Representation of latent tie preferences by group membership

Determine how latent tie-formation preferences are internally represented as a function of individuals’ group memberships across multiple identity dimensions in social networks; specifically, ascertain whether these preferences depend on full multidimensional source–target group memberships or decompose into independent one-dimensional components.

Background

The paper argues that prior network studies often make strong simplifying assumptions about homophily and out-group preferences, which can obscure actual latent tendencies. Because attributes such as race, gender, and socioeconomic status can be correlated, observed connection patterns may conflate opportunity with preference, complicating inference.

To resolve this, the authors develop a framework that allows multiple plausible structures of latent preferences—from fully multidimensional to separable one-dimensional forms—and propose to identify which structure best explains empirical networks. The quoted sentence frames the foundational unresolved issue the framework seeks to address.