In a new article in Royal Society Open Science, CSDE External Affiliate Aja Sutton (Population Research Center at Portland State University) with co-authors Adam Z. Reynolds (University of New Mexico), Matthew A. Turner (Stanford University), and James Holland Jones (Stanford University) examine how (digital) social influencers can modify epidemics by affecting social learning of health-protective behaviors in group-structured populations. Using agent-based models that incorporate both small protective and anti-protective nudges from social influencers into an epidemic scenario, they test how—under varying conditions of group structure modified by homophily and out-group aversion—competing influence messages affect health-protective behavioral diffusion between two behaviorally naive groups, and by extension infection transmission dynamics and outcomes. In heterogeneous populations, social influencers were protective of the whole population by increasing behavioral diffusion—independent of homophily—and flattening the epidemic curve, even in the equal presence of anti-protective messaging. Stronger group structure—especially, homophily—produced behavioral segregation and modified infection growth rates by accelerating within-group behavioral diffusion, leading to a lag between groups’ epidemic peak intensity and total infection burden. This work suggests contexts through which public health messaging is shared—such as social media sites, which exhibit a high degree of homophily—can produce substantial differences in disease transmission dynamics and epidemic outcomes.