Aggregation mechanism for multidimensional identity preferences

Identify the aggregation mechanism by which individuals integrate one-dimensional latent preferences across multiple identity dimensions into an overall tie-formation probability, discriminating among candidate rules such as weighted averaging, independent conjunctive evaluation (product/AND), or disjunctive evaluation (OR).

Background

There is limited systematic evidence on how people combine preferences across multiple identity dimensions when deciding to form social ties. Crossed-categorization studies suggest additive combination, while common network models implicitly assume multiplicative aggregation, but the underlying behavioral mechanism remains contested.

The authors formalize several candidate aggregation functions and compare them empirically using Bayesian model selection. The quoted sentence states the long-standing open question motivating this comparison.

References

Finding the information aggregation mechanism people use in interactions is a long-standing open question that has been formulated in different ways.

— The hidden architecture of connections: How do multidimensional identities shape our social networks?  (2406.17043 - Martin-Gutierrez et al., 2024) in Section 2.2: Modeling preference aggregation mechanisms

As we will discuss below, a question for further research is what a ``sufficient" sharing of identity markers is.

— Going Beyond the In-/Out-Group Dichotomy: Investigating Altruism towards Middle-Groups  (2609.09046 - Houf et al., 8 Sep 2026) in Section 4, subsection “T2 (2-3_{P,U}): Introduction of a Middle-Group: A Clear Hierarchy of Groups Emerges”

How important is each single, isolated identity marker? Which role do markers from lived experience play? How are their conjoint effects? Is there a tipping point in markers where a neutral' behaviour turnshostile', or would it be a gradual transition?

— Going Beyond the In-/Out-Group Dichotomy: Investigating Altruism towards Middle-Groups  (2609.09046 - Houf et al., 8 Sep 2026) in Section 5, Discussion and Conclusion