Algorithmic impact assessments decouple bias monitoring from stated goals

Evidence

A recent report found that, while pretrial risk assessments may lead to biased outcomes if used carelessly, targeted and teleological use leads to both significantly reduced pretrial detention and less racial disparities—in short, a win by all accounts (Picard, Watkins, Rempel, & Kerodal, 2019). “Teleological” means that the use must be guided by clear goals, such as decreased disparities, increased efficiency, or fewer incarcerations. “Targeted” means that leadership must consider whether the tool should be used only in a subset of cases, such as those that are most closely related to the stated goals (e.g., if the goal is to decrease incarceration, then there should be a focus on violent crimes committed by high-risk offenders). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1191 (#complicates, via linked claim)

Categorization

• XREF: connects to: Goal-guided targeted use of risk assessments reduces both detention and bias — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 1191; Optimizing for stated goals produces monstrous outcomes — , The Braided Reading EPUB: 5-Year…, 655

Broader Connections

• Goal-guided targeted use of risk assessments reduces both detention and bias #complicates (0.72) — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1191