Acquisition of Social Network Graph Structure

Abstract

Two hypotheses concerning the acquisition of network graph structures were tested. Strong support was found for the hypothesis that the deep structure of a graph affects how quickly it will be learned. Specifically, scale-free graphs were acquired more readily than random graphs. Much less support was found for the hypothesis that the acquisition of social networks is exceptional. The results suggest that the surface description given a graph does not affect how quickly it will be learned. Given similar stimuli, computer or transportation networks were learned just as efficiently as social networks. Evidence is also presented that subjects bring a set of expectations to the task of graph structure acquisition. They exhibit a bias toward responding yes when asked if nodes are connected. In social networks, they assume that intragender friendships are more common than intergender friendships. Acquisition is positively correlated with graph transitivity and negatively correlated with connectedness and efficiency.

Keywords: learning, social networks