New perspectives on A.I. in sentencing. Human decision-making between risk assessment tools and protection of human rights

The aim of this thesis is to investigate a field that until a few years ago was foreign to and distant from the penal system. The purpose of this undertaking is to account for the role that technology could plays in the Italian Criminal Law system.
More specifically, this thesis attempts to scrutinize a very intricate phase of adjudication. After deciding on the type of an individual's liability, a judge must decide on the severity of the penalty. This type of decision implies a prognostic assessment that looks to the future. It is precisely in this field and in prognostic assessments that, as has already been anticipated in the United, instruments and processes are inserted in the pre-trial but also in the decision-making phase. In this contribution, we attempt to describe the current state of this field, trying, as a matter of method, to select the most relevant or most used tools. Using comparative and qualitative methods, the uses of some of these instruments in the supranational legal system are analyzed. We do so to better enable policy makers and academics to understand the nuance that might arise from the introduction of such instruments, trying to abandon an approach of total closure and caution, but trying to glimpse and focus the analysis on certain instruments that might prove useful at a limited stage of the decision.
This approach makes it possible to take a closer look at the impacts on criminal and procedural rights and guarantees and the benefits they could provide.
Focusing attention on the Italian system, an attempt was made to investigate the nature of the element of an individual's ‘social dangerousness’ (pericolosità sociale) and capacity to commit offences, types of assessments that are fundamental in our system because they are part of various types of decisions, including the choice of the best sanctioning treatment. It was decided to turn our attention to this latter field because it is believed that the judge does not always have the time, the means and the ability to assess all the elements of a subject and identify the best 'individualizing' treatment in order to fully realize the function of Article 27, paragraph 3 of the Constitution.
Clearly, it has been acknowledged that the introduction of such instruments must necessarily be confronted with a system of substantive and procedural guarantees that must remodel or innovate in the presence of such instruments. Indeed, in a procedural dialectic that has hitherto seen only the defendant, the judge and other marginal subjects as protagonists, the I.A. instruments are part of this dialogue.
An attempt has been made in this paper to show an optimistic outlook towards an introduction of such tools, albeit only limited to a certain point in time and initially applied only in bonam partem. The proposal considers ad hoc regulation of artificial intelligence in the first place and of such instruments thereafter as necessary. Reflections can at present only stop at an embryonic stage, since it must first be understood through a more in-depth study whether it is really acceptable to link the quantification of the sentence to the assessment of the risk of reoffending. We ask to what extent and in what ways some A.I. tools may prove not only useful but also crucial in a decision as support to the adjudicating body.
The significance of this study is that it informs our theoretical understanding of the relation between A.I. tools and criminal sectors by introducing a focus on risk assessment tools in the assessment of social dangerousness in the choice of the best correctional treatment. These findings indicate the need for resources in order to investigate this field.