A Commentary on Statistical Assessment of Violence Recidivism Risk

Abstract

Increasing integration and availability of data on large groups of persons has been accompanied by proliferation of statistical and other algorithmic prediction tools in banking, insurance, marketing, medicine, and other fields (see, e.g., Steyerberg 2009a, b). Controversy may ensue when such tools are introduced to fields traditionally reliant on individual clinical evaluations. Such controversy has arisen about "actuarial" assessments of violence recidivism risk, that is, the probability that someone found to have committed a violent act will commit another during a specified period. Recently, Hart, Michie, and Cooke (2007a) and subsequent papers from these authors in several reputable journals have claimed to demonstrate that statistical assessments of such risks are inherently too imprecise to be useful, using arguments that would seem to apply to statistical risk prediction quite broadly. This commentary examines these arguments from a technical statistical perspective, and finds them seriously mistaken in many particulars. They should play no role in reasoned discussions of violence recidivism risk assessment.

Methodological limitations in the measurement and statistical modeling of violence among adults with mental illness

“…Challenges to methodological or conceptual issues in violence risk assessment research are not new (Buchanan, 2008;Buchanan, 2014;Hart, Michie, & Cooke, 2007;Imrey & Dawid, 2015;Large & Singh, 2014;Mossman, 2014;Wand & Large, 2013). Our hope is that the current empirical examination of three limitations found frequently in the extant research will spur serious consideration about ways to improve the science and practice of clinical violence risk screening, assessment, and management.…”

Diagnosing Crime and Diagnosing Disease: Bias Reduction Strategies in the Forensic and Clinical Sciences

“…Instructive examples can be found by examining the history and current research on the prediction of violent or dangerous behavior. The prognostic models used in violence prediction, including sexual violence, are generalizable across a broad range of future risks (73), including the prediction of cardiovascular events (74), property losses and other insurance risks (75), and breast cancer survival (76).…”

On individual risk

“…This has been initiated and promoted by Hart et al (2007), Cooke and Michie (2010), Hart and Cooke (2013), whose analysis has been strongly influential in some quarters (see e.g. Singh and Petrila 2013), but has been widely criticised for serious technical statistical errors and confusions (Harris et al 2008;Hanson and Howard 2010;Imrey and Dawid 2015;Mossman 2015). But to date this debate has (to its own disadvantage) largely avoided questioning the meaning of an individual risk.…”

“…from a pragmatic standpoint, such information about frequencies in a group of people "like Sam" (or "like Angelina") can be extremely helpful and informative. In Imrey and Dawid (2015) such a group frequency, taken as relevant to an individual member of the group, is termed an "individualized risk." The foundational philosophical question, however, is: Can we consider this as supplying a measure of "individual risk"?…”