Risk-scoring error types depend on which label counts as positive

Evidence

Even though “the detailed differences between [these] multiple definitions are difficult to grasp” (Verma & Rubin, 2018, p. 1) and full explication is beyond the scope of this chapter, it is illustrative to go over a few examples. Let’s consider the case of determining a defendant’s risk score. For simplicity, let’s assume that the score is either a “good” risk score (where good means that the defendant is unlikely to commit additional crimes and likely to show up to future court dates) or a “bad” risk score (likely to commit additional crimes and unlikely to show up to court). A true positive is when a defendant is given a good risk score, and this score accurately reflects the reality (the defendant is unlikely to commit additional crimes and likely to show up to future court dates). A false positive means that a defendant is given a good risk score, but this is not accurate (i.e., the defendant should have been given a bad risk score). A true negative is when a defendant is given a bad risk score, and this accurately reflects the reality. A false negative is when a defendant is given a bad risk score, but this does not reflect the reality (i.e., the defendant should have been given a good risk score). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1156

Categorization

• DEFINE: Carefully defines true/false positives and negatives applied to the specific case of criminal risk scores.
• corroborated: 2 independent passages support this claim — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 1156