Artificial Intelligence-Based Detection of Suicidal Ideation among University Students: A Mixed Methods Approach

Author(s): Ali J. Eidan1, Hassam M. Al-Amarei2, Zainab Ammar Mohammed3, Zahraa Hamid Mohsen3, Kawther Abdul Hussein Abdulhadi3, Zahraa Hameed Oda Alquraishi4
Date: 08/10/2026
doi: https://doi.org/10.5281/zenodo.8425890

Abstract

Objective: This study aimed to assess the prevalence of suicidal ideation among university 
students using the Beck Scale for Suicide Ideation (BSSI), evaluate the concordance of Beck 
suicide risk with three artificial intelligence platforms (DeepSeek, Gemini, and ChatGPT), and 
identify thematic drivers of suicidal ideation through qualitative analysis of free-text responses. 
Methods: A descriptive cross-sectional study was conducted with 495 university students 
using a self-reported questionnaire that included sociodemographic variables, the Beck Scale 
for Suicide Ideation, and an open-ended question about current psychological state and 
stressors. 
Results: Among participants, 52.93% were classified as low risk, 37.78% as moderate risk, 
and 5.26% as high and very high risk for suicidal ideation. Significant associations were 
observed between suicidal ideation and female gender and economic status, with students 
reporting insufficient income demonstrating the highest risk levels. Moreover, analysis of AI 
concordance revealed significant differences in BSSI scores across DeepSeek-derived risk 
categories (F=39.251, p<0.0001), with fair agreement with the BSSI (Quadratic Weighted 
Kappa (QWK)=0.248). In contrast, Gemini (F=1.952, p=0.101; QWK=0.231) and ChatGPT 
(F=1.994, p=0.114; QWK=0.119) demonstrated no statistically significant differences between 
risk groups, with Gemini showing fair and ChatGPT showing slight agreement, respectively. 
Thematic analysis of free-text responses identified academic pressure, particularly examination 
stress, workload, and fear of failure, as the primary contributing factors. 
Conclusion: Suicidal ideation represents a critical mental health concern among university 
students. The concordance of AI platforms with BSSI remains varied, ranging from fair to 
slight, establishing them as decision-support platforms rather than alternatives for clinical 
assessment.

Download Full Text