Socio-Computational Analyses of Stigma Disclosures from People Who Use Drugs on Reddit with LLM-Based Interventions: A Thesis Proposal

Layla Bouzoubaa, Rezvaneh Rezapour

Keywords: Stigma, Substance Use, Social Media, Reddit, Natural Language Processing (NLP), Large Language Models (LLMs), Narrative Analysis, Empathetic Response Generation, Computational Social Science

TL;DR: This thesis proposes NLP methods to analyze how people who use drugs disclose and internalize stigma on Reddit, and explores LLM-based methods for generating empathetic, context-aware responses to stigmatizing language.

Abstract: This thesis proposal explores how NLP can be used to characterize, model, and mitigate the stigma associated with substance use. Drawing on social theory and lived experience, the work is organized across three aims: (1) identifying patterns of stigma expression in Reddit communities through large-scale annotation and clustering, (2) modeling internalized stigma as a temporal and narrative process using user-level affect trajectories, and (3) designing generative tools to offer stigma-responsive support. Completed studies demonstrate how theory-grounded annotation pipelines and large language models (LLMs) can surface rich typologies of stigma and self-disclosure, while proposed work extends these insights toward context-sensitive intervention. By integrating insights from public health, narrative theory, and computational social science, this research contributes new methods for understanding and responding to stigmatizing language in real-world online settings.

Archival Status: Archival

Paper Length: Long Paper (up to 8 pages of content)

Submission Number: 189