Governing What Law Does to People: Trauma-Informed Practice and Algorithmic Accountability

1. Technology competence: ABA Model Rule 1.1 Comment 8 requires attorneys to keep abreast of changes in the law and its practice, including the benefits and risks associated with relevant technology. For James, competence requires: (1) understanding what COMPAS is and what it claims to assess; (2) knowing the Dressel and Farid accuracy findings (65.2%) and the racial disparity in false-positive rates; (3) being able to challenge the reliability of the score in the context of Daquan’s matter; (4) requesting methodology and demographic performance data from the court.
2. Systemic concern: Dressel and Farid (2018) established COMPAS accuracy at 65.2% — no better than untrained lay prediction and not materially better than chance. Black defendants are incorrectly flagged as high-risk at approximately twice the rate of white defendants. Berk et al. (2024) established that this bias originates in the training data and cannot be fixed through technical calibration. For a defendant who is Black, the prior probability of a false-positive high-risk classification is systematically elevated by the algorithm’s training data, independent of his actual risk level.
3. Rule 3.3 (candor): If James discovers the court is relying on a tool with documented accuracy limitations and racial disparities, and the court appears unaware, Rule 3.3’s candor obligation may require bringing this to the court’s attention. If the prosecution presents the COMPAS score as reliable evidence of risk, James may have an obligation to disclose the Dressel and Farid findings as directly contrary authority. Jurisdictional verification required.
4. Sample motion: ‘DEFENDANT’S MOTION TO EXCLUDE OR LIMIT ALGORITHMIC RISK ASSESSMENT SCORE. Defendant Daquan [surname] moves to exclude or limit reliance on the COMPAS score in the court’s bail determination on the grounds that: (1) COMPAS has documented accuracy limitations below the standard of reliability required for liberty decisions (Dressel & Farid, Science Advances 2018); (2) COMPAS produces systematically elevated false-positive risk scores for Black defendants due to racially disparate training data (Berk et al., 2024); and (3) no demographic impact assessment of COMPAS has been conducted for this county’s population. Defendant requests disclosure of COMPAS methodology, feature weights, and demographic performance data for this jurisdiction.’

This model analysis represents the depth of reasoning expected in the practice application exercise. The post-test (minimum 80%) assesses conceptual mastery of the session’s evidence base and governance framework.