Navigating the Rabbit Hole: Emergent Biases in LLM-Generated Attack Narratives Targeting Mental Health Groups

2025-04-08

AI Summary

This study reveals that large language models (LLMs) spontaneously generate stigmatizing and adversarial narratives targeting mental health populations—even in zero-shot settings—exacerbating bias and harm toward high-risk groups. It presents the first systematic audit of LLMs’ spontaneous adversarial behavior toward mental health entities. Methodologically, it introduces an interdisciplinary bias propagation framework integrating network centrality analysis (closeness centrality; *p* = 4.06×10⁻¹⁰) and Gini-coefficient-based clustering (*G* = 0.7), grounded in sociological stigma theory to quantify labeling effects. Results show that mental health–related entities occupy statistically significant central positions within adversarial narrative networks, and labeling intensity progressively amplifies along attack chains. The work advances LLM bias evaluation through a rigorous, cross-disciplinary methodology and facilitates the substantive implementation of AI ethics in safeguarding vulnerable populations.

Technology Category

• Natural Language Processing: Ethics — Bias, Fairness, Transparency & Privacy
• Machine Learning: Ethics, Bias, and Fairness
• Multiagent Systems: Adversarial Agents

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Large Language Models (LLMs) have been shown to demonstrate imbalanced biases against certain groups. However, the study of unprovoked targeted attacks by LLMs towards at-risk populations remains underexplored. Our paper presents three novel contributions: (1) the explicit evaluation of LLM-generated attacks on highly vulnerable mental health groups; (2) a network-based framework to study the propagation of relative biases; and (3) an assessment of the relative degree of stigmatization that emerges from these attacks. Our analysis of a recently released large-scale bias audit dataset reveals that mental health entities occupy central positions within attack narrative networks, as revealed by a significantly higher mean centrality of closeness (p-value = 4.06e-10) and dense clustering (Gini coefficient = 0.7). Drawing from sociological foundations of stigmatization theory, our stigmatization analysis indicates increased labeling components for mental health disorder-related targets relative to initial targets in generation chains. Taken together, these insights shed light on the structural predilections of large language models to heighten harmful discourse and highlight the need for suitable approaches for mitigation.

Problem

• Evaluates LLM-generated attacks on vulnerable mental health groups
• Studies bias propagation in attack narratives using network analysis
• Assesses stigmatization levels in LLM-generated mental health attacks

Innovation

• Evaluates LLM attacks on mental health groups
• Uses network framework for bias propagation
• Assesses stigmatization degree from LLM attacks