Human-Inspired AI for Suicide Prediction and Prevention

Speaker: Yaniv Yacoby, Assistant Professor of Computer Science, Wellesley College
Series: Centre for Human-Inspired AI (CHIA)

Suicide is a difficult topic, and understanding how to prevent it remains a major challenge. AI could help researchers better understand changes in suicide risk and inform more effective prevention and treatment.

Transitions from suicidal thoughts to action can happen rapidly, leaving narrow windows for observation and intervention. Smartphone studies offer valuable insights by collecting frequent self-reports of emotions and suicidal thoughts in daily life. However, these datasets are often small and noisy, and connecting them with the psychological theories that guide clinical practice remains difficult.

In this talk, Yaniv Yacoby will share ongoing research developing models of psychological state that combine machine learning, psychological theories, clinician and patient knowledge, and smartphone self-report data. He will explore how verbal theories can be translated into mathematical models that can be tested against data, with the aim of making predictions more interpretable and reliable and supporting more effective, theory-informed suicide prevention.

About the speaker

Yaniv Yacoby is an Assistant Professor of Computer Science at Wellesley College, where he leads the Model-Guided Uncertainty (MOGU) Lab. His research combines machine learning and psychological science to better understand, predict and prevent suicide. He recently received a $3 million research award from the US National Institute of Mental Health to support this work.

Previously, Yaniv was a postdoctoral fellow in the Nock Lab at Harvard University and Massachusetts General Hospital. He completed his PhD at Harvard, advised by Finale Doshi-Velez, and holds a master’s degree in music from the New England Conservatory.