{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,28]],"date-time":"2026-07-28T06:00:20Z","timestamp":1785218420646,"version":"3.55.0"},"reference-count":58,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,7,28]],"date-time":"2026-07-28T00:00:00Z","timestamp":1785196800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>Large language models (LLMs) have been proposed as decision support tools in medicine, yet their role in forensic cause of death analysis remains unexplored.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>In this study, we used 118 real-world cases spanning diverse categories of death to systematically evaluate the performance of four representative LLMs (GPT-4o, OpenAI o3, Gemini-2.5pro, and DeepSeek-R1) in forensic cause of death analysis. Two senior forensic pathologists independently evaluated each model\u2019s decision-making capabilities regarding inference quality and conclusion accuracy. These metrics were assessed using an expert scoring system with a 5-point Likert scale, with original analytical statements and legally valid expert opinions serving as objective gold standards. In a sub-study, we examined the application potential of the locally deployed open-source model DeepSeek-R1:32b. Additionally, a targeted retrospective analysis was conducted to quantify the incidence and typologies of AI hallucinations.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      DeepSeek-R1 demonstrated a statistically significant advantage in inference quality scores over GPT-4o (\n                      <jats:italic>p<\/jats:italic>\n                      \u202f=\u202f0.015,\n                      <jats:italic>r<\/jats:italic>\n                      <jats:sub>rb<\/jats:sub>\n                      \u202f=\u202f0.28) and Gemini-2.5pro (\n                      <jats:italic>p<\/jats:italic>\n                      \u202f=\u202f0.000003,\n                      <jats:italic>r<\/jats:italic>\n                      <jats:sub>rb<\/jats:sub>\n                      \u202f=\u202f0.46), while no statistically significant differences were observed among the four models in terms of conclusion accuracy scores. The locally deployed DeepSeek-R1:32b model also showed no statistically significant difference from GPT-4o in conclusion accuracy scores. However, hallucinations persistently appear in the response reports of all LLMs.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Discussion<\/jats:title>\n                    <jats:p>LLMs can provide limited auxiliary value in cause of death analysis but should not replace the final judgment of forensic experts. LLMs still require expert oversight to ensure evidence integrity and mitigate risks such as hallucination. Open source LLMs can further mitigate data privacy concerns and provide practical support for cause of death analysis.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/frai.2026.1752043","type":"journal-article","created":{"date-parts":[[2026,7,28]],"date-time":"2026-07-28T05:37:26Z","timestamp":1785217046000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Can large language models serve as consultants for forensic cause of death analysis? A multidimensional evaluation"],"prefix":"10.3389","volume":"9","author":[{"given":"Enhao","family":"Fu","sequence":"first","affiliation":[{"name":"Academy of Forensic Science, Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Key Laboratory of Forensic Medicine, Ministry of Justice","place":["Shanghai, China"]},{"name":"School of Basic Medicine and Forensic Medicine, Institute of Medical Aspects of Specific Environments, Henan University of Science and Technology, Forensic Medicine Identification Center","place":["Luoyang, Henan, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Haojie","family":"Qin","sequence":"additional","affiliation":[{"name":"School of Basic Medicine and Forensic Medicine, Institute of Medical Aspects of Specific Environments, Henan University of Science and Technology, Forensic Medicine Identification Center","place":["Luoyang, Henan, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhiling","family":"Tian","sequence":"additional","affiliation":[{"name":"Academy