To Believe, or Not to Believe: Modeling Information Cascades and Sequential Learning Equilibria

Abstract

Are perceived "hot" businesses with high ratings on social networks truly superior to average businesses? Does the process of crowdsourcing information lead to more accurate decision-making? How does fake news become so prevalent? Previous research done on information cascades fall into three different camps: theory-based, simulation-based, and data-based. This paper attempts to deploy these three approaches in concert for a holistic understanding of information cascades and sequential decision-making behavior. First, we propose a new model for studying these network interactions. Second, we use theory to determine whether an equilibrium exists in sequential learning; that is, if individuals in groups converge to one choice of action. Then, we use simulations and data-based analysis to understand the factors that affect when the equilibrium is reached, if at all. We find that social learning equilibria exist in theory, but in practice, a wide variety of environmental factors such as the number of individuals in the network, the accuracy of each individual's private signal, and the presence of fraud can impact the accuracy of the information cascades. This study of information cascades and sequential learning has applications that extend far beyond the realm of social networks. Specifically, the analyses done in this paper will complement other studies on how virality can spread economic sentiments, especially negative sentiments of "fake news," and rests at the intersection of network theory, behavioral finance, and narrative economics.