when peer pressure makes teens safer

a quick game

let’s start with a game. pick the gamble you’d rather play. after four rounds, you’ll see what two other players chose before you decide.

teens in our study played a version of this, sometimes alone and sometimes after seeing what peers chose. we wanted to know whether a peer’s choice made that option more appealing, and whether it mattered if the peers played it safe or took the risk.

what 78 teens chose

78 teens made ‘gamble’ choices like the ones you just made, sometimes alone and sometimes after seeing two peers’ choices. in the plot on the right, each dot is one teen. the higher the dot, the more extra value a peer’s safe choice added to that option for them.

blue dots are teens who said they’d never used substances. coral dots are teens who reported some use.

measuring the value a peer’s choice adds

how did we quantify extra value? we used a computational model, a series of equations with parameters that are fit to the choices of each participant, that first estimates how much each person likes a gamble on its own, given how much risk they’re comfortable with (the ‘risk preference’ parameter). then it estimates how much seeing a peer pick that gamble adds on top (the ‘social value’ parameter, called other-conferred utility in the paper).

teens who had never used substances got a bigger value boost from peers’ safe choices than teens who had.

0.5 on this scale means a peer’s choice added nothing.

when peers take the risk

now look at trials where the peers picked the riskier gamble. the same teens’ dots move to show how much those risky choices added.

here the two group averages sit close together, and we didn’t find a difference tied to substance use. the difference only showed up for safe choices. in other words, differences between substance-use groups seemed specific to safe influence.

the same pattern in vmPFC

31 of these teens also had fMRI scans, which track changes in blood flow as a stand-in for brain activity. we looked at vmPFC, a region whose activity tracks the value of a given option.

now each dot shows one teen’s vmPFC response to the value added by peers’ safe choices. the same pattern appeared: the response was stronger in teens who had never used substances. their choices and their brain’s value signal told the same story.

only the 31 teens who were scanned appear here, so there are fewer dots.

risky peers in vmPFC

on the other hand, when we look at brain responses in vmPFC to risky peers, we didn’t find a difference in vmPFC between the two groups. again, this mirrors our computational model findings that there is no difference in risky influence between substance use groups.