You Have to See It to Believe It: Social Exposure as a Near-Necessary Condition for Conspiracy Adoption Online
Date:
Presented at Sunbelt 2026, the annual meeting of the International Network for Social Network Analysis (INSNA), Thursday, June 25, 2026, 9:20–9:40am, Coquina C. Session: Contagion and Diffusion Processes Through Social Networks.
Authors: Zach Brown (University of Washington, presenting) and Zack W. Almquist (University of Washington).
Abstract: How does conspiracy discourse spread through online communities, and what role does social exposure play relative to independent uptake? We investigate the case of 3i/Atlas on Reddit, a comet that upon discovery was framed as proof of non-human intelligence in UFO-focused subreddits and attracted sustained engagement over several months before belief cratered in the face of scientific consensus. We construct a bipartite thread–user network from 825,287 posts by 109,143 users across six subreddits, enabling counterfactual identification of network exposure effects by comparing users exposed to “infected” threads against users who participated only in never-infected threads within the same communities. Cox proportional hazards models estimated on 1.3 million person-days reveal network exposure as a near-necessary condition for keyword adoption: only 0.01% of never-exposed users adopted (6 out of 47,004), compared with 47.2% of highly-exposed users (3,042 out of 6,443). Any exposure is associated with a 59% increase in adoption hazard. Following the conspiracy theory’s disconfirmation, network-central users were disproportionately likely to pivot to an alternative narrative rather than exit quietly, with Pivoters exhibiting 2.6 times the mean degree of Exiters. These findings contribute a replicable bipartite network framework for measuring information contagion and provide substantive evidence that social reinforcement drives conspiracy belief adoption online.