Self-Stigma Is Not a Monolith, but Generic Empathy Is: Persona-Conditioned LLM Support for People Who Use Drugs
2026-06-22 • Computation and Language
Computation and Language
AI summaryⓘ
The authors studied how people who use drugs express self-stigma online and found that there are four different types or "personas." They created a method to identify which persona someone fits based on their posts using machine learning models, which worked better than usual approaches. Experts testing responses from language models showed that matching responses to personas helped change behavior but people preferred more general empathetic replies. The authors highlight that balancing personalized support and overall empathy is tricky and needs new ways to evaluate these systems.
self-stigmapeople who use drugs (PWUD)latent profile analysispersonasnatural language processingmachine learning classifierslanguage models (LLMs)empathytreatment avoidancebehavioral health support
Authors
Layla Bouzoubaa, Rezvaneh Rezapour
Abstract
Self-stigma predicts treatment avoidance and disengagement among people who use drugs (PWUD), yet conversational systems aiming to provide support typically treat self-stigma expression as a uniform signal. We present a three-phase, proof-of-concept study of a persona-aware approach to LLM support. Latent Profile Analysis (LPA) on indicator-level features from 1,174 self-stigma expressors on Reddit yields a four-persona typology validated against held-out behavioral and linguistic features. Sequential Bayesian and recurrent neural classifiers recover these personas from limited posting histories, substantially outperforming batch and few-shot LLM baselines (macro-F1 = 0.74 at 30 posts). Evaluation by eight clinical experts across three contemporary LLMs revealed a misalignment: persona-matched responses successfully achieved targeted behavioral shifts, yet raters holistically preferred the generic empathy of the persona-neutral baseline. Our findings suggest that holistic empathy judgments and clinically-aligned response design can pull in opposite directions, and that evaluating LLM-based stigma support requires rubrics capable of decomposing the two.