Punished by Algorithm: The Ethics of Artificial Intelligence at Sentencing

Should artificial intelligence (AI) be used to determine the sentences of criminal offenders? If so, how should AI instruments be employed by the criminal court?

Punished by Algorithm: The Ethics of Artificial Intelligence at Sentencing constitutes the first single-authored book on the ethical advantages and disadvantages of the use of AI in sentencing. This book tackles some of the most pressing ethical and practical questions raised by the use of artificial intelligence in criminal sentencing. It explores whether sentencing is an inherently human endeavour, examines the challenges posed by the "input problem", and evaluates how the performance of sentencing algorithms ought to be assessed.

Can AI ever make better sentencing decisions than human judges? Should algorithmic systems be fully transparent, and if so, how does this compare with the opacity that can also characterize human decision-making? What constitutes algorithmic bias, and why does it matter in the context of criminal justice?

The book further investigates the role of trust in the justice system, asking whether AI can be trusted and how its deployment might reshape public confidence in legal institutions. It considers whether AI could help improve sentencing practices within imperfect real-world systems and what criteria should govern decisions about introducing AI into judicial decision-making. Finally, it examines why many AI ethics frameworks struggle to connect penal theory with penal practice and considers the case for establishing a national AI Sentencing Commission to oversee the responsible use of these technologies.