Race-blind procedures fail because correlated proxy variables still leak bias

Evidence

Another commonly cited solution is to blind criminal justice actors to defendants’ race (see Sah, Robertson, & Baughman, 2015). This practice is currently being implemented by the San Francisco District Attorney’s Office (Gecker, 2019). But there are limitations to such an approach: even if a defendant’s skin color is hidden, cues and correlated variables, such as defendants’ zip codes, names, or colleges attended, may still be present (Sah et al., 2016). In addition, prosecutors must see defendants at some point, so blinding can only impact initial case decision-making. — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1170 (#corroborates, via linked claim)

To organize across the discourses of blackness and masculinity, of race and gender, and forget the way in which, at the same moment, blacks in the underclass were being positioned in class terms, in similar work situations, exposed to the same deprivations of poor jobs and lack of promotion that certain members of the white working class suffered, was to leave out the critical dimension of positioning. — Stuart Hall, Essential Essays, Volume 2, 251 (#complicates, via linked claim)

Categorization

  • XREF: connects to: Blind coding cannot fix bias already introduced during data collection.; Race-blind prediction models can encode bias; active debiasing improves fairness. — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 1185; Omitted variables correlated with race bias the discrimination estimate — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 736 (+9 more)

Broader Connections