Elamur Research⌕
☰
home / papers / 2026 / Can LLMs Understand the Impact of Trauma? Costs and Benefits of LLMs Coding the Interviews of Firearm Violence Survivors

Can LLMs Understand the Impact of Trauma? Costs and Benefits of LLMs Coding the Interviews of Firearm Violence Survivors

Paper · 2026 · Jessica H Zhu, Shayla Stringfield, Vahe Zaprosyan et al.

Summary

This study asks whether small, open-source LLMs can reliably perform inductive qualitative coding of long, sensitive interviews. The interviews come from 21 Black men who survived community firearm violence. They were interviewed in 2013–2015 after admission to a DC-area trauma center. The authors argue that firearm violence research is chronically underfunded, which makes cheap automation appealing. However, commercial AI coding tools have mostly been tested on short, standard-English texts and rely on large closed models.

The team built a machine coding pipeline using Llama models run locally to protect the data. They tried several ways of splitting transcripts into chunks and compared the machine-generated codes with codes from Black women human coders who used grounded theory.

Some configurations captured important codes, but overall relevance was low and depended heavily on preprocessing. The models also fixated on African American English colloquialisms. Safety guardrails caused substantial narrative erasure, meaning the models silenced accounts of violence.

The work shows that current LLMs cannot replace human coders for marginalized communities. It calls for low-resource tools that are fair and reliable.

Key points

  • Models and hardware: The pipeline used full-precision Llama 3.2 1B and Llama 3.1 8B on university servers. Llama 1B needed about 6GB of VRAM versus 34GB for 8B, yet came close to 8B's performance, though less reliably.
  • Preprocessing strategies: Transcripts were split into chunks of up to 256 tokens in three ways:
    • "paired chunks," pairing each interviewer turn with the subject's reply;
    • "question chunks," matching responses to protocol questions using concatenated all-mpnet-base-v2 and multiqa-MiniLM-L6-cos-v1 embeddings with a 20% cosine-similarity threshold;
    • "full text," passing whole interviews, tested only for the 1B model.
  • Time trade-off: Human coding took over 10 times as many hours, and the automated pipeline achieved high Percent Captured with roughly 12 GPU hours. However, validating the machine codes eroded much of that saving.
  • Validation support: Data tracking combined with interpretable clustering such as BERTopic can shorten expert validation of the machine codes.
  • Availability: Code is released at github.com/jhzsquared/AIvsHumanCoding. The interview data is withheld to protect participants.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of Maryland, College ParkCollege Park · United States132 works · 25,380 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Arts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSafety Research · 543 worksGeneral Social Sciences · 227 worksEducation · 203 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Jessica H Zhu, Shayla Stringfield, Vahe Zaprosyan, Michael Wagner, Michel Cukier and Joseph Richardson. 2026. Can LLMs Understand the Impact of Trauma? Costs and Benefits of LLMs Coding the Interviews of Firearm Violence Survivors. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.591

@inproceedings{zhu2026can,
  title = {{Can LLMs Understand the Impact of Trauma? Costs and Benefits of LLMs Coding the Interviews of Firearm Violence Survivors}},
  author = {Jessica H Zhu, Shayla Stringfield, Vahe Zaprosyan, Michael Wagner, Michel Cukier and Joseph Richardson},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.591},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.591}
}

