Following the Crowd: How Herding May Affect Binary Decisions

“Network”, referring to the concept information flow among agents, is ubiquitous today. The networks we looked at consist of individuals tasked with making a binary decision based on partial information about the true outcome of an experiment. This decision could have positive or negative consequences. It is up to the agents in the network, by observing others’ choices, to discover what they should do about the faced decision: either adopt the action or refrain from the action. The first network we looked at involved a sequence of agents, tasked with making one, permanent decision. Those agents could see who made what decision in front of them; this network is known as the Sequential Social Learning Model. The second network type we looked at involved several agents making simultaneous decisions. In this network each agent makes multiple decisions over time, with the ability to see some individuals in the network. This second network is known as the Social Network Model. The methods we used involved looking at each agent’s perceived probabilities of a positive and negative consequence. Furthermore, when observing the actions of others, we implemented Bayes’ Rule to further improve the initial perceived probabilities and ultimately improve the individuals’ decisions. Generally, we found that, with enough neighbors performing the same action, an agent would follow, regardless of his partial information. We followed this “herding effect” and its effects on the entire population’s probability of success.