Coming to a Consensus During Risky Decision-Making

Relationships are all about compromise. From deciding on where to eat dinner with a friend to negotiating chore lists at home, we often experience situations that require some flexibility. But what happens when we must work with others—especially people we don't know—to make a risky decision? That's what Caltech's Dean Mobbs, professor of cognitive neuroscience, and members of his lab set out to explore in a recent study.

"We found that most people compromise, settling on something between their two preferences," says Ketika Garg, a postdoctoral scholar research associate in the Mobbs lab. "But we also found that when outcomes are shared, people are biased in how they divvy up responsibility."

According to the researchers, people who claim more credit for successes and deflect blame for failures also compromise less with their partners, which might sound familiar to anyone who has been randomly paired for a joint task at school or work. A paper outlining the team's findings was published on June 29 in the Proceedings of the National Academy of Sciences (PNAS).

Research by Mobbs and his lab often explores risk-and-reward trade-offs, so he, Garg, and graduate student Wenning Deng initially aimed to investigate what happens when people must make risky decisions in collaboration with others who may not have the same risk preferences and how people navigate that tension.

Previous studies have shown that people tend to arrive at a consensus together and their risk preferences can be shaped by the social context, says Garg, who is a co-senior author on the PNAS paper, along with Mobbs. "Now, we show how people compromise in a joint real-time task by treating it as a dynamic process," she explains. "There's a tendency to cooperate, but the idea of compromise is not studied as much; that is a new angle we present with this work."

In the study, participants first repeatedly played a video game to assess their individual risk tolerance. In the game, players represented by avatars had to forage for rewards in a virtual environment made up of nine locations, or boxes, while a cartoon predator who could capture them appeared in certain locations. Individuals played the game 120 times against two different predators, choosing one box for each round, with each predator differing in their overall attack probability. The goal of the game was to learn the probability that a predator would attack a chosen box to maximize rewards while avoiding capture.

In the social phase of the experiment, participants were paired with a stranger to play the game together online 60 more times. Individuals chose locations on their own and the avatars would move to the middle box in between their two choices. A successful trial meant the player or players got to keep the reward points allocated to their chosen box but capture cost them points. In both scenarios, players received a reward proportional to the amount of points they collected in the game.

"We asked them to predict their partner's choice before they made their own to test how they assess a chess move, basically," Garg explains. "You predict where your partner will go, and then, if you do not want to compromise, you have to counteract and move backwards or forward to pull the middle location where your avatars will end up in your favorable direction. But if you want to compromise, then you will move toward your partner's perceived choice."

Next, the researchers asked participants to take credit and assign blame for the outcome of their collectively played games. They found that people were more likely to take credit for wins than blame for losses even when the contributions and outcomes are clear.

"This result was initially surprising because the egocentric bias in responsibility attribution is typically observed under ambiguity—people take credit and deflect blame largely because ambiguous situations leave room to interpret the evidence in their own favor," says Deng, lead author on the paper. "However, we found that even when feedback about both players' choices makes each player's contribution clear, people still show the egocentric bias, claiming more credit for wins than blame for losses."

On the other hand, Deng says this reasoning also makes sense especially when it's correlated with willingness to compromise.

"My guess is that when people take more risk and the group wins, they think their choice leads to more rewards for the group, but when they take less risk and the group wins, they think they are keeping the group safe from danger," Deng says.

The team notes that while previous work has shown that diverse preferences and knowledge can make groups collectively intelligent and improve performance, much of that work treats collective intelligence as an aggregate outcome without examining the interpersonal friction that diverse preferences create.

"Here, we zoom in to the dyadic, or two-person, level and ask: When two people have genuinely different risk preferences and their decisions are coupled, how do they coordinate? How do they attribute responsibility?" says Mobbs, who is also the Allen V. C. Davis and Lenabelle Davis Leadership Chair and director of the Caltech Brain Imaging Center. "We find that compromise is a primary mechanism, that it is reinforced by reciprocity, and that metacognitive biases in responsibility attribution predict who compromises and who doesn't."

This bias, the researchers say, has consequences for whether collaboration succeeds or breaks down; partners who differ but are less biased tend to converge toward a shared middle ground, while partners with strong biases can spiral into conflict, counteracting rather than accommodating each other in a self-reinforcing cycle.

"Whether a team ends up aligned or in conflict may depend less on how different people are and more on how they assign responsibility when things go wrong," Garg says.

In future work, the team would like to increase the group size in the experiments to explore how a larger team dynamic might change how decisions are made. They also want to make a decision-making game that is more naturalistic and represents real-world risks and rewards.

The PNAS paper is titled "Linking Compromise and Responsibility Attribution to Risky Decision-Making in Dyadic Foraging." Funding for the research was provided by the National Institute of Mental Health.