Do You Believe What You Know? Stereotype Learning in the Context of Social Networks

Julia Eberlen¹, Olivier Klein¹, Matteo Gagliolo¹

“Belgians like to drink beer” is a common stereotype in- and outside of Belgium. However, we would be hard-pressed to indicate when we learned about it, as clearly, nobody is born with that knowledge. In addition, knowing about the stereotype is not the equivalent of believing it. We propose that stereotypes are the product of social learning that depends on an individual’s position in a network. We created an agent-based model in order to investigate how people might come to know about stereotypes and are currently implementing the difference between knowledge and endorsement. In the model, agents are connected in a scale-free network generated by the powerlaw cluster graph algorithm (Holme & Kim, 2002) with a clustering coefficient of 0.14. Each node represents one individual who has the ability to learn and remember. The model contains one single stereotype and learning is formalized as purely social. This means that each agent takes into account what it already knows about the stereotype in addition to the stereotype-relevant information presented by its network neighbors. While what the agent “knows” is a nuanced, continuous stereotype, what its neighbors see is different: from their point of view, the agent will show dichotomous stereotype information, namely whether or not it believes in the stereotype. This reflects people’s shortcomings in presenting their knowledge in a differentiated, nuanced manner, as well as our tendency to simplify and categorize the information surrounding us. The model is currently being tested in a number of configurations, differing in network structure, allocation of initial biases towards one or another stereotype to the nodes, and parameters of the learning rule. Ultimately, this work will give us further evidence on the impact of network structure on the formation and transmission of stereotypes.