but these days, a fair amount of it happens in economics departments as economists explore the relatively new field of "behavioral economics" and attempt to quantify the behavior and incentive structures of groups: consumers, businesses, governments, and even humanity as a whole. A lot of effort goes into designing experiments that will allow the researchers to measure behavior, hopefully in ways that generalize to other situations.

In this case, they designed two experiments: one to test groups, and another to test the individual participants to learn more about characteristics each brought to the group dynamic. The first experiment assembled groups of 2 to 5 people, in which they repeatedly rolled a six-sided die. Each group received a reward depending upon the result of the die roll they reported. To be clear: The reward was based upon the number they reported, not the number that was actually rolled, and each individual reported the die roll separately. If they unanimously reported a roll of 5 or less, then the reward was half of a euro multiplied by the die roll (i.e., they got half of a euro for a reported roll of one, up to 2 euros for a five). If they reported a six, then the reward was zero. So, regardless of what was actually rolled, there was a strong incentive for them to report a five each time to maximize their rewards, though there was room for them to lie less and receive a smaller reward. If the group was not unanimous in its report, then the group's reward was zero.

The researchers discovered that the larger the group, the more frequently it would lie. But the size of the lie didn't increase: Once a group of any size chose to lie, it almost always chose to report a roll of five and maximize its reward.

The researchers also found that all-male groups lie considerably more than groups with at least one female member, though there were some interesting nuances to this. It didn't matter much how many women were in the group; the first female member provoked an "honesty shift" that persisted at the same level all the way up through to the all-female groups. However, the researchers couldn't find any hard evidence as to why the addition of any number of women to the group provoked this change in the group's willingness to lie. Mixed-gender groups didn't deliberate longer than all-male groups, and they didn't have higher levels of disagreements, so there was little evidence that women were "enforcing honesty" in the group. The researchers offered two hypotheses for future testing: that men cared more about their reputations in a mixed-gender environment than in an all-male one; or that men tend to believe that women are more honest and that it would be difficult to convince the women in the group to lie so that they could achieve a unanimous report. The researchers suggested that they could test the second theory by changing the rules to require a majority instead of a unanimous report and observe whether majority-male groups behave differently.

The second test they performed was designed to identify which individual members of the groups were more inclined to lie (or "cheat") by testing them one at a time. Each participant was given a quiz with six questions related to obscure classical music trivia. They were told that they would receive a reward for answering all six questions correctly, but that they were not allowed to use the internet to look up answers. The quiz was difficult enough that there was no way that any of the participants could answer all six questions correctly without cheating. So, by inference, every participant who got all six questions correct — 31% of males and 25% of females — was a "cheater."

Knowing which of the participants in the first test were likely cheaters gave the researchers another dimension to explore. They found that even in two-person groups that have no cheaters, all-male groups lie much more (44%) than all-female groups (only 9%). The researchers also discovered, perhaps not surprisingly, that the more cheaters that were in a group, the more frequently that group would lie about the die roll.

This is an interesting, and somewhat amusing, research study, but one that comes with some caveats that will require additional experiments to sort out. First, it's not clear that the rewards — and the negative consequences — in the experiment were enough to fully drive participants' behavior. Was the amount of money being offered meaningful to them? Were they really concerned about the consequences of lying? Second, we're left to wonder how much these results generalize to other situations, though the researchers point out that their findings align with previous experiments along similar lines. Third, are there cultural influences that would result in a different outcome if the experiment was run in other parts of the world? Do societal expectations for honesty, or religious doctrines that are strictly enforced, change the amount of pressure on individuals to remain honest? Fourth, are there aspects of participants' backgrounds, such as education, income level, or active participation in a religious community, that affect how they respond in this situation? And finally, how do these results extend to participation by nonbinary and transgender persons?

Still, it's a thought-provoking result. It's also a good argument for gender diversity in the workplace: Introducing even one woman into an all-male group resulted in a significant shift toward greater honesty.

Honesty of Groups: Effects of Size and Gender Composition

Featured image via Gorodenkoff/Shutterstock.com.