Homophily

Homophily — from the Greek homos (same) and philos (friend) — is the principle that similar individuals tend to associate, form ties, and cluster together more frequently than dissimilar ones. Coined by sociologists Paul Lazarsfeld and Robert Merton in 1954, the concept predates network science but has become one of its foundational explanatory mechanisms. Homophily is not merely a correlation; it is a generative force that shapes social networks from the bottom up, producing assortativity by attributes such as age, ethnicity, education, religion, and political belief.

The mechanism is deceptively simple: people meet through shared activities, institutions, and spaces, and similarity breeds trust, understanding, and repeated interaction. But homophily's consequences are profound and sometimes pernicious. It produces segregated communities, echo chambers, and inequality of opportunity. A job market that operates through social networks will reproduce the demographic composition of those networks; if the networks are homophilous, advantage and disadvantage both become self-perpetuating.

Homophily operates at multiple levels simultaneously. Status homophily involves ties between people with similar socioeconomic standing. Value homophily involves ties between people with similar beliefs and attitudes. A single network can be assortative by one attribute and disassortative by another — degree and age may correlate positively while degree and geography correlate not at all.

The measurement of homophily parallels the measurement of assortativity: one computes the fraction of edges that connect same-type nodes, normalized by random expectation. High homophily implies strong community structure; zero homophily implies a randomly mixed network. But real networks almost never show zero homophily. The default social state is not diversity but clustering.

Homophily is usually discussed as a social fact to be documented and perhaps regretted. But it is also a design principle. Recommendation algorithms, dating apps, and professional networking platforms all exploit homophilous preferences because they increase engagement. The result is not neutral curation but active segregation — a system that discovers your similarities and then eliminates your exposure to difference. Homophily is not just what networks are; it is what networks are being optimized to become.