Learning resource · Teaching note

Violence Risk Tools: What AUC Does and Does Not Tell You

Prepared by MindFrame Editorial

2026

Purpose

An evidence-synthesis object unpacking what the Area Under the Curve (AUC) statistic, commonly reported for forensic violence-risk tools, actually measures, and the calibration, sensitivity, specificity and base-rate information it leaves out.

Psychology Lens

What the work portrays

This object is a standing evidence-synthesis object, not a dramatized title, focused on the statistical literacy needed to interpret forensic risk-assessment research responsibly.

Clinical & psychological context

A 2023 systematic review and meta-analysis of 50 publications and over 10,000 participants found that most studies of violence risk instruments reported AUC (a measure of discrimination between higher- and lower-risk groups) but omitted calibration and other metrics needed to judge real-world predictive accuracy for individuals.

Representation note

A high AUC shows a tool can rank people by relative risk better than chance, but it does not by itself show that the tool's specific probability estimates are accurate, nor does it account for the base rate of violence in the population being assessed; MindFrame cautions against citing AUC alone as proof a tool is clinically dependable.

Themes

Concepts

Study Mode

Watch / read → Observe → Reflect → Discuss

What to observe before interpreting

  • The distinction between a tool that discriminates well (AUC) and a tool that is well calibrated (its stated probabilities match observed outcomes)
  • How a low base rate of violence in a population can make even a 'good' tool generate many false positives
  • What information beyond AUC would be needed to judge whether a specific risk score is trustworthy for an individual decision

Discussion questions

  • Why can a violence-risk tool have a statistically impressive AUC and still perform poorly in practical, individual-level decision-making?
  • What did the cited systematic review find about how consistently calibration and other key metrics are reported alongside AUC?

Sources

A listed source is a traceable reference; it should not be assumed to support every statement on this page unless the claim has been individually reviewed.

Review status

Identity: Pending legacy review · Psychology annotation: Pending legacy review · Source support: Pending legacy review · Rights: Metadata only

Legacy pilot record retained for continuity; publication status does not imply independent verification of every identity, source or psychological claim.