An agent model of social network and travel behavior interdependence

Travel is a prerequisite for activities which maintain social and business connections, building the vital social networks which conduct the flow of values, services, and opportunity. This paper presents a multi-agent simulation to study linked geographical and social spaces. The model simultaneously generates a social network and travel behavior by defining social-networking visits as travel activities. Information about space and other agents flows only via the social network. Social ties are added/removed depending on co-presence/lack of visits, introducing the dynamic feedback to the model. A random utility model (RUM) trades off socializing with the cost of travel and the activity attributes to generate probabilities for link attachments. The focus on link attachment probabilities allows the model to be verified within the analytical framework of random, small-world, and exponential (preferential attachment) graphs. New measures are suggested to compare social and geographical space. The model response to changes in utility parameters is sometimes unexpected due to the feedback. Ensemble results of a base case and sensitivity tests are presented.