of Forensic Science, Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Key Laboratory of Forensic Medicine, Ministry of Justice","place":["Shanghai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hewen","family":"Dong","sequence":"additional","affiliation":[{"name":"Academy of Forensic Science, Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Key Laboratory of Forensic Medicine, Ministry of Justice","place":["Shanghai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Donghua","family":"Zou","sequence":"additional","affiliation":[{"name":"Academy of Forensic Science, Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Key Laboratory of Forensic Medicine, Ministry of Justice","place":["Shanghai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaotian","family":"Yu","sequence":"additional","affiliation":[{"name":"Academy of Forensic Science, Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Key Laboratory of Forensic Medicine, Ministry of Justice","place":["Shanghai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ningguo","family":"Liu","sequence":"additional","affiliation":[{"name":"Academy of Forensic Science, Shanghai Key Laboratory of Forensic Medicine, Shanghai Forensic Service Platform, Key Laboratory of Forensic Medicine, Ministry of Justice","place":["Shanghai, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,7,28]]},"reference":[{"key":"ref1","first-page":"4232","author":"Ahuja","year":"2023"},{"key":"ref2","year":"2025"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"30","DOI":"10.1186\/s41935-025-00445-1","article-title":"Assessing ChatGPT-4\u2019s ability to generate forensic reports: a study of artificial intelligence in forensics","volume":"15","author":"Aydogan","year":"2025","journal-title":"Egypt. J. Forensic Sci."},{"key":"ref4","first-page":"3","volume-title":"Teoria statistica delle classi e calcolo delle probabilita","author":"Bonferroni","year":"1936"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"33","DOI":"10.1007\/s10916-023-01925-4","article-title":"Evaluating the feasibility of ChatGPT in healthcare: an analysis of multiple clinical and research scenarios","volume":"47","author":"Cascella","year":"2023","journal-title":"J. Med. Syst."},{"key":"ref6","year":"2025"},{"key":"ref7","article-title":"Chatbot arena: an open platform for evaluating LLMs by human preference","author":"Chiang","year":"2024"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"301814","DOI":"10.1016\/j.fsidi.2024.301814","article-title":"Exploring the potential of large language models for author profiling tasks in digital text forensics","volume":"50","author":"Cho","year":"2024","journal-title":"Forensic Sci. Int. Digit. Investig."},{"key":"ref9","doi-asserted-by":"publisher","first-page":"545","DOI":"10.1080\/13683500.2023.2300032","article-title":"Understanding the role and impact of generative artificial intelligence (AI) hallucination within consumers\u2019 tourism decision-making processes","volume":"28","author":"Christensen","year":"2025","journal-title":"Curr. Issue Tour."},{"key":"ref10","doi-asserted-by":"publisher","first-page":"444","DOI":"10.1007\/s10956-023-10039-y","article-title":"Examining science education in ChatGPT: an exploratory study of generative artificial intelligence","volume":"32","author":"Cooper","year":"2023","journal-title":"J. Sci. Educ. Technol."},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2501.12948","article-title":"DeepSeek-R1: incentivizing reasoning capability in LLMs via reinforcement learning","author":"Guo","year":"2025","journal-title":"arXiv"},{"key":"ref12","doi-asserted-by":"publisher","first-page":"1082","DOI":"10.3390\/diagnostics14111082","article-title":"Chatgpt\u2019s efficacy in queries regarding polycystic ovary syndrome and treatment strategies for women experiencing infertility","volume":"14","author":"Devranoglu","year":"2024","journal-title":"Diagnostics"},{"key":"ref13","doi-asserted-by":"publisher","first-page":"275","DOI":"10.1093\/fsr\/owad039","article-title":"ChatGPT in forensic sciences: a new pandora\u2019s box with advantages and challenges to pay attention","volume":"8","author":"Dinis-Oliveira","year":"2023","journal-title":"Forensic Sci Res"},{"key":"ref14","doi-asserted-by":"publisher","first-page":"111","DOI":"10.1007\/s12599-023-00834-7","article-title":"Generative AI","volume":"66","author":"Feuerriegel","year":"2024","journal-title":"Bus. Inf. Syst. Eng."