The ability to accurately predict violence and other forms of serious antisocial behavior would provide important societal benefits, and there is substantial enthusiasm for the potential predictive accuracy of neuroimaging techniques. Here, we review the current status of violence prediction using actuarial and clinical methods, and assess the current state of neuroprediction. We then outline several questions that need to be addressed by future studies of neuroprediction if neuroimaging and other neuroscientific markers are to be successfully translated into public policy. Violent behavior is a costly large-scale societal problem. There is growing interest in using neuroscience data to assess risk for future violent behavior, but the utility of neuroscience for violence risk assessment remains to be established. We review what is currently known about the underlying neurobiological mechanisms of violence, and evaluate recent neuroprediction efforts. Finally, we outline a set of practices for enhancing the validity and reliability of future risk assessment based on neuroscientific measures.
- machine learning
- predictive modeling