Detecting misconception surges in opioid policy: a zero-shot LLM analysis of British Columbia's decriminalization pilot.
5 affiliations
- 1Institute of Health Policy, Management and Evaluation, Dalla Lana School of Public Health, University of Toronto, Toronto, Canada. [email protected].
- 2Faculty of Information, Toronto, Canada. [email protected].
- 3Schwartz Reisman Institute, Toronto, Canada. [email protected].
- 4Institute of Health Policy, Management and Evaluation, Dalla Lana School of Public Health, University of Toronto, Toronto, Canada.
- 5Institute of Health Policy, Management and Evaluation, Dalla Lana School of Public Health, University of Toronto, Toronto, Canada. [email protected].
Abstract
Background: In January 2023, British Columbia (BC) launched Canada’s first drug decriminalization pilot in response to persistently high opioid-related deaths. This initiative removed criminal penalties for possession of up to 2.5 grams of certain illicit drugs to reduce stigma and increase service use. Social media platforms such as Reddit provide a real-time window into public perceptions and misconceptions about such policies.
Methods: We analyzed \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}12,537\end{document} Reddit comment records posted between January 2023 and October 2024 from subreddits discussing decriminalization in British Columbia. A zero-shot large language model applied a multi-label taxonomy with four misconception classes and one no-claim class. Expert review of 547 comments with model disagreement supported selection of the primary classifier for analysis. We tracked monthly misconception prevalence and used subreddit-week reply premiums and lattice clustering to detect engagement bursts. Negative binomial models quantified reply differences across periods, regions, and comment-level harm reduction references.
Results: GPT-4o assigned at least one substantive misconception label to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}2,457\end{document} comment records (19.60%). Label 5 was assigned to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}10,080\end{document} records (80.40%), consistent with a conservative taxonomy that separates checkable policy claims from opinion, sarcasm, anecdote, and moral judgment. After the May 2024 amendment that restricted possession in public spaces, legal status and enforcement misconceptions increased. Legal or enforcement misconception records had a higher adjusted reply-rate estimate than other records, although the main effect had limited precision (IRR 1.13, 95% CI 1.00-1.27, p=0.059). The post-May interaction was positive (IRR 1.14, 95% CI 1.02-1.27, p=0.019). Reply premiums clustered into three premium bursts from late March to mid-May 2024. Harm reduction references had higher adjusted replies (IRR 1.14, 95% CI 1.07-1.22, p<0.001), while co-occurrence of harm reduction references with legal or enforcement misconceptions was associated with a smaller added reply premium (IRR 0.80, 95% CI 0.66-0.97, p=0.022).
Conclusions: Public discussion of BC’s decriminalization policy revealed persistent confusion, and legal and enforcement confusion intensified around the May 2024 revision. Burst detection on social platforms can identify when and where confusion concentrates, which supports timely public clarification and service signposting. During future amendments, concise messages about permissible locations and remaining police authority may reduce avoidable enforcement contact and misinformation.
Related
- The symbolic power of decriminalization and recriminalization in shaping stigma toward people who use drugs in British Columbia.
- Lived experiences of psychosocial and quality of life outcomes for adults newly diagnosed with breast cancer in interior British Columbia: qualitative research
- Changes in Post-Transplant Vaccine Coverage After Implementing Structured Vaccine Reviews in Pediatric Solid Organ Transplant Recipients in British Columbia, Canada: A Canadian Immunization Research Network Study.
- How relational continuity shapes recommended chronic disease management: analysis of population-based administrative health data in British Columbia
- Differences in A1C trajectory by sex and BMI in paediatric type 1 diabetes: a retrospective cohort study using a diabetes registry in British Columbia, Canada
BibTeX
@article{shakeri2026detecting,
title = {{Detecting misconception surges in opioid policy: a zero-shot LLM analysis of British Columbia's decriminalization pilot.}},
author = {Zahra Shakeri and Marwa Sharifi and Sepehr Harfi Moridani and Chan Pang Yang and Konrad Samsel and Christoffer Dharma and Mohammad Noaeen},
journal = {BMC public health},
year = {2026},
month = aug,
doi = {10.1186/s12889-026-28661-x},
url = {https://europepmc.org/article/MED/42972500},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Detecting misconception surges in opioid policy: a zero-shot LLM analysis of British Columbia's decriminalization pilot.
AU - Zahra Shakeri
AU - Marwa Sharifi
AU - Sepehr Harfi Moridani
AU - Chan Pang Yang
AU - Konrad Samsel
AU - Christoffer Dharma
AU - Mohammad Noaeen
PY - 2026
DA - 2026/08/06
JO - BMC public health
DO - 10.1186/s12889-026-28661-x
UR - https://europepmc.org/article/MED/42972500
AB - Background: In January 2023, British Columbia (BC) launched Canada’s first drug decriminalization pilot in response to persistently high opioid-related deaths. This initiative removed criminal penalties for possession of up to 2.5 grams of certain illicit drugs to reduce stigma and increase service use. Social media platforms such as Reddit provide a real-time window into public perceptions and misconceptions about such policies. Methods: We analyzed \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}12,537\end{document} Reddit comment records posted between January 2023 and October 2024 from subreddits discussing decriminalization in British Columbia. A zero-shot large language model applied a multi-label taxonomy with four misconception classes and one no-claim class. Expert review of 547 comments with model disagreement supported selection of the primary classifier for analysis. We tracked monthly misconception prevalence and used subreddit-week reply premiums and lattice clustering to detect engagement bursts. Negative binomial models quantified reply differences across periods, regions, and comment-level harm reduction references. Results: GPT-4o assigned at least one substantive misconception label to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}2,457\end{document} comment records (19.60%). Label 5 was assigned to \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}10,080\end{document} records (80.40%), consistent with a conservative taxonomy that separates checkable policy claims from opinion, sarcasm, anecdote, and moral judgment. After the May 2024 amendment that restricted possession in public spaces, legal status and enforcement misconceptions increased. Legal or enforcement misconception records had a higher adjusted reply-rate estimate than other records, although the main effect had limited precision (IRR 1.13, 95% CI 1.00-1.27, p=0.059). The post-May interaction was positive (IRR 1.14, 95% CI 1.02-1.27, p=0.019). Reply premiums clustered into three premium bursts from late March to mid-May 2024. Harm reduction references had higher adjusted replies (IRR 1.14, 95% CI 1.07-1.22, p<0.001), while co-occurrence of harm reduction references with legal or enforcement misconceptions was associated with a smaller added reply premium (IRR 0.80, 95% CI 0.66-0.97, p=0.022). Conclusions: Public discussion of BC’s decriminalization policy revealed persistent confusion, and legal and enforcement confusion intensified around the May 2024 revision. Burst detection on social platforms can identify when and where confusion concentrates, which supports timely public clarification and service signposting. During future amendments, concise messages about permissible locations and remaining police authority may reduce avoidable enforcement contact and misinformation.
SN - 1471-2458
ER -