Risk Assessment, Predictive Algorithms and Preventive Justice

Abstract

The term “preventive justice” was first used in the late eighteenth century and linked to laws aimed at preventing future crime by intervening where, according to Blackstone (Commentaries on the Laws of England in Four Books. Routledge, 1753), there was a “probable suspicion, that some crime is intended or likely to happen”. The past few decades have seen preventive justice schemes reinvigorated in association with a focus on risk assessment tools aimed at predicting the risk of future harmful behavior. While there has been considerable criticism of the use of risk assessment tools to predict rather than manage behavior, “structured professional judgment”, which combines statistical or actuarial risk prediction with clinical methods, has become an accepted forensic method to help identify those who are at low, moderate or high risk of harming others. Recently, predictive machine learning algorithms have been used to inform judicial decision-making, including sentencing, in the absence of expert testimony about their proper use. These algorithmic assessments may be viewed as an extension of a previous trend toward actuarial prediction tools aimed at assessing the risk of recidivism. This chapter analyzes some of the issues raised by the use of risk assessment tools in predicting the risk of harm. It will argue that there remain serious ethical and human rights concerns with the movement toward predictive algorithms in predicting the risk of future harmful behavior and that structured professional judgment, with all its faults, may in fact be the “least worst” option.