Related Experiment Video
Updated: Jul 10, 2026
07:31
Implementation of a Real-Time Psychosis Risk Detection and Alerting System Based on Electronic Health Records using CogStack
Published on: May 15, 2020
Investigating the Relationship Between Suicide Risk and Warning Signs Within China and the United States: Using
Zizhuo Grace Yin1, Martin Swanbrow Becker2, Yin Yang3
1Counseling and Psychological Service, Purdue University, West Lafayettle, Indiana, USA.
Summary
Machine learning analyzed social media posts to find cultural differences in suicide risk expressions. Chinese posts often mentioned sleep and fatigue, unlike English posts, highlighting the need for nuanced interpretation.
Area of Science:
- Computational social science
- Digital mental health
- Cross-cultural psychology
Background:
- Social media platforms are crucial for understanding public health issues.
- Analyzing online suicide-related content requires sophisticated methods to account for cultural nuances.
- Previous research often lacks large-scale, cross-cultural comparisons of online suicidality expressions.
Purpose of the Study:
- To investigate cross-cultural differences in suicide risk expressions on social media using machine learning.
- To apply computational methods to analyze large datasets of social media posts from different cultural contexts.
- To identify specific linguistic and thematic patterns associated with suicide risk in Chinese and English online communities.
Main Methods:
- Collected 3545 English posts (Reddit) and 2326 Chinese posts (Zhihu) from 2010-2019.
- Employed a supervised machine learning algorithm for content screening and analysis of suicide-related expressions.
- Utilized logistic regression and binomial distribution to analyze the relationship between post content, nationality, and suicide risk indicators, including Patient Health Questionnaire (PHQ-9) items.
Main Results:
- Significant cross-cultural differences were found in suicide risk expressions.
- Chinese social media posts were more likely to express issues related to sleep, fatigue, eating, and concentration.
- Two Patient Health Questionnaire (PHQ-9) items (low interest and feeling bad) showed a negative association with suicide risk, potentially indicating general distress rather than acute risk.
Conclusions:
- Cultural context significantly influences how individuals express suicidality online.
- Machine learning is a valuable tool for identifying complex, culturally specific patterns in online suicide-related content.
- Findings have implications for culturally sensitive suicide prevention and intervention strategies informed by digital data.