The Pile of Law paper provides empirical evidence for (a), and the authors indeed suggest that such systems could be integrated into data workflows to accomplish (b).[6]
Based on § 4.2., it seems that the AI would assist humans to catch up with already accepted antidiscrimination decisionmaking, such as that mitigating racial biases. Human internalization of these norms could lead to further bias reduction in law, to which AI would help catch up some individuals who could otherwise decide based on earlier legislation.
§ 3.2 suggests that a variety of microaggressions would be detected, which could increase the focus on subjective experiences of the legal subjects. Anonymization in immigration and civil litigation could initially support but eventually contradict this effort. If decisionmakers could be biased by a name or a characteristic, then anonymization could improve the legal compliance of their judgments. However, if some individuals need to be protected by anonymization, then their subjective experience can still be worse than that when they do not. Thus, if legal compliance can be improved without de facto anonymization, that can be better. Alternatively, it should be continuously phased out for specific groups.
Based on § 4.2., it seems that the AI would assist humans to catch up with already accepted antidiscrimination decisionmaking, such as that mitigating racial biases. Human internalization of these norms could lead to further bias reduction in law, to which AI would help catch up some individuals who could otherwise decide based on earlier legislation.
§ 3.2 suggests that a variety of microaggressions would be detected, which could increase the focus on subjective experiences of the legal subjects. Anonymization in immigration and civil litigation could initially support but eventually contradict this effort. If decisionmakers could be biased by a name or a characteristic, then anonymization could improve the legal compliance of their judgments. However, if some individuals need to be protected by anonymization, then their subjective experience can still be worse than that when they do not. Thus, if legal compliance can be improved without de facto anonymization, that can be better. Alternatively, it should be continuously phased out for specific groups.