Individual Submission Summary
Share...

Direct link:

Predicting Nationwide Campaign Effects From Multivariate Brain Activity Signatures

Sun, May 27, 8:00 to 9:15, Hilton Prague, Floor: L, Barcelona

Abstract

Billions of dollars per year are put towards public service messaging, but it is difficult to predict which messages will be the most persuasive and successful in changing behavior. Recent work has found that neuroimaging can improve predictions of message success and behavior change, but has primarily focused on average brain activity in single brain regions. By contrast, distributed brain activity signatures have been used to predict outcomes in specific cognitive and emotional domains, but have not been applied in the context of persuasive messaging, nor to link brain responses to large-scale human behavior. Here we record brain activity in a sample of smokers exposed to anti-smoking messages and test whether message-wise response patterns can forecast large-scale effectiveness of the same messages in the field. Specifically, we assessed the similarity between the neural response for each message and theoretically motivated multivariate brain signatures for positive valence, negative emotion, and vividness. We find that the latter two metrics predict population-level message success, operationalized as the click-through rate (CTR) each banner generated when it was viewed during the nationwide online campaign. Our work suggests that such multivariate signature maps, assessed in a subgroup of the population, can forecast meaningful, real-world behaviors, and may ultimately provide insight into improving message success in the new media environment.

Authors