Toward a Fourth Law of Robotics: Preserving Attribution, Responsibility, and Explainability in an Algorithmic Society

摘要

Jack Balkin makes several contributions in this Lecture.He proposes a set of "laws of robotics" for an "Algorithmic Society" (i.e., one characterized by "social and economic decision-making by algorithms, robots, and AI agents"). 1These laws both elegantly encapsulate and add new principles to a growing movement for accountable design and deployment of algorithms. 2 My purpose in commenting on his Lecture is threefold: (1) to contextualize his proposal as a kind of regulation of regulation, familiar from the perspective of administrative law, (2) to expand the range of methodological perspectives capable of identifying "algorithmic nuisance," a key concept in this Lecture, 3 and (3) to propose a fourth law of robotics to ensure the viability of Balkin's first three laws.Balkin argues that "algorithms (a) construct identity and reputation through (b) classification and risk assessment, creating the opportunity for (c) discrimination, normalization, and manipulation, without (d) adequate transparency, accountability, monitoring, or due process." 4 In response to these