On the Influence of Gender and Race in Romantic Relationship Prediction from Large Language Models

TL;DR

This paper investigates biases related to gender and race in large language models' predictions of romantic relationships, revealing disparities influenced by name ethnicity and gender, with implications for social equity.

Contribution

It uncovers biases in language models regarding romantic relationship predictions based on gender and race, highlighting the need for more inclusive AI development.

Findings

Models less likely to predict same-gender relationships.

Models show bias against interracial pairs involving Asian names.

Asian names' gender features are less distinguishable in embeddings.

Abstract

We study the presence of heteronormative biases and prejudice against interracial romantic relationships in large language models by performing controlled name-replacement experiments for the task of relationship prediction. We show that models are less likely to predict romantic relationships for (a) same-gender character pairs than different-gender pairs; and (b) intra/inter-racial character pairs involving Asian names as compared to Black, Hispanic, or White names. We examine the contextualized embeddings of first names and find that gender for Asian names is less discernible than non-Asian names. We discuss the social implications of our findings, underlining the need to prioritize the development of inclusive and equitable technology.