Machine Learning Model for Predicting Suicide Risk on Social Media
Author Information
Author(s): Mavragani Amaryllis, Shang Lili, Tillmann Taavi, Washburn Jason, Kaminsky Zachary, McQuaid Robyn J, Hellemans Kim GC, Patterson Zachary R, Saad Mysa, Gabrys Robert L, Kendzerska Tetyana, Abizaid Alfonso, Robillard Rebecca
Primary Institution: University of Ottawa Institute of Mental Health Research at The Royal
Hypothesis
Can the Suicide Artificial Intelligence Prediction Heuristic (SAIPH) accurately predict suicidal ideation using social media data?
Conclusion
The study validates the SAIPH model's ability to predict suicidal ideation and assess changes in suicidal trajectory over time.
Supporting Evidence
- SAIPH scores were significantly associated with suicidal ideation metrics in both cohorts.
- Individuals with responses to suicidal mentions showed improved SAIPH scores over time.
- SAIPH effectively predicted suicidal ideation using social media data collected before suicidal mentions.
Takeaway
Researchers created a computer program that can help predict if someone might be thinking about suicide based on what they post on social media.
Methodology
The study analyzed Twitter data from two cohorts, scoring them with the SAIPH algorithm and comparing results to standardized suicidal ideation metrics.
Potential Biases
Potential bias due to reliance on self-reported data and the nature of social media interactions.
Limitations
The study relied on single-question metrics for suicidal ideation and was conducted during the COVID-19 pandemic, which may have influenced results.
Participant Demographics
Participants included 159 university students and 307 individuals from a national COVID-19 survey, with varying ages and a majority being female.
Statistical Information
P-Value
p=.01
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website