Descriptive norms can "backfire" in hyper-polarized contexts

Descriptive social norms interventions, where a behavior is promoted by learning that others engage in that behavior, are a cornerstone of behavior change research and practice. Here, we examine the effect of learning about the behavior of outgroup members in a hyper-polarized context: mask-wearing during the COVID-19 pandemic. Contrary to prior findings, we find a descriptive social norm "backfire": Across three experiments, Biden supporters increased their mask-wearing intentions after being informed that most Trump supporters never wore masks. We also provide evidence consistent with a mechanism whereby this effect was driven by changes in perceptions about how negatively ingroup members view nonmask wearing. Finally, in a fourth study, Biden supporters show the traditional descriptive norms effect (rather than a backfire) from the same treatment when in a nonpolarized context: dishonesty in a coin-flipping task. These findings help to clarify why descriptive social norm interventions promote behavior change, and underscore the importance of social norms in motivating prosocial behaviors. They also suggest an update to current best practices in the design of descriptive norm interventions: in polarized contexts, it can be beneficial to publicize antisocial behavior of outgroup members.