Abstract
Despite its increasing influence on our everyday life, many aspects of Artificial Intelligence (AI) remain a mystery to the human mind. Therefore, it is understandable that in the past, proposals to introduce AI into an already indeterminate and uncertain domain such as sentencing law are often met with scepticism and apprehension. Such a proposal, at first instance, would seem to some, ambitious or even naive. However, considering that sentencing is a complex legal domain involving the balancing of various difficult considerations which has traditionally remained within the realm of human judges, such an assertion is reasonable. In this dissertation, however, I will make a modest case for the introduction of such technology into this domain and how best to regulate it. This dissertation contends that a new regulatory model in the form of an Independent Monitor would be adequate to address the harms associated with an algorithmic sentencing system.
My argument is modest in two respects. Firstly, this dissertation considers an incremental establishment of an algorithmic sentencing system that would operate alongside New Zealand's existing sentencing process. The predictions generated by the system in the form of a recommended sentence will serve nothing more than a non-binding recommendation. Secondly, the algorithmic sentencing system discussed in this dissertation does not aim to replicate the moral reasoning involved in the judicial decision-making process but rather to systemise past decisions made by human judges to inform the judge of how their colleagues in the past facing similar situations have reacted.