SSR: Can Simulated Patients Learn to Stigmatize Themselves? Modeling Self-Stigma through Internal Monologue

2026-06-06

AI Summary

Current large language model–driven patient simulations struggle to authentically reproduce the dynamic behaviors—such as avoidance, denial, or self-blame—that arise from internalized stigma in individuals with mental health conditions, often resulting in overly static interactions. To address this limitation, this work proposes a novel patient simulation framework grounded in the psychological 3A1H model of self-stigma. The approach integrates stigma-aware inner monologues into dialogues and leverages a newly curated dataset featuring Stigma-Sensitive Reflections (SSR) alongside chain-of-thought–based fine-tuning to enable context-sensitive modeling of self-stigma. Experimental results demonstrate that the proposed method significantly outperforms existing baselines, generating patient responses that are both more realistic and adaptive, thereby offering a new paradigm for clinical training and empathetic dialogue systems.

Abstract

Simulating patients with large language models (LLMs) is a promising tool for mental health training, but existing approaches fail to capture a key clinical reality: self-stigma. Patients experiencing self-stigma, the internalization of negative stereotypes, often exhibit context-sensitive resistance, such as avoidance, denial, or self-blame, which current models render as static or uniformly compliant behavior. To address this, we introduce a novel simulation framework grounded in the psychological 3A1H model of self-stigmatization. Our core innovation is the creation of a \textbf{Stigmatized Self-Reflection} (\textbf{SSR}) dataset, where we augment mental health dialogues with internal monologues that reflect stigma-aware reasoning. By fine-tuning LLMs with this data using a chain-of-thought approach, we train patient agents to dynamically adjust their level and expression of stigma based on conversational triggers. Evaluations demonstrate that our approach significantly outperforms specialized baselines, generating more authentic and situationally appropriate patient responses. This work provides a crucial step towards realistic stigma simulation for clinical training and empathetic dialogue systems.

Problem

• self-stigma
• simulated patients
• mental health training
• internalized stereotypes
• stigmatization

Innovation

• self-stigma
• internal monologue
• large language models
• mental health simulation
• chain-of-thought