Ask a bride before walking down the aisle “How likely are you to get divorced?” and most will respond “Not a chance!” Tell her that the average divorce rate is close to 50 percent, and ask again. Would she change her mind? Unlikely. Even law students who have learned everything about the legal aspects of divorce, including its likelihood, state that their own chances of getting divorced are basically nil. How can we explain this?
Psychologists have documented human optimism for decades. They have learned that people generally overestimate their likelihood of experiencing positive events, such as winning the lottery, and underestimate their likelihood of experiencing negative events, such as being involved in an accident or suffering from cancer. Informing people about their statistical likelihood of experiencing negative events, such as divorce, is surprisingly ineffective at altering their optimistic predictions, and highlighting previously unknown risk factors for diseases fails to engender realistic perceptions of medical vulnerability. How can people maintain their rose-colored views of the future in the face of reality? Which neural processes are involved in people’s optimistic predictions?
To answer these questions we have investigated optimism by using a recent, burgeoning approach in neuroscience: Describing neural activity related to complex behavior with the simple concept of “prediction errors.” Prediction errors are the brain’s way of keeping track of how well it is doing at predicting what is going to happen in the future.
The concept of prediction errors was initially put forward in research on artificial intelligence. By now, scientists have used the basic concept of prediction errors in several domains and have come up with various ways of describing prediction errors in mathematical equations. Let me give you the basics without any mathematics: Imagine your granny tells you that she will give you some money next time she visits. You estimate how much money she will give you, maybe 10, maybe 100 dollars depending on how rich (and generous) your granny is. When she gives you the money you will not only be happy about the money but you will also see how much your prediction differed from what you actually got; in other words, you calculate a prediction error. Knowing this prediction error will help you to estimate how much money you will get the next time your granny comes along. It’s an essential part of learning, and the brain is doing it all the time.
How have neuroscientists employed the idea of prediction errors to study brain activity? In dozens of studies, researchers have looked for and identified brain regions that are related to the calculation of prediction errors. They do this in various ways, but the typical experiment consists of having participants gamble for money on computerized versions of slot machines. At the same time, participants’ brains are monitored in functional magnetic resonance imaging (fMRI) scanners.
Interestingly, similar patterns of brain activity seem to be at play when participants gamble for money and when they engage in complex social interactions. For example, in our everyday life, we often have to track how good or bad the advice of another person is. Timothy Behrens and colleagues from Oxford University used prediction errors to model how humans incorporate advice from social partners into their decisions. Participants repeatedly had to choose which one of two options would yield a higher reward. Before they made their decision, they saw which option another person would advise them to choose. So participants had to form prediction errors for two types of information: non-social (how rewarding are the two options) and social (how good is the other person’s advice). The two kinds of prediction errors were processed in a similar fashion, suggesting conceptual links between processing social and non-social information.