On the Interplay of Gullibility, Plausibility, and Criticism: A Computational Model of Epistemic Vigilance

Humans heavily depend on communication. We constantly share new ideas, catch up on current news, and exchange gossip. Much of the information conveyed in this way is, however, not first-hand. As a result, we run the risk of being misinformed and of spreading potentially harmful messages via large social networks. Current research argues that we are endowed with a set of cognitive mechanisms capable of targeting such risks. These mechanisms, known as mechanisms of epistemic vigilance, help us evaluate communicated information by i) critically evaluating presented arguments, ii) checking the plausibility of messages against pre-existing background beliefs, and iii) assessing the competence of a sender based on cues of trustworthiness. So far, the mechanisms exist only as verbal theory, which do not allow a thorough systemic analysis of the interplay between them. In this paper, we implement an agent-based computational model of epistemic vigilance to add to the existing microscopic (individual level) and macroscopic (societal level) understanding of the mechanisms. Through simulations of different multi-agent societies we are able to show that the mechanisms of epistemic vigilance are sufficient to explain a wide variety of phenomena: (a) The locality of critics in social groups is a deciding factor when it comes to quickly correcting false messages. (b) Plausibility checking can create impeding group structures that exclude other agents from receiving surrounding information. (c) Impeding group structures can be overcome through competence checking. (d) And on a societal level, increasing the proportion of agents performing plausibility checks, creates an abrupt shift from consensus to polarization.