Assessing suicide risk from social media using targeted model improvements
Bag of Tricks or Bag of Myths? Reducing Modeling Complexity with Task Knowledge in Explainable Suicide Risk Assessment
Computation and LanguageArtificial Intelligence
Summary
Predicting suicide risk from social media posts is hard because there’s little data and it requires careful explanations. The authors tested many common computer language processing tricks but found most did not actually help in this tough setting. They built a new system that predicts risk levels, points to evidence in the text, and identifies important clinical factors by using special techniques tailored to the task. Their work ranked third in a competition and shows that choosing methods based on the problem itself is important.
What this means in practice
- •For mental health app developers: Incorporate more accurate and explainable suicide risk assessments from social media text into apps supporting mental health professionals.
- •For social media moderation teams: Improve tools that detect high suicide risk posts by using models tailored to handle rare risk cases and provide clear evidence highlights.
Authors
Shlok Shelat, Shrey Salvi, Souvik Roy, Manas Gaur, Amit Sheth
Abstract
Assessing suicide risk from social media text is a small-data, high-stakes setting requiring not only severity prediction but also supporting evidence and clinically relevant risk and protective factors. Yet common NLP techniques, including model scaling, synthetic data, loss reweighting, ensembling, and threshold tuning, are often applied without testing whether their gains hold up under severe class imbalance, coupled outputs, and limited author-level data. We study 1,635 clinician-annotated posts and audit 31 pre-specified techniques from 7 methodological families through roughly 300 controlled experiments on author-disjoint partitions. We found no prior audit of this playbook in this regime. The findings guide a task-grounded system for three outputs: 4-level suicide risk, evidence spans, and 24 clinical risk and protective factors. Only 5 of 31 comparisons produced reliable gains. We reformulate factor prediction as entailment between each post and its codebook definitions, using an architecturally diverse ensemble with class-balanced training and score rescaling. Risk predictions condition a 7-model evidence tagger ensemble; evidence restricts symbolic risk rules; and a difficult risk class is routed separately. The factor predictor remains independent because risk evidence provides no additional factor signal. We also correct a mismatch between validation scores used for threshold fitting and test-time ensemble scores through deployment-consistent calibration, yielding the largest improvement to the factor system. The final system achieves 0.8203 for risk, 0.7953 for evidence, and 0.7045 macro-F1 for factors, with a 0.7781 composite, ranking third among 53 teams. We call the underlying principle task-conditioned technique selection: retain techniques only when task-specific knowledge, structure, or empirical evidence justifies them.