AI Will Always Love You: Studying Implicit Biases in Romantic AI Companions

2025-02-27

AI Summary

This work investigates overlooked implicit gender bias in romantic AI companions, specifically examining how persona instantiation triggers stereotypical gender-role responses. We systematically evaluate multiple large language models (LLMs) under controlled prompts using three behavioral experiments: an adapted Implicit Association Test (IAT), sentiment lexicon analysis, and quantification of sycophantic tendencies—each applied to gendered and romantic personas. Our study is the first to uncover the emergence mechanism of persona-driven implicit bias in intimate contexts and proposes a novel bias measurement framework tailored to affective human-AI interaction. Empirical results demonstrate that gendered personas significantly distort model outputs, reinforcing traditional gender roles; moreover, bias intensity varies non-monotonically with model scale. These findings provide both theoretical foundations and methodological tools for fairness-aware design and evaluation of AI companions.

Technology Category

• Humans and AI: Emotional Intelligence
• Philosophy and Ethics of AI: Bias, Fairness & Equity
• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

While existing studies have recognised explicit biases in generative models, including occupational gender biases, the nuances of gender stereotypes and expectations of relationships between users and AI companions remain underexplored. In the meantime, AI companions have become increasingly popular as friends or gendered romantic partners to their users. This study bridges the gap by devising three experiments tailored for romantic, gender-assigned AI companions and their users, effectively evaluating implicit biases across various-sized LLMs. Each experiment looks at a different dimension: implicit associations, emotion responses, and sycophancy. This study aims to measure and compare biases manifested in different companion systems by quantitatively analysing persona-assigned model responses to a baseline through newly devised metrics. The results are noteworthy: they show that assigning gendered, relationship personas to Large Language Models significantly alters the responses of these models, and in certain situations in a biased, stereotypical way.

Problem

• Examining implicit biases in romantic AI companions
• Evaluating gender stereotypes in AI relationships
• Assessing biased responses in gendered AI personas

Innovation

• evaluates implicit biases
• uses gender-assigned AI companions
• devises new bias metrics