Lie Detection Takes a Village - Pacific Standard

Lie Detection Takes a Village

Researchers find that groups are better than individuals at catching dissemblers—though not by much.
These storks know how to root out a liar. (Photo: Navaneeth KN/Flickr)

These storks know how to root out a liar. (Photo: Navaneeth KN/Flickr)

Lie detection is at best an imprecise science; at worst, a guessing game. Whatever techniques people try—and there's no shortage of techniques—very few do better than chance at telling truth from fiction. But in the never-ending search to find a way to find the truth, researchers have found something that works better, if only a little bit: group discussion.

Nearly all of the research on how to determine whether someone's lying—whether it's robbing a bank or cheating on a spouse—has focused on how an individual can ferret out a lie. There is, of course, the polygraph machine, which most psychologists view as completely bogus. And there's the micro-expressions technique, which looks at tiny changes in facial expressions—and probably doesn't work very well.

Groups correctly distinguished lies from truths about 61.7 percent of the time, while individuals did so only 53.6 percent of the time.

But if individuals aren't so great at distinguishing truths from lies, perhaps groups are, suggest Nadav Klein and Nicholas Epley of the University of Chicago's Booth School of Business. For one, groups may be collectively more accurate in detecting lies than any one individual, an effect sometimes called "the wisdom of crowds." Plus, groups may be less trusting than individuals, which could help overcome any one person's tendency to believe a peer. There's also evidence that if groups are encouraged to share and discuss different kinds of information, they may make better decisions.

Importantly, that last effect differs from the wisdom of crowds. In the wisdom of crowds scenario, everyone estimates how likely it is a person's lying and then pool those estimates together. In the information-sharing scenario, groups pool the information underlying each individual's estimate, rather than the estimates themselves. Sharing the underlying knowledge, previous research suggests, may produce better overall estimates.

To see whether group discussion actually does improve lie detection—and if so, which of those three explanations might be at work—Klein and Epley conducted three experiments in which individuals and groups of three had to decide whether another individual depicted in a video was lying. Groups always performed better than individuals. In the first version of the experiment, groups correctly distinguished lies from truths about 61.7 percent of the time, while individuals did so only 53.6 percent of the time.

The advantage came almost entirely from groups' improved ability to detect lies. While groups did little better than individuals at identifying truthful statements, they were much more successful—63.9 percent versus 51.2 percent correct—at identifying lies.

Most important, however, was the reason behind why the groups did better: discussion. Klein and Epley found that out by comparing real groups with nominal groups, formed by averaging the judgments of three individuals who hadn't spoken to each other. Nominal groups barely outperformed the individuals that made them up. In the first experiment, for example, nominal groups averaged just 54 percent, a tiny increase over the individual average of 53.6 percent. In other words, a group's advantage stemmed from sharing different kinds of information, rather than averaging individual estimates of a person's honesty.

Quick Studies is an award-winning series that sheds light on new research and discoveries that change the way we look at the world.