Every day, people make countless decisions, big and small: Should I buy that new house? Do I want chocolate or vanilla ice cream? A recent study suggests that when faced with uncertainty regarding a choice, how a person evaluates their options may not fit the rational mathematical model that scholars had assumed to be the case.
“How we make decisions is not optimal,” says Alireza Soltani, an assistant professor of psychological and brain sciences at Dartmouth College in Hanover, New Hampshire, and senior author of the study published in Nature Human Behavior.
Economists have claimed that we assess uncomplicated decisions optimally, by multiplying the size of—or how much we personally value—a reward by the likelihood of getting that reward. If we don’t precisely know the probability, then we just plug in an estimate.
To test this assumption, Soltani and his team focused on a very simple decision: a choice between pairs of gambles, each with a reward and a probability. Because monkeys are often studied as a model for human decision-making, the researchers decided to validate that practice by also testing for differences in gambling behavior between species. They enlisted seven rhesus macaques and about 100 undergraduate students. Winning gambles paid out as points that were converted to money and extra credit for the students, and as drops of juice for the monkeys.
In one experiment, the researchers gave monkeys and students the reward probabilities and magnitudes explicitly, represented by the length, area, and color of rectangular bars on a computer screen. In other experiments, monkeys and students chose between two colored options. For the monkeys, the magnitude was represented by the number of dots around each choice; for the students the magnitude was a number. In these gambles, they had to learn the odds of winning through feedback—with probabilities volatile rather than fixed in certain scenarios.
Researchers examined choice behavior of more than 100 students in about 50,000 trials, and of seven monkeys performing more than 500 sessions and about 320,000 trials. They found that while both monkeys and students tended to multiply the reward by the probability when given the exact probability, the brains of both primates opted for a different sort of calculation in cases of uncertainty. They applied weights to the reward and the estimated probability, and then added those together. In other words, they adopted an additive model rather than a multiplicative model. When the researchers incorporated volatility—manipulating the experiment so that the probabilities of winning with each choice in the pair reversed every 80 or 20 gambles—monkeys and humans tended to put more weight on the size of the reward than on the probability.
“If you go to get chocolate ice cream at Cold Stone, you don’t actually know if they will have chocolate,” explains Paul Zak, director of the Center for Neuroeconomics Studies at Claremont Graduate University in California. “So, in a world that has a lot of uncertainty, which is the world a lot of animals and humans live in, we basically ignore stuff we don’t know and focus on how much we are going to enjoy the chocolate ice cream.”
Zak says the findings are compelling—and make good sense. “Evolution gives you good enough; it doesn’t give you the best,” he notes, suggesting that adoption of the additive strategy under uncertainty provides a decent enough approximation while conserving the brain’s energy. “It is computationally simpler and therefore less cognitively costly.”
Research had shown that for decisions with a lot of variables—say, choosing a car or a college to attend—people might go through this same kind of back-and-forth, comparing attributes and applying weights. But no one had looked at what happens with simple decisions.
Measures of brain activity bolstered the conclusions of Soltani’s team. Scans in two monkeys, taken as they were gambling, showed that uncertainty modulated the way some neurons in a region of the prefrontal cortex encoded for the magnitude of reward.
“Adding probability and magnitude seems absurd. It’s like apples and oranges. But this is effectively what we have evolved to do,” says Soltani, noting that this strategy allows us to adjust the weight we give the probability as we collect more information or as the environment shifts.
“Being the smartest species is more about being flexible than about being optimal,” Soltani adds. “When you optimize certain things your behavior could be a disaster for another environment.”
Next steps, he says, include using this new lens to reevaluate all the decision-making strategies and biases that scholars have previously measured, as well as perhaps redefining how we judge human behavior. “Maybe we can better explain suboptimal performance by thinking, ‘Oh, wait a second, this makes us flexible,’” says Soltani.