Brain Scans Predict Human Choices At Scale

Neuroimaging can capture brain activity in response to stimuli before a person decides how to respond. Initial affective responses—broadly good or bad feelings about a stimulus—have been associated with activity in evolutionarily conserved subcortical and cortical circuits including the Nucleus Accumbens (NAcc) and Anterior (AIns). Activity then continues through integrative circuits associated with more deliberative and reflective processing. Previous work has suggested that the early affective responses may be more broadly shared across individuals than the final choice behavior. Alexander Genevsky and colleagues tested the utility and generalizability of "neuroforecasting"—the use of neural data collected in the laboratory to forecast real-world aggregate-level behavior. In a series of experiments they compared functional magnetic resonance imaging (fMRI) data from groups of approximately 40 people to internet survey data from thousands of people. One experiment asked participants to decide whether to fund real film projects posted on the crowd-funding website Kickstarter. In another, participants decided whether to continue to watch short videos from the video sharing website Youtube. While choices made by people in the fMRI group were not always significantly associated with those of people online, activity in the NAcc in the fMRI group was consistently associated with the choices of the online participants. The authors attribute this result to the generalizability of activity in the NAcc compared to more idiosyncratic responses of other brain regions that ultimately contributed to the final choice. NAcc data even forecast the choices of groups of internet participants who were unrepresentative of the fMRI group. According to the authors, brain data collected from even relatively small groups of people may effectively forecast choices that involve good or bad feelings.