Punishment and Artificial Intelligence

Suppose that computer scientists and engineers have developed a brand-new algorithm which is designed to determine sentences in individual criminal cases. Suppose, furthermore, that it seems that this algorithm is actually very good at doing this job. Would it then be justified to replace human sentencing judges with the sentencing algorithm? The answer to this question obviously depends upon many things. However, in this chapter focus is placed narrowly on the question of how we should compare human judges and the algorithm with regard to how well each is doing the job of determining the severity of the appropriate sentences. It is argued that, due to both theoretical and practical reasons, we do to a large extent currently lack the penal ethical resources to answer this question. The impetus behind these considerations is the assumption that the question may well become urgent in a not very distant future and that, in the absence of the requisite ethical considerations, there is a significant risk that decisions on this matter will be made on insufficient or even ethically arbitrary grounds.