2021/1 – Algoritmi ed esperienza giuridica

The proliferation of algorithms throughout society has resulted in significant challenges to the law in several important areas. For example, for constitutional law the use of algorithms may lead to challenges to free speech rights and to the fundamental rights defendants are afforded in court proceedings. In intellectual property law, the use of algorithms may lead to creative works of authorship or to inventions which may be eligible for protection under copyright or patent law respectively, yet no human author or inventor may have been involved in producing the work. Additionally, governments regularly use algorithms in the process of governing itself, which raises challenging issues within administrative law, contract law, and others. Industry is also using algorithms to interface with customers, supervise business transactions, and to make decisions which impact commercial transactions. And there is interest among legal scholars to discuss the role of algorithms as actors in discriminatory behavior directed against historically protected individuals and groups. This indicates that constitutional rights as well as basic human rights may be involved when algorithms are used in decision making. With these issues in mind, this paper presents a broad overview of the law which is being challenged by the use of algorithms with the goal to generate further discussion on the topic of how the use of algorithms affect people’s basic rights under the law.

The article aims to provide an overview about some implications of emerging AI applications in justice. In particular some of the issues raised by the use of Predictive Justice in criminal law are examined. For example the enhancement of the bias contained in the algorithms due to the programming and the opacity of big data decisions due to machine learning. The Author argues that Predictive Justice – antithetical to Interpretative Justice – is likely to overwhelm fundamental requirements of law, such as the transparency in producing law and the connection between law and fact. On the basis of EU guidelines, it is proposed to adopt a trusthworthy approach with respect to AI, with the aim of reconciling machine learning with the principle of non-discrimination and strict protection of fundamental rights.

Information Society has evolved to Algorithmic Society, where decision-making processes are done not only by humans but also by artificial agents. However, opacity usually characterizes algorithms, and it means that machine reasoning could be understood only looking at each output, while keeping their manufacturers substantially unaccountable. Starting from the perspective of legal informatics, this paper aims at investigating the dichotomy between opacity and confidentiality, on the one hand, and explainability and transparency, on the other hand, taking into account the current debate. Eventually, challenges and possible solutions for a fair Algorithmic society are discussed.

The spread and pervasiveness of algorithms have led to the definition of today’s society as an “algorithmic society”. Resuming Latour’s description of the “Berlin key” and its power to program human action, the paper focuses on the algorithm as a non-neutral and normative tool. Social regulation by means of algorithms aims to obtain total effectiveness by rendering everything measurable.

Big data and artificial intelligence have opened up unprecedented perspectives for our ability to predict future states of the world. The same is true in the legal field, where scholars and practitioners are growingly drawn to the forecasting capabilities of algorithms. In recent years, prediction models have not only fed a lively theoretical debate on legal computability and predictive justice. Still, they have also inspired a number of applications spanning from intelligent platforms for workforce management to innovative tools for the judicial assessment of recidivism risk. In such a scenario, we need a reflection on the impact that computational heuristics can have on the very complexion of law. Seen up close, the use of predictive analytics techniques in legal settings is often affected by issues that range from inherent epistemic fragilities to the risk of turning into rights violations. This paper provides a critical account of computational prediction and its hidden pitfalls. Our first goal is to lay the groundwork for an in-depth analysis of the theoretical and practical implications that predictive heuristics may have for law. The second one is to present augmented intelligence – the cooperative integration between humans and machines – as a reference paradigm to mitigate the risks of prediction and, more in general, to inspire the computational evolution of legal science and practice.

The text develops some remarks starting from the juxtaposition between legal science and artificial intelligence, in particular by focusing on the logics on which these two paradigms are based. The remarks stem from some basic aspects of the law, on the one hand, and, on the other hand, of contemporary artificial intelligence. The perspective adopted considers in particular the origins of contemporary artificial intelligence, and problematizes some of the main purposes of Turing’s project of constructing a universal machine. From the point of view of the law, the text adopts the perspective of the so-called tradition of «legal experience», which allows to develop some interesting remarks about the present use of algorithms in legal science.

