Can a Machine Sentence Justly?

Jesper Ryberg and Julian Roberts teach ethics and criminology at Roskilde University and the University of Oxford respectively. In Sentencing and Artificial Intelligence, they have curated a powerful and compelling collection of essays on the application of a new technology to an old problem. Their edited volume focuses on the application of predictive computational tools, often called “artificial intelligence” (AI) or “machine learning,” to the task of determining the punishment that a convicted offender will receive. As their professional orientations might lead one to expect, their book focuses on normative rather than empirical or technical questions. The resulting essays, written by scholars from the United States, Canada, the United Kingdom, and the European Union (but not the Global South), do not advance a single thesis or follow a singular argumentative thread. They instead engage with in interrelated suite of normative questions. These include due process (or the procedural obligations owed by the state to individuals); non-discrimination entitlements; and the defendant’s claims to an “accurate” judgment (so far as facts go) or the opportunity to call for “mercy” in the exercise of sentencing-related discretion. All of these issues, as I shall discuss below, are addressed at what we might call a “mid-level” of generality—i.e. by taking as the object of study institutions and practices at a certain level of abstraction. This decision results in discussion generic enough to be located in any one of a number Western European or North American jurisdictions, whether guided by civil or common law. Indeed, it is quite striking that the reader does not encounter much by way of a case study or a specific description of AI-driven sentencing for the first sixty pages of the book. When specific tools are discussed, moreover, details are also scant. For Aziz Z. Huq is Professor at the University of Chicago Law School, Chicago, IL, USA. Email: huq@uchicago.edu Criminal Justice Ethics, 2022 Vol. 41, No. 3, 268–277, https://doi.org/10.1080/0731129X.2022.2136867