Context presentation and native conversation structure in large language model responses to escalating suicide risk: a controlled multimodel evaluation
4 affiliations
- 1Department of Plastic and Aesthetic Surgery, The Second Affiliated Hospital of Guilin Medical University
- 2School of Clinical Medicine, Chengdu University of Traditional Chinese Medicine
- 3School of Clinical Medicine, Xiangnan University
- 4School of Economics, Zhejiang University of Science & Technology
Abstract
Background: Large language models are increasingly used for informal mental health support, yet their safety during escalating suicide-risk conversations may depend on how prior information is presented. Native multi-turn conversations and reconstructed cumulative prompts differ not only in the presence of prior model responses but also in prompt reconstruction, information concentration, message positioning, explicit integration cues, and conversation-state processing.
Objective: This study compared suicide-risk recognition, direct safety inquiry, response urgency, and timely real-world intervention across isolated current-turn prompts, reconstructed cumulative user-only prompts, and native continuous multi-turn conversations.
Methods: We conducted a controlled repeated-measures evaluation of ChatGPT, Gemini, and Doubao using 12 standardized five-turn trajectories: 4 direct suicide-risk trajectories, 4 implicit or internally inconsistent risk trajectories, and 4 non-suicidal high-distress controls. Each trajectory was evaluated under three context-presentation formats. The 24 model-trajectory combinations were treated as matched repeated-measures units nested within 8 independent risk-trajectory blocks. Primary contrasts were estimated using exact trajectory-level sign-flip tests and trajectory-block bootstrap confidence intervals. Turn-level outcomes were analyzed using multilevel mixed-effects models. Exploratory condition-by-model heterogeneity was assessed using a trajectory-blocked exact permutation sensitivity analysis, with a trajectory-clustered interaction model used only as a secondary asymptotic check.
Results: Timely appropriate intervention occurred in 10/24 isolated current-turn observations (41.7%), 21/24 reconstructed cumulative observations (87.5%), and 12/24 native continuous observations (50.0%). On average across the three systems, reconstructed cumulative presentation exceeded isolated input by 45.8 percentage points (95% trajectory-block bootstrap CI, 25.0 to 66.7; Holm-adjusted exact P = .031), whereas native continuous performance was 37.5 percentage points lower than reconstructed cumulative performance (95% trajectory-block bootstrap CI, −50.0 to −25.0; Holm-adjusted exact P = .031). Exploratory model-specific analyses indicated substantial heterogeneity: the native-versus-reconstructed contrast was largest for Doubao, smaller and statistically imprecise for Gemini, and absent for ChatGPT.
Conclusions: On average across the three evaluated systems, reconstructed cumulative user-only prompts outperformed native continuous conversations, although exploratory model-specific patterns were heterogeneous. Because the formats also differed in prompt reconstruction, information salience, message positioning, explicit integration instructions, and native conversation-state processing, the contrast cannot isolate the causal effect of prior model responses.
Related
- German Adaptation and Psychometric Evaluation of the Conversation Questionnaire
- Navigating Dynamic Risks: Outcome Monitoring and Evaluation under Escalating Risk
- Large Language Models' Simulation of Patient Comprehension of Discharge Summaries: Multimodel Evaluation Study.
- Double-blind, placebo-controlled randomised clinical trial on flexible doses of escitalopram and naltrexone in the treatment of men with compulsive sexual behaviour disorder: study protocol presentation
- De-escalating Therapy in Obstructive Sleep Apnoea After Glucagon-Like Peptide-1 Receptor Agonist-Induced Weight Loss: A Narrative Review
BibTeX
@article{liu2026context,
title = {{Context presentation and native conversation structure in large language model responses to escalating suicide risk: a controlled multimodel evaluation}},
author = {Jueqiao Liu and Xiaoyi Chen and Jun Cai and Xuchen Luo and Wei Xiao and Jiehua Deng},
journal = {Frontiers in Psychiatry},
year = {2026},
month = oct,
doi = {10.3389/fpsyt.2026.1962098},
url = {https://www.frontiersin.org/articles/10.3389/fpsyt.2026.1962098/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Context presentation and native conversation structure in large language model responses to escalating suicide risk: a controlled multimodel evaluation
AU - Jueqiao Liu
AU - Xiaoyi Chen
AU - Jun Cai
AU - Xuchen Luo
AU - Wei Xiao
AU - Jiehua Deng
PY - 2026
DA - 2026/10/16
JO - Frontiers in Psychiatry
DO - 10.3389/fpsyt.2026.1962098
UR - https://www.frontiersin.org/articles/10.3389/fpsyt.2026.1962098/full
AB - Background: Large language models are increasingly used for informal mental health support, yet their safety during escalating suicide-risk conversations may depend on how prior information is presented. Native multi-turn conversations and reconstructed cumulative prompts differ not only in the presence of prior model responses but also in prompt reconstruction, information concentration, message positioning, explicit integration cues, and conversation-state processing. Objective: This study compared suicide-risk recognition, direct safety inquiry, response urgency, and timely real-world intervention across isolated current-turn prompts, reconstructed cumulative user-only prompts, and native continuous multi-turn conversations. Methods: We conducted a controlled repeated-measures evaluation of ChatGPT, Gemini, and Doubao using 12 standardized five-turn trajectories: 4 direct suicide-risk trajectories, 4 implicit or internally inconsistent risk trajectories, and 4 non-suicidal high-distress controls. Each trajectory was evaluated under three context-presentation formats. The 24 model-trajectory combinations were treated as matched repeated-measures units nested within 8 independent risk-trajectory blocks. Primary contrasts were estimated using exact trajectory-level sign-flip tests and trajectory-block bootstrap confidence intervals. Turn-level outcomes were analyzed using multilevel mixed-effects models. Exploratory condition-by-model heterogeneity was assessed using a trajectory-blocked exact permutation sensitivity analysis, with a trajectory-clustered interaction model used only as a secondary asymptotic check. Results: Timely appropriate intervention occurred in 10/24 isolated current-turn observations (41.7%), 21/24 reconstructed cumulative observations (87.5%), and 12/24 native continuous observations (50.0%). On average across the three systems, reconstructed cumulative presentation exceeded isolated input by 45.8 percentage points (95% trajectory-block bootstrap CI, 25.0 to 66.7; Holm-adjusted exact P = .031), whereas native continuous performance was 37.5 percentage points lower than reconstructed cumulative performance (95% trajectory-block bootstrap CI, −50.0 to −25.0; Holm-adjusted exact P = .031). Exploratory model-specific analyses indicated substantial heterogeneity: the native-versus-reconstructed contrast was largest for Doubao, smaller and statistically imprecise for Gemini, and absent for ChatGPT. Conclusions: On average across the three evaluated systems, reconstructed cumulative user-only prompts outperformed native continuous conversations, although exploratory model-specific patterns were heterogeneous. Because the formats also differed in prompt reconstruction, information salience, message positioning, explicit integration instructions, and native conversation-state processing, the contrast cannot isolate the causal effect of prior model responses.
SN - 1664-0640
ER -