Abstract
A model of opinion dynamics amongst agents embedded on an adaptive social network is extended to introduce tuneable spatial embedding. As in the original model, the opinions and social connections of a population of model agents change due to three social processes: conformity, homophily and neophily. Here, however, direct interactions are constrained to take place only between pairs of agents that are linked by short spatial connections, or between pairs of agents that have benefited from some degree of random rewiring of these spatial connections. This introduction of spatiality could be expected to either reduce the ability of extreme agents to connect with one another in order to form extreme communities, or to increase their ability to influence the opinions of the community of agents clustered around them. Results demonstrate that the latter is the case. Spatial constraints tend to encourage extreme communities (relative to comparable nonspatial networks) due to the increased number and strength of distinct agent communities in spatial networks. These results suggest that the presence of strong community structure (rather than high clustering coefficients or short characteristic path lengths) may promote extremist communities in realworld populations.