Logic, Belief Propagation and Misinformation

We are broadly concerned with security risks associated with the spread of misinformation. Misinformation is spread when messages are exchanged on a network, and some agents erroneously believe reports from agents that do not have suitable expertise. So the spread of misinformation is closely connected to the notion of belief change, which is a topic that has been studied extensively in the formal Artificial Intelligence (AI) community. The problem is that formal models of belief change are generally focused on fully rational agents. In practice, however, the trust that agents have in information sources is influenced by non-logical factors. For example, agents might be more likely to trust sources that have demonstrated alignment on some unrelated set of values. In this short paper, we propose a formal approach to modelling belief change that is based on a flexible kind of trust transformation, where the extent to which a source is trusted is not entirely dependent on expertise or reliability. We demonstrate how this approach can be used to capture the propagation of false facts over a social network. This is a preliminary paper outlining a basic approach, which we intend to use as the basis for software that provides logic-based recommendations related to the accuracy of information received on a social network. We see this as a kind of reasoning prosthetic, which limits the spread of information by providing transparent information that helps users understand the consequences of their own beliefs.