Juror Perceptions of Machine Learning Based Diagnoses of Mental Disorders

Abstract

Advances in technology have led to ever more sophisticated methods for identifying and diagnosing mental disorders. This study aims to understand how jurors will react to machine learning based diagnoses of mental disorders in the sentencing process. Participants read a summary of a criminal court case that included either psychological or neuroscientific expert evidence that was or was not based on machine learning outputs. We found expert evidence generally produced contradictory effects (decreasing culpability but increasing perceived dangerousness), but the effects were consistent across psychological, neurological, and machine learning-based evidence.