},{"key":"ref15","doi-asserted-by":"publisher","first-page":"13","DOI":"10.1038\/d41586-025-00229-6","article-title":"China\u2019s cheap, open AI model DeepSeek thrills scientists","volume":"638","author":"Gibney","year":"2025","journal-title":"Nature"},{"key":"ref16","doi-asserted-by":"publisher","first-page":"20230023","DOI":"10.1259\/bjr.20230023","article-title":"AI pitfalls and what not to do: mitigating bias in AI","volume":"96","author":"Gichoya","year":"2023","journal-title":"Br. J. Radiol."},{"key":"ref17","year":"2025"},{"key":"ref18","doi-asserted-by":"publisher","first-page":"150","DOI":"10.1177\/00258024231191829","article-title":"ChatGPT: forensic, legal, and ethical issues","volume":"64","author":"Guleria","year":"2024","journal-title":"Med. Sci. Law"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"4019","DOI":"10.1016\/j.csbj.2024.10.050","article-title":"Assessment of large language models (LLMs) in decision-making support for gynecologic oncology","volume":"23","author":"Gumilar","year":"2024","journal-title":"Comput. Struct. Biotechnol. J."},{"key":"ref20","doi-asserted-by":"publisher","first-page":"111807","DOI":"10.1016\/j.forsciint.2023.111807","article-title":"Mapping AI-ethics\u2019 dilemmas in forensic case work: to trust AI or not?","volume":"350","author":"Hefetz","year":"2023","journal-title":"Forensic Sci. Int."},{"key":"ref21","doi-asserted-by":"publisher","first-page":"284","DOI":"10.1089\/cyber.2024.0407","article-title":"Generative artificial intelligence and misinformation acceptance: an experimental test of the effect of forewarning about artificial intelligence hallucination","volume":"28","author":"Hwang","year":"2025","journal-title":"Cyberpsychol. Behav. Soc. Netw."},{"key":"ref22","year":"2025"},{"key":"ref23","doi-asserted-by":"publisher","first-page":"114","DOI":"10.1177\/00258024221102203","article-title":"Current state of cause of death determinations in Japan and the need to list the precise underlying cause of death","volume":"63","author":"Inoue","year":"2023","journal-title":"Med. Sci. Law"},{"key":"ref24","doi-asserted-by":"publisher","first-page":"1553","DOI":"10.1002\/ksa.12628","article-title":"DeepSeek versus ChatGPT: multimodal artificial intelligence revolutionizing scientific discovery. From language editing to autonomous content generation\u2014redefining innovation in research and practice","volume":"33","author":"Kayaalp","year":"2025","journal-title":"Knee Surg. Sports Traumatol. Arthrosc."},{"key":"ref25","doi-asserted-by":"publisher","first-page":"301939","DOI":"10.1016\/j.fsidi.2025.301939","article-title":"Digital forensics in law enforcement: a case study of LLM-driven evidence analysis","volume":"54","author":"Kim","year":"2025","journal-title":"Forensic Sci. Int. Digit. Investig."},{"key":"ref26","doi-asserted-by":"publisher","first-page":"159","DOI":"10.2307\/2529310","article-title":"The measurement of observer agreement for categorical data","volume":"33","author":"Landis","year":"1977","journal-title":"Biometrics"},{"key":"ref27","doi-asserted-by":"publisher","first-page":"100221","DOI":"10.1016\/j.teler.2025.100221","article-title":"Generative AI risks and resilience: how users adapt to hallucination and privacy challenges","volume":"19","author":"Lee","year":"2025","journal-title":"Telemat. Inform. Rep."},{"key":"ref28","doi-asserted-by":"publisher","first-page":"110382","DOI":"10.1016\/j.clinimag.2024.110382","article-title":"Comparative diagnostic accuracy of GPT-4o and LLaMA 3-70b: proprietary vs. open-source large language models in radiology","volume":"118","author":"Li","year":"2025","journal-title":"Clin. Imaging"},{"key":"ref29","doi-asserted-by":"publisher","first-page":"21133","DOI":"10.1038\/s41598-025-08424-1","article-title":"Comparative analysis of ChatGPT 3.5 and ChatGPT 4 obstetric and gynecological knowledge","volume":"15","author":"\u0141ugowski","year":"2025","journal-title":"Sci. Rep."