The Evolution of Online Virality

In social psychology and epistemology, the dominant approach to online content sharing focuses on individual, evidence-based decisions and downplays users´ anticipations of engagement (reactions, comments, etc.). Thus, it struggles to capture engagement-seeking ´influencers´ and ´followers´ who spread and relay content without necessarily believing it factual. This, in turn, impacts our ability to prevent harmful content from going viral.In earlier work, we proposed a strategic model of content-sharing informed by speech-act theory. We demonstrated, analytically and in agent-based simulations, how a few engagement-oriented users can flood a network with content that reflects neither the preferences nor the beliefs of most users. Our results are consistent with empirical findings tracking viral content to few accounts and suggest that ´influencers´ don´t need to leverage beliefs. This project will investigate simulated networks with engagement-oriented ´influencers,´ where other users alternate between content- and engagement-oriented strategies. We will determine which populations and network structures generate the same output as evolutionary models of opinion dynamics (particularly, of deception) known to capture real-world online social network behavior. This, in turn, will enable novel preventive measures targeting influencers (e.g., increasing the social cost of deception) rather than their messages only (e.g., fact-checking and moderation).