References

30 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sirwan Khalid Ahmed, Ribwar Arsalan Mohammed, Abdulqadir J. Nashwan, Radhwan Hussein Ibrahim, Araz Qadir Abdalla, Barzan Mohammed M. Ameen, and Renas Mohammed Khdhir. 2025. Using thematic analysis in qualitative research. Journal of Medicine, Surgery, and Public Health, 6:100198. in human-AI collabo… find
  2. Manuel J. Galvan and B. Keith Payne. 2024. Implicit bias as a cognitive manifestation of systemic racism. Daedalus, 153(1):106–122. find
  3. Jie Gao, Yuchen Guo, Gionnieve Lim, Tianqin Zhang, Zheng Zhang, Toby Jia-Jun Li, and Simon Tangi Perrault. 2024. Collabcoder: A lower-barrier, rigorous workflow for inductive collaborative qualitative analysis with large language models. In Proceedings of the 2024 CHI Conference on Human Factors in… find
  4. Simret Araya Gebreegziabher, Zheng Zhang, Xiaohang Tang, Yihao Meng, Elena L. Glassman, and Toby Jia-Jun Li. 2023. PaTAT: Human-AI collaborative qualitative coding with explainable interactive rule synthesis. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems. ACM. find
  5. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Atlas.ti. Quickly gain insights with AI-powered Abhinav Pandey, Abhishek Kadian, Ahmad Al-tools. — The Llama 3 Herd of Models https://atlasti.com/trainings/ Dahle, Aiesha Letman, Akhil Mathur, Alan Schel-quickly-gain-insights-with-ai-powered-tools. ten, Alex Vaughan, Amy Yang,… ↗
  6. Chip Brownlee. 2026. Trump has slashed federal Valentin Hofmann, Pratyusha Ria Kalluri, Dan Jurafsky, funding for gun violence prevention. but how much and Sharese King. 2024. AI generates covertly racist exactly? https://www.thetrace.org/2026/01/ decisions about people based on their dialect. Natur… ↗
  7. Kevin Crowston, Eileen E. Allen, and Robert Heckman. 2012. Using natural language processing technology for qualitative data analysis. International Journal of Social Research Methodology, 15(6):523–543. find
  8. Shih-Chieh Dai, Aiping Xiong, and Lun-Wei Ku. 2023. LLM-in-the-loop: leveraging large language model for thematic analysis. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 9993–10001. Association for Computational Linguistics. find
  9. Gregory Dam and Stefan Kaufmann. 2008. Computer assessment of interview data using latent semantic analysis. Behavior Research Methods, 40(1):8–20. find
  10. DontPlanToEnd. UGI leaderboard: Uncensored general intelligence. https://huggingface.co/ spaces/DontPlanToEnd/UGI-Leaderboard. Accessed: 2025-12-20. ↗
  11. Jessica L. Feuston and Jed R. Brubaker. 2021. Putting tools in their place: The role of time and perspective Johns Hopkins. 2024. New report highlights U.S. 2022 gun-related deaths: Firearms remain leading cause of death for children and teens, and disproportionately affect people of color. https://… find
  12. Robert P Lennon, Robbie Fraleigh, Lauren J Van Scoy, Aparna Keshaviah, Xindi C Hu, Bethany L Snyder, Erin L Miller, William A Calo, Aleksandra E Zgier-ska, and Christopher Griffin. 2021. Developing and testing an automated qualitative assistant (AQUA) to support qualitative analysis. Family Medicine… find
  13. Victoria R Li, Yida Chen, and Naomi Saphra. 2024. ChatGPT doesn’t trust chargers fans: Guardrail sensitivity in context. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 6327–6345, Miami, Florida, USA. Association for Computational Linguistics. find
  14. Megh Marathe and Kentaro Toyama. 2018. Semi-automated coding for qualitative research. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems. ACM. find
  15. Bruce Sherin. 2013. A computational study of commonsense science: An exploration in the automated analysis of clinical interview data. Journal of the Learning Sciences, 22(4):600–638. find
  16. MAXQDA. AI assist: AI-powered qualitative data analysis with MAXQDA. https://www.maxqda.com/ products/ai-assist. Accessed: 2025-12-20. ↗
  17. Satpreet Harcharan Singh, Kevin Jiang, Kanchan Bhasin, Ashutosh Sabharwal, Nidal Moukaddam, and Ankit Patel. 2024. RACER: An llm-powered methodology for scalable analysis of semi-structured mental health interviews. In Proceedings of the 1st Workshop on NLP for Science (NLP4Science), pages 73–98. As… find
  18. Varun Nagaraj Rao, Eesha Agarwal, Samantha Dalal, Dana Calacci, and Andrés Monroy-Hernández. 2025. QuaLLM: An llm-based framework to extract quantitative insights from online forums. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 1355–1369. Association for Computatio… find
  19. OpenAI. 2025. ChatGPT \| the intelligence age. https://www.maxqda.com/products/ai-assist. Accessed: 2025-12-20. ↗
  20. Angelina Parfenova, Andreas Marfurt, Jürgen Pfeffer, and Alexander Denzler. 2025. Text annotation via inductive coding: Comparing human experts to llms in qualitative data analysis. In Findings of the Association for Computational Linguistics: NAACL 2025, pages 6456–6469. Association for Computation… find
  21. Desmond Upton Patton, Aviv Y. Landau, and Siva Mathiyazhagan. 2023. ChatGPT for social work science: Ethical challenges and opportunities. Journal of the Society for Social Work and Research, 14(3):553–562. find
  22. Jaspreet Ranjit, Hyundong J. Cho, Claire J. Smerdon, Yoonsoo Nam, Myles Phung, Jonathan May, John R. Blosnich, and Swabha Swayamdipta. 2025. Uncovering intervention opportunities for suicide prevention with language model assistants.
  23. Nils Reimers and Iryna Gurevych. 2020. Making monolingual sentence embeddings multilingual using knowledge distillation. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics.
  24. Tim Rietz and Alexander Maedche. 2021. Cody: An AI-based system to semi-automate coding for qualitative research. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. ACM. find
  25. Maria Y. Rodriguez and Heather Storer. 2019. A computational social science perspective on qualitative data exploration: Using topic models for the descriptive analysis of social media data. Journal of Technology in Human Services, 38(1):54–86. find
  26. Angelina Wang, Michelle Phan, Daniel E. Ho, and Sanmi Koyejo. 2025. Fairness through difference awareness: Measuring Desired group discrimination in LLMs. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 6867–6893, Vienna, Aust… find
  27. Michael Williams and Tami Moser. 2019. The art of coding and thematic exploration in qualitative research. International Management Review, 15:45. find
  28. Mian Zhong, Pristina Wang, and Anjalie Field. 2025. Hicode: Hierarchical inductive coding with llms. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 31048–31066. Association for Computational Linguistics. find
  29. Jessica Zhu, Michel Cukier, and Jr Richardson, Joseph. 2024. Nutrition facts, drug facts, and model facts: putting AI ethics into practice in gun violence research. Journal of the American Medical Informatics Association, 31(10):2414–2421. find
  30. Caleb Ziems, William Held, Omar Shaikh, Jiaao Chen, Zhehao Zhang, and Diyi Yang. 2024. Can large language models transform computational social science? Computational Linguistics, 50(1):237–291.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhu-2026-can-llms-understand-impact-trauma/