},{"key":"ref30","doi-asserted-by":"publisher","first-page":"101123","DOI":"10.1016\/j.orgdyn.2024.101123","article-title":"Whom to sue? Liability of unaccountability in AI decisions","volume":"54","author":"Ma","year":"2025","journal-title":"Organ. Dyn."},{"key":"ref31","doi-asserted-by":"publisher","first-page":"221","DOI":"10.1007\/s00194-021-00548-8","article-title":"Feststellung der todesursache","volume":"32","author":"Madea","year":"2022","journal-title":"Rechtsmedizin"},{"key":"ref32","doi-asserted-by":"publisher","first-page":"30397","DOI":"10.1038\/s41598-025-15416-8","article-title":"\u201cMy AI is lying to me\u201d: user-reported LLM hallucinations in AI mobile apps reviews","volume":"15","author":"Massenon","year":"2025","journal-title":"Sci. Rep."},{"key":"ref33","doi-asserted-by":"publisher","first-page":"313","DOI":"10.1007\/s12024-020-00220-9","article-title":"A review of causal inference in forensic medicine","volume":"16","author":"Meilia","year":"2020","journal-title":"Forensic Sci. Med. Pathol."},{"key":"ref34","author":"Merken","year":"2023"},{"key":"ref35","doi-asserted-by":"publisher","first-page":"e48392","DOI":"10.2196\/48392","article-title":"The ChatGPT (generative artificial intelligence) revolution has made artificial intelligence approachable for medical professionals","volume":"25","author":"Mesko","year":"2023","journal-title":"J. Med. Internet Res."},{"key":"ref36","doi-asserted-by":"publisher","first-page":"301683","DOI":"10.1016\/j.fsidi.2023.301683","article-title":"ChatGPT, llama, can you write my report? An experiment on assisted digital forensics reports written using (local) large language models","volume":"48","author":"Michelet","year":"2024","journal-title":"Forensic Sci. Int. Digit. Investig."},{"key":"ref37","doi-asserted-by":"publisher","first-page":"122","DOI":"10.1177\/0025802418754576","article-title":"Medical evidence and proof of cause of death in nigerian courts","volume":"58","author":"Obafunwa","year":"2018","journal-title":"Med. Sci. Law"},{"key":"ref38","year":"2025"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2303.08774","article-title":"GPT-4 technical report","author":"Achiam","year":"2023","journal-title":"arXiv"},{"key":"ref40","year":"2025"},{"key":"ref41","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2203.02155","article-title":"Training language models to follow instructions with human feedback","author":"Ouyang","year":"2022","journal-title":"arXiv"},{"key":"ref42","doi-asserted-by":"publisher","first-page":"101626","DOI":"10.1016\/j.lanwpc.2025.101626","article-title":"Use of the international classification of diseases to perinatal mortality (ICD-PM) with verbal autopsy to determine the causes of stillbirths and neonatal deaths in rural Cambodia: a population-based, prospective, cohort study","volume":"60","author":"Patel","year":"2025","journal-title":"Lancet Reg. Health West. Pacific"},{"key":"ref43","doi-asserted-by":"publisher","first-page":"102122","DOI":"10.1016\/j.ijlp.2025.102122","article-title":"A case study of forensic psychiatry experts\u2019 reports analysis through large language models","volume":"102","author":"Petroni","year":"2025","journal-title":"Int. J. Law Psychiatry"},{"key":"ref44","doi-asserted-by":"publisher","first-page":"104667","DOI":"10.1016\/j.amjoto.2025.104667","article-title":"Evaluating advanced AI reasoning models: ChatGPT-4.0 and DeepSeek-R1 diagnostic performance in otolaryngology: a comparative analysis","volume":"46","author":"Prasad","year":"2025","journal-title":"Am. J. Otolaryngol."},{"key":"ref45","doi-asserted-by":"publisher","first-page":"2546","DOI":"10.1038\/s41591-025-03727-2","article-title":"Benchmark evaluation of DeepSeek large language models in clinical decision-making","volume":"31","author":"Sandmann","year":"2025","journal-title":"Nat. Med."