Choice Polarization on a Social Influence Network

We are studying here a model of social influence in which actors are dwelling upon a social network and are influencing/influenced by each other through the recurrent and reproduced pattern of their relationships. Furthermore, we assume that influence outcomes represent preference orderings (among a number of alternatives) and, thus, they are located on a certain relational structure (a graph of orderings). Because of the structural arrangement of influence outcomes, actors’ orderings may occupy antipodal positions. In this setting, after following a social influence process of random dyadic interactions among actors, the system is stabilized (absorbed) by two possible configurations of actors’ influence outcomes: either a unanimous arrangement of orderings or a polarization across dipoles of antipodal orderings. We conduct computer simulations in order to study how this process of polarized social influence depends on certain parameters of the system (such as the size and connectivity of the social network) and the degree of its randomization. We find that polarization appears to be unexpectedly high over social networks of the ‘small world’ type. Furthermore, we discuss the relevance of this model of polarized social influence on relationally structured outcomes with Arrow’s Impossibility Theorem and we outline some possible further extensions of this model.