Constraining Learning, Expanding Competencies: Lawyer-in-the-Loop in Algorithmic Discovery

• In Event: Algorithms at Work: The Practice of Prediction - I: Algorithms-in-Practice in Institutional and Organizational Perspective

Abstract

Often framed as a recent phenomenon, applying algorithmic intelligence to legal decision-making is well established in US civil litigation practice. From 2012-2015, civil discovery evolved from an overwhelmingly manual process of attorney document review to a sophisticated methodology driven by machine learning (ML) governed by case law, professional standards of conduct, and rules of legal procedures. My current research examines how accountability mechanisms evolved in this early adopter application, paying specific attention to the technical competencies lawyers were newly required to develop in order to adequately perform their discovery duties. This work is informed by an in-depth qualitative analysis of relevant case law, law review articles, professional ethic codes, and semi-structured interviews with litigation practitioners. Despite its remarkable early success, I argue that algorithmic civil discovery still exhibits characteristics in practice that are odds with legal goals, norms, and education. Specifically, my analysis explores the notion that ML-driven discovery has unwittingly shifted the framing of legal discovery practice away from data-driven iterative sensemaking, to that of metrics-driven automated classification. This conceptual and operational shift risks constraining the learning afforded to lawyers during the discovery process, while in parallel requiring them to perform technical and managerial tasks their formal training does not adequately prepare them for. Critical study of these issues is of pressing importance to civil litigation practice. It also presents itself as a fertile empirical case study for the broader study of workplace algorithmic transformation especially as it relates to ML applications supporting or supplanting traditional white-collar work.

Author

• Fernando Delgado, Cornell University