Rethinking Magnitude: Toward Empirically Calibrated Benchmarks in Forensic Risk Assessment

Effect size reporting is ubiquitous in the risk assessment literature and is routinely used to demonstrate the predictive value of structured tools, compare instruments, and support claims about their adequacy and utility. However, far less attention has been given to the interpretive conventions used to assign meaning to these estimates. Many commonly invoked benchmarks, including those derived from Cohen’s work on effect size interpretation in the early 1990s, were developed as broad behavioural science heuristics rather than context-specific standards for high-stakes applied prediction. Despite this, such benchmarks are frequently applied in forensic risk assessment with limited scrutiny of their origins, assumptions, or contextual fit. In this article, we argue that effect size interpretation in risk assessment requires a more contextually grounded approach. Focusing on the area under the curve as a widely reported index of predictive accuracy, we examine the limitations of decontextualised magnitude labels and synthesise Area Under the Curve (AUC) estimates for existing tools across the published risk assessment literature. Drawing on these estimates, we propose provisional, field-specific benchmarks for interpreting AUCs across violence, sexual offending, and general recidivism domains. These benchmarks are not intended to function as rigid thresholds, but as descriptive reference points for situating findings within the empirical distribution of published risk assessment effects. In doing so, the article aims to support more transparent, defensible, and contextually meaningful interpretation of predictive performance, and to promote further discussion about how effect sizes should be evaluated in forensic and other high-stakes assessment settings.