For some time now, the use of intelligent machines has radically changed the nature of the juridical experience. Algorithms have become common in courtrooms, in administration archives and in parliamentary assemblies, showing the effects of their formidable power. This paper aims to show how, rather than resuscitating the Leibnizian dream of a mathematical certainty, artificial intelligence can allow a judge to weigh up the possible dynamics of the system and thus to take full responsibility for decisions which aim to guarantee the defence of increasingly full and inclusive rights.

The algorithm is a sublimation of statistics. If statistics is still a reflection of reality, the algorithm becomes reality in itself, or self-reference, or implicit knowledge: a pure abstraction that evolves independently of any human intervention. It is no longer man who knows by algorithms, but the algorithms know by themselves. Artificial intelligence is the reflection of statistics; it is the extreme and definitive passage from the real to the virtual. We have reached the decisive turning from the humanization of the divine, the sovereign as a mortal God, to the deification of computation; a technological, social and cultural revolution called infinite computing. We don’t know how many autonomous software agents reshaping our lives, our interactions, and our environments. Faced with a systematic and overall change in human condition, what kind of society can we expect?

The rationale behind the deployment of algorithms in judicial activities and, in particular, in the investigative field is to answer to the enquiries of a judge, prosecutor or any other legal professional. Legal professionals and computer scientists must understand the manifold challenges of the problem under scrutiny to identify the best solution, on the one hand, and, on the other hand, to put into question the above findings. This article maps out the properties of the algorithms deployed in the field of digital forensics with the aim of detecting and analysing any potential digital evidence. Theo verarching needs are to monitor, understand and justify algorithms’ outcomes. Thus, in view of factors that may render the algorithmic logic procedure less clear, the technical activity shall preserve digital evidence’s usability in trials, in order to safeguard the guarantees of due process according to national and supranational constitutional rights.

The aim of the paper is to explore the broader scope of the Ethics Guidelines for Trustworthy AI. In particular, the paper focuses on the reasons that led EU to develop an ethical approach to AI, seeking to investigate to which extent it is arguable that the ethical principles for a trustworthy AI should be based on the compliance with fundamental rights. It points out that the symbolic value of fundamental rights as embedded within this non-binding tool, shows the normative vision of EU, mitigating the possible conflict between institutional and private actors involved as well as different interests at stake. Finally, it argues that neither the ethical approach nor the legal design of AI can effectively address the issue of algorithmic inferences and their impact on individuals and society.

The text provides for a picture of the main research lines addressed by Paolo Grossi in his academic work and underlines its contribution and significance to the understanding of law, even nowadays.

Il volume di Nicola Lettieri affronta il problema della tecnoregolazione, o regolazione algoritmica, da una prospettiva filosofico-giuridica, soffermandosi in particolare sugli effetti della rivoluzione digitale sulla fenomenologia delle norme (pp. 25-34), sul “lato oscuro” delle meccaniche algoritmiche e sulle possibili insidie in termini di libertà e diritti (pp. 35-49), e infine sulle (ridotte) possibilità di critica ammissibili nella sfera digitale (pp. 50-60).
l volume ha il merito di portare l’attenzione su una serie di problemi ormai ineludibili per la cultura giuridica, i quali rendono, per certi aspetti, datata e superata la tradizionale distinzione tra “apocalittici” e “integrati”. Come ha scritto di recente Tommaso Edoardo Frosini, si tratta di una questione che verrà presto risolta dall’avvicendarsi delle generazioni, dal momento che coloro che hanno oggi vent’anni “non conoscono altra civiltà che quella digitale”.
L’aspetto di maggiore originalità del lavoro di Lettieri consiste nel suo approdo, ossia nella formulazione di proposte concrete capaci di ibridare l’indagine filosofico-giuridica con la scienza informatica, al fine di individuare correttivi e soluzioni equitative concretamente capaci di riequilibrare le pressoché inevitabili imprecisioni, gli errori e le distorsioni delle procedure algoritmiche.