Study identifies neurons that predict what another individual will do
By SUE McGREEVEY
Researchers have identified neurons in primates that anticipate others’ intentions and actions. Video: Cell
Every day we make decisions based on predicting what someone else will do—from deciding whether the driver approaching an intersection will stop for the red light to determining whether a particular negotiation strategy will result in a desired outcome.
Now a study by Harvard Medical School investigators at Massachusetts General Hospital has discovered two groups of neurons that play key roles in social interactions between primates.
One group is activated when deciding whether to cooperate with another individual and another group is involved in predicting what the other will do. The findings appear in Cell.
“For a long time we have been interested in understanding how complex social interactions between individuals are mediated by neurons within the brain,” said Keren Haroush, HMS instructor in neurosurgery at Mass General and lead author of the Cell paper.
“We found that part of the frontal lobe called the anterior cingulate cortex plays an essential role in mediating cooperative social interactions in Rhesus monkeys. Some neurons encode the animal’s decision whether or not to cooperate with another monkey, and a separate group of neurons was activated in predicting what the other monkey would do before it had made its decision. The activity of those other-predictive neurons was uniquely affected by the social context of the interaction,” Haroush said.
The anterior cingulate cortex is broadly connected with other brain regions known to be involved in interactive behavior, and damage to this structure results in reduced interest in other individuals compared with inanimate objects. In fact, people with autism spectrum disorders or other conditions affecting social interactions, such as antisocial personality disorder, have been found to have abnormalities in the anterior cingulate cortex.
To better understand the role of the anterior cingulate cortex in making one’s own decisions and predicting what another individual will do, Haroush and senior author Ziv Williams, HMS associate professor of neurosurgery at Mass General, tested pairs of monkeys in a version of the classic prisoner’s dilemma game.
In the game, each monkey is given a choice—in this instance which of two displayed symbols to choose—and the relationship between the two animals’ choices determines how much of a reward each will receive. In repeated trials with the monkeys sitting next to each other, the animals learn through experience that one symbol represents cooperation with the other monkey and the other represents a lack of cooperation called defection.
If both animals choose the cooperation symbol, both get an equally large drink of juice, but if one chooses defection and the other chooses cooperation, the defector gets the largest amount of juice and the cooperator gets the smallest. However, if both animals choose to defect, both get an equally small drink of juice; so deciding how to get more juice involves predicting what the other animal will choose.
Each trial randomly alternated which animal was given the opportunity to choose first. After both had made their choices, the monkeys could see what each had chosen and detect how much juice each received. While the animals were more likely to select defection versus cooperation overall, they were less likely to cooperate if the other monkey had defected on the previous trial. Mutual cooperation between both monkeys increased the likelihood of cooperation on future trials.
Two versions of the trial that changed the social context of the experiment—one in which the monkeys were in separate rooms and the other in which a monkey played the game against a computer—significantly reduced the likelihood of cooperation and of reciprocation after previous mutual cooperation.
Measuring the activity of 353 individual neurons within the anterior cingulate cortex while the monkeys performed the trials revealed that about half were activated during the task. Of these task-responsive neurons, a quarter showed differences in activation based on the animals’ individual choice, and an even larger group—a third of those involved in the task—showed changes in activation corresponding with the as-yet unknown choice of the other monkeys. The predictions made by the activity of these neurons were as accurate as those made by an algorithm that evaluated the animals’ previous choices.
“We also found that these ‘other-predictive’ neurons were uniquely affected by the social context of the interaction and were much less active when the animals were separated, supporting the role of these neurons in anticipating another individual’s intentions or covert state of mind,” Williams noted. “In addition, temporarily disrupting the activity of the ACC during a series of trials reduced the overall likelihood of cooperation and specifically of reciprocal cooperation, which is in line with previous studies that have found ACC involvement in disorders affecting social interaction.
“Social interactions are complex, and here we touched on only a small aspect of how individuals interact,” Williams added. “Our eventual hope is to better understand how these complex, multifaceted interactions are encoded within the human brain and use this understanding to develop new, targeted treatment for disorders such as autism and antisocial behavior, which are often characterized by difficulty with social interaction.”
The study was supported by National Institutes of Health grant 5R01-HD059852, the Presidential Early Career Award for Scientists and Engineers, and the Whitehall Foundation.