"Tell me who you hang out with, and I will tell you who you are": A Bayesian approach to inferring leadership profiles in social systems

Abstract

The popular proverb "tell me who you hang out with, and I will tell you who you are" captures an accurate statistical intuition regarding homophily in social structures. This paper mathematically formalizes this adage by proposing a probabilistic model for the classification of social leaders through latent variable analysis. Using a Bayesian approach, we demonstrate how the composition of a sample of followers enables the inference of the probability that a leader belongs to a specific ethical or ideological category. The model provides a methodological framework for understanding belief updating under uncertainty and the impact of system noise on the reliability of follower networks.