Measuring the Degrees of Separation of a group of Minds

Abstract

The degrees of separation are essential in the study of social networks. We develop a graph theoretical methodology inspired in this concept to study behavioral data. Having a similarity measure appropriate to the data, we propose to build a graph connecting each agent its nearest n neighbors, having that their similarity exceeds a threshold p. From here, graph theoretical indicators such as connectivity and clustering can be studied as functions of p and n. We apply this methodology to a psychological experiment where 97 participants estimate the typicality of 8 objects with respect to the concept “Hat”, in 2 contexts. Using correlation distance as a similarity function, we compute clustering properties, and average path length as functions of n and p. Interestingly, only n=20 connections are required to keep the similarity structure of the system, having an average path length near 2.


Back to Table of Contents