Why do people cooperate? Why do we band together in extraordinary numbers to solve problems? Why do we commit acts of “heroism” to protect or save others, including sometimes people we don’t even know personally?

The level of cooperation that humans routinely exhibit poses an evolutionary puzzle and an enigma to economists. The essence of cooperation is the fact that a “public good” is delivered despite the risk of “free riding”—that is, the possibility that some members of the community enjoy the public good (a playground) but do not contribute to it (do not volunteer their time to help build the playground). But if free riding is possible, why do we not all ride for free? Indeed, that’s what classic economics and game theory predict: Homo economicus is thought to act in his or her own selfish interests only, and this selfishness should put an end to cooperation whenever free-riding is possible.

One approach to explaining human cooperation assumes that people have the ability to detect and avoid cheaters or free riders in social exchange situations. Along those lines, it can be shown that unless people can punish free riders, cooperation will falter.

Does it follow that we have evolved a “cheater detection module” that is uniquely dedicated to identifying cheaters?

Researchers Kroneisen, Woehe, and Rausch from the University of Mannheim, Germany, recently set out to test this idea. Their findings, reported in a recent article in Psychonomic Bulletin & Review, involved a study that set out to test the idea that people have selectively better memories for “cheaters”: If you can remember that someone has cheated in the past, then you would presumably be less likely to cooperate with them in the future. Over time, this would avoid damage to your welfare and might therefore bestow an evolutionary advantage.

Kroneisen and colleagues presented participants with photographs of faces that were accompanied by behavioral descriptions that were either aggressive, prosocial, or neutral. For example, an aggressive description might proclaim that “X often provokes a fight with other soccer fans”, whereas a prosocial person might be characterized as “Y often provides the homeless with a tasty meal.” Those descriptions were randomly assigned to faces.

During the first phase of the experiment, participants were presented with the photograph and the description before they rated how “likeable “ they found the person. During the second phase, the same faces were shown again (but without the description) intermingled with a new set of faces that were not shown during the first phase. Participants again had to rate the photos for their likeability, and they then had to indicate whether or not they had seen the face during the first phase of the study. If they identified a face as having been seen before, participants additionally had to indicate whether that person had been identified as aggressive, social, or neutral during the first phase.

The analysis focused on people’s ability to remember the putative character of people shown in the first phase.

As expected on the basis of the idea that we have a “cheater detection module”, Kroneisen and colleagues found that people were better at identifying aggressive faces than faces that were presented as social or neutral.

The results suggest that you would be quite likely to recognize a person as a potential threat if you had previously seen them behave aggressively—much more so than you would recognize someone as a philanthropist even if you had previously observed them feeding the hopeless.

Does this support the notion of a “cheater detection module”?

Not entirely.

The study by Kroneisen and colleagues also manipulated how participants approached the task. In another condition that was identical to the one just described in all other respects, people were instructed to imagine that they had moved in to a new home “in a very bad neighborhood where mainly aggressive and violent people live and the police are seen quite often,” and that the stimuli they were about to see were pictures and descriptions of their new neighbors.

Imagining this unpleasant situation is likely to have one important consequence: In a bad neighborhood, social behaviors should be more unusual than they normally are. If your (pretend) neighbors are all aggressive, then the person who feeds the homeless is likely to be unexpected or unusual, and hence distinctive. Distinctiveness, in turn, has long been known to be a major variable that determines the quality of memories—if you ever saw a pink poodle on a bus, you’d be unlikely to forget the experience.

Intriguingly, Kroneisen and colleagues found that people’s ability to recognize a person as being social was greatly improved in the condition in which participants had to imagine their move to a bad neighborhood. In fact, their ability to recognize a person’s character was identical for ostensibly aggressive and ostensibly social faces in that condition.

In a nutshell, if you expect people to behave badly, then you remember kindness as well as you would remember aggressive behavior when you have no specific expectations about how people should behave.

We may not have evolved to have a “cheater detection module”, but we certainly have memories that remember things to the extent that they are unusual or unexpected.

What color was that poodle on the bus again?