Belief dynamics driven by social influence

Abstract

Agent-based modelling (ABM) is a widespread and commonly used technique for the understanding of social system behaviour. Despite ABM being a commonly used method for the analysis of complex social systems, simulated agents still can best be described as autistic before a specific set of social rules is ascribed to them. This discrepancy leads to model outcomes that can differ from reality. For improving agent-based models on this aspect, a social practice theory-based approach can be used as a theoretical framework. The main question to be researched is: How does enabling agents to reason about others' beliefs on social practices affect dynamics of belief and behaviour formation as opposed to action adaption models, with respect to transport mode choices? Two agent-based models have been created within this study to provide an answer to this research question. Through including a case study within the transport domain, model results can be hold against macro-behaviour within the state of the art of models on transport mode choices. The case study depicts a group of students making a decision for a transport mode to move from their university faculty to a compulsory seminar. The students influence each other, leading to belief adaption. Within the transport domain, usually this social influence is translated to compared and adapted actions instead. This study differs itself by letting beliefs be the reference point for social comparison. Therefore, belief dynamics and agent behaviour for the models with belief adaption are compared to the state of the art action adaption models. Lastly, the effect of limitations towards having insight to other agents' personal characteristics is analysed. The analysis provides the following conclusions. Within the field of transport mode choices, enabling agents to reason about others' beliefs on social practices results in the same opinion dynamics and behaviour as is presented by action adaption models. However, an agent-based model that allows agents to reason about others' beliefs on social practices with limited insight in others' private characteristics through a theory of mind capacity, does provide a different reaction to an increase in the social capital of agents. More specifically, enhancing social network connectivity through increasing the group size or similarity threshold does not lead to more consensus within the theory of mind model, which is the case in the other models presented within this study. This stresses the importance of considering the extent to which are capable of forming a just conjection on others' private characteristics when designing an agent-based model with social interaction.