A network-level measure of the tendency for nodes to connect preferentially to others that are similar with respect to a chosen attribute; typically reported as a scalar r ∈ [−1,1] (Newman's assortativity) where r > 0 indicates assortative mixing (similar nodes connect), r < 0 indicates disassortative mixing, and r = 0 indicates no assortative preference. Different formulations apply for categorical versus scalar attributes.

Principle

Positive assortativity reflects homophily or structurally constrained mixing by the chosen attribute and tends to produce assortatively clustered subnetworks that affect diffusion, segregation and resilience properties of the network.

Demonstration

Illustrative scenario: A social network is annotated by age group. Computation of the assortativity coefficient by age yields r = 0.6 (positive). Recognition: friendships are concentrated within age cohorts. Action: public-health messaging that initially targets one cohort spreads mainly within that cohort before crossing to others. Consequence: diffusion speed and reach across age groups are modulated by the observed assortativity.

Misapplication

Interpreting observed assortativity as evidence of individual preference without considering alternative mechanisms (opportunity structure, forced matching, network sampling bias). The semantic error is conflating pattern (mixing) with mechanism (homophily as choice).

Consequence

Assortativity shapes network segregation, the formation of echo chambers, uneven diffusion across attribute categories and the network's structural robustness; it therefore conditions the effectiveness of interventions that rely on cross-attribute bridging.

Reversal

Assortativity computed for one attribute can coexist with disassortativity on another (for example assortative by education but disassortative by occupation), producing complex mixing regimes; moreover, in temporal or multiplex networks assortativity values can change and have different implications across layers or timescales.

Boundary

Clearly within: the network-level scalar r computed according to a specified attribute and formula. Boundary case: small sample networks yield unstable r estimates requiring caution. Clearly outside: node-level homophily indices or correlation between attribute and centrality—related but distinct measures that do not substitute for network-level assortativity without reformulation.

Semantic Tension

Tension between homophily-driven cohesion (assortativity) and the normative or instrumental value of cross-group connectivity and diversity; policies promoting bridging may reduce assortativity while increasing cross-cutting exposure.

Synthesis

The assortativity coefficient quantifies structural mixing by an attribute and predicts how attribute-aligned clusters form and influence diffusion and segregation, but it is silent on causal mechanisms and must be interpreted alongside opportunity structure, sampling and temporal dynamics.