Anxiety Hampers Learning About Untrustworthy People

Anxiety isn't bad for learning in general, just for learning about people.

Posted Sep 21, 2020

In all facets of our lives, we have to figure out who to trust. Trust involves a level of vulnerability. You have to give something of yours to someone else with the expectation that they will use it for mutual benefit rather than exploiting what is given to them. The actions other people take when they are in a position of trust give you some sense of how they are likely to act in the future.

Image by Ppart on Shutterstock
Source: Image by Ppart on Shutterstock

In general, people are good at learning who they should trust. This skill is critical for dealing effectively with the people around you. An interesting paper in the May 2020 issue of Psychological Science by Amrita Lamba, Michael Frank, and Oriel FeldmanHall explored whether anxiety might make it harder for people to learn who to trust.

The design of this study is complex because it aimed to isolate the influence of anxiety on social learning and was set up to be able to track changes in people’s behavior over the course of the study.

Participants performed two tasks via a computer. One was a “trust game” in which participants were led to believe they were playing with three other people over the internet. In fact, the computer was choosing responses for participants. The way the game was set up, on each of the 84 trials, participants started with $1. They could give as much of that dollar to their opponent as they wanted. The amount they gave was then multiplied by 4, and the opponent was allowed to share back with the participant as much of that money as they wanted. For example, if the participant gave their opponent $0.40, the opponent would then get $1.60 (the 40 cents multiplied by 4) and they could choose to give as much back as they wanted. They might give the participant $0.50 and keep the $1.10 for themselves.

The three “opponents” were programmed so that one started by returning almost half of the money on each trial and gradually decreased the amount they gave and increasing toward the end. A second “opponent” started by returning a moderate amount, then decreasing, then increasing. Finally, a third “opponent” started by returning a small amount, gradually increasing over the course of the study and then decreasing. The average amount returned by each “opponent” was the same over the course of the study.

Throughout the study, participants were randomly selected to play one of the three opponents on each trial, and they play each of the opponents 28 times during the study. The aim was for them to learn how each opponent responds and to wager an appropriate amount given how much they expected that the opponent would return. Technically, to maximize the amount of money they make, they should give it all away if they think the opponent will return at least 25 percent and to give none of it away if they think the opponent will return less than 25 percent.

To compare social learning with nonsocial learning, the participants also played a round in which they had choices among three slot machines. The slot machines were programmed to respond the same way as the people in the social learning version of the game. Overall, participants invested about 5 cents more per trial in the slot machine game than in the social game, suggesting that they trust people less than machines.

There was a tendency for people to let first impressions play a big role in their decisions in the game both for the machines and the humans. The opponent (or machine) that started out returning the most money was given more money across the study than the machine that started off neutral. This difference was larger for the social game than for the slot-machine game.

Participants also filled out an anxiety inventory. A key question in this study is how participants with high levels of anxiety differ in their performance from those who do not have a high level of anxiety.

Anxious participants performed similarly to nonanxious participants for the slot machines, but they had a hard time in the social game. In particular, they tended to continue to give money to people, even when they were not returning enough to cover what they gave. Anxious people’s learning suffered when they had to learn from the loss of money to another person.

Overall, participants were able to adjust their play of the game over time. Participants were able to give away less money to people or slot machines when they were being less generous. However, anxious participants found it harder to learn about people than about machines. Anxious people had a particularly hard time adjusting their behavior after losing money to a person.


Lamba, A. Frank, M.J., FeldmanHall, O. (2020). Anxiety impedes adaptive social learning under uncertainty. Psychological Science, 31(5), 592-603.