},{"key":"ref46","doi-asserted-by":"publisher","first-page":"2050","DOI":"10.1038\/s41467-024-46411-8","article-title":"Systematic analysis of ChatGPT, google search and llama 2 for clinical decision support tasks","volume":"15","author":"Sandmann","year":"2024","journal-title":"Nat. Commun."},{"key":"ref47","doi-asserted-by":"publisher","first-page":"166","DOI":"10.1080\/13218719.2017.1379113","article-title":"The murder trial of gerard Baden-clay: admissibility of expert opinion evidence of injuries and cause of death","volume":"25","author":"Scott","year":"2017","journal-title":"J. Law Med."},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2403.05530","article-title":"Gemini 1.5: unlocking multimodal understanding across millions of tokens of context","author":"Georgiev","year":"2024","journal-title":"arXiv"},{"key":"ref49","doi-asserted-by":"publisher","first-page":"e79221","DOI":"10.7759\/cureus.79221","article-title":"DeepSeek in healthcare: revealing opportunities and steering challenges of a new open-source artificial intelligence frontier","volume":"17","author":"Temsah","year":"2025","journal-title":"Cureus"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2305.04388","article-title":"Language models don\u2019t always say what they think: unfaithful explanations in chain-of-thought prompting","author":"Turpin","year":"2023","journal-title":"arXiv"},{"key":"ref51","doi-asserted-by":"publisher","first-page":"301974","DOI":"10.1016\/j.fsidi.2025.301974","article-title":"Forensic analysis and privacy implications of LLM mobile apps: a case study of ChatGPT, copilot, and gemini","volume":"54","author":"Tyagi","year":"2025","journal-title":"Forensic Sci. Int."},{"key":"ref52","doi-asserted-by":"publisher","first-page":"owae017","DOI":"10.1093\/fsr\/owae017","article-title":"Beyond pandora\u2019s box: vast potential with significant challenges","volume":"10","author":"Wang","year":"2025","journal-title":"Forensic Sciences Research"},{"key":"ref53","doi-asserted-by":"publisher","first-page":"41","DOI":"10.1038\/s41746-024-01029-4","article-title":"Prompt engineering in consistency and reliability with the evidence-based guideline for LLMs","volume":"7","author":"Wang","year":"2024","journal-title":"npj Digit Med"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2201.11903","article-title":"Chain-of-thought prompting elicits reasoning in large language models","author":"Wei","year":"2023","journal-title":"arXiv"},{"key":"ref55","doi-asserted-by":"publisher","first-page":"301859","DOI":"10.1016\/j.fsidi.2024.301859","article-title":"Exploring the potential of large language models for improving digital forensic investigation efficiency","volume":"52","author":"Wickramasekara","year":"2025","journal-title":"Forensic Sci. Int. Digit. Investig."},{"key":"ref56","doi-asserted-by":"publisher","first-page":"2106","DOI":"10.1016\/j.procs.2025.09.331","article-title":"Mitigating automation bias in generative AI through nudges: a cognitive reflection test study","volume":"270","author":"Wingerter","year":"2025","journal-title":"Procedia Comput. Sci."},{"key":"ref57","doi-asserted-by":"publisher","first-page":"189","DOI":"10.1016\/j.aopr.2025.05.001","article-title":"DeepSeek-R1 outperforms gemini 2.0 pro, OpenAI o1, and o3-mini in bilingual complex ophthalmology reasoning","volume":"5","author":"Xu","year":"2025","journal-title":"Adv. Ophthalmol. Pract. Res."},{"key":"ref58","doi-asserted-by":"publisher","first-page":"104844","DOI":"10.1016\/j.jbi.2025.104844","article-title":"Evaluating large language models for information extraction from gastroscopy and colonoscopy reports through multi-strategy prompting","volume":"168","author":"Yu","year":"2025","journal-title":"J. Biomed. Inform."}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1752043\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,28]],"date-time":"2026-07-28T05:37:29Z","timestamp":1785217049000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1752043\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,28]]},"references-count":58,"alternative-id":["10.3389\/frai.2026.1752043"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1752043","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,28]]},"article-number":"1752043"}}