Search papers, labs, and topics across Lattice.
This study investigates the impact of self-stigma on treatment engagement among people who use drugs (PWUD) and introduces a persona-aware approach to LLM support that recognizes the diversity of self-stigma expressions. Using Latent Profile Analysis on data from 1,174 Reddit users, the authors identify a four-persona typology that enhances the effectiveness of LLM responses, achieving a macro-F1 score of 0.74 when predicting personas from limited posting histories. Despite the success in behavioral shifts through persona-matched responses, clinical experts preferred the generic empathy of a persona-neutral baseline, highlighting a tension between targeted support and holistic empathy in LLM design.
Persona-aware LLM responses can drive behavioral change in PWUD, but clinical experts still favor generic empathy over tailored interactions.
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.
