Giovanni Sartor

This chapter proposes a dialectical approach to legal interpretation, consisting of three dimensions: (1) a formalization of the canons of interpretation in terms of argumentation schemes; (2) a dialectical classification of interpretive schemes; and (3) a logical and computational model for comparing the arguments pro and contra an interpretation. The traditional interpretive maxims or canons used in both common and civil law are translated into defeasible patterns of arguments, which can be evaluated through sets of corresponding critical questions. These interpretive argumentation schemes are classified in general categories and a distinction is drawn between schemes supporting and rebutting an interpretation. This framework allows conceiving statutory interpretation as a dialectical procedure consisting in weighing arguments pro and contra an interpretation. This procedure is formalized and represented computationally through tools from formal argumentation systems.

This chapter provides an analysis of defeasible legal reasoning as argumentation. It first provides a general account of the idea of defeasibility and introduces the idea of nonmonotonic reasoning. It then focuses on defeasible argumentation, considering how defeasible arguments can be constructed and how they can be defeated by rebutting and undercutting counterarguments. The dialectical interactions of defeasible arguments are further explored by focusing on reinstatement and reasoning about priorities. The idea of legal systems as the basis for argumentation frameworks is then investigated. The rationale for defeasibility in law is discussed, along with the possibility of using different approaches, such as revision or probability, to deal with uncertainty in legal reasoning. Finally, an account is provided of the emergence of theories of defeasibility in philosophy, logic, and legal theory.

This chapter is meant to address the extent to which value-based reasoning—as involved in balancing and proportionality—may include quantitative reasoning, according to arithmetic constraints. Relying on some work on cognitive and evolutionary psychology, it is argued that processing non-symbolic approximate magnitudes is a fundamental cognitive capacity, which is deployed also when we are reasoning with goals and values. A model is developed for determining the impact of alternative choice on multiple values, assessing the utilities so produced and merging these utilities into an overall evaluation, which may be used in comparisons. The model is applied to assessing legislative choice on rights and social values, according to proportionality. Finally, it is shown how proportionality assessments may be constrained by the requirement of consistency with precedents.

Argumentation schemes are forms of reasoning that are fallible but correctable within a self-correcting framework. Their use provides a basis for taking rational action or for reasonably accepting a conclusion as a tentative hypothesis, but they are not deductively valid. We argue that teleological reasoning can provide the basis for justifying the use of argument schemes both in monological and dialogical reasoning. We consider how such a teleological justification, besides being inspired by the aim of directing a bounded cognizer to true belief and correct choices, needs to take into account the attitudes of dialogue partners as well as normative models of dialogue and communicative activity types, in particular social and cultural settings

Inspired by legal reasoning, this paper presents a formal framework for assessing conflicting arguments. Its use is illustrated with applications to realistic legal examples, and the potential for implementation is discussed. The framework has the form of a logical system for defeasible argumentation. Its language, which is of a logic-programming-like nature, has both weak and explicit negation, and conflicts between arguments are decided with the help of priorities on the rules. An important feature of the system is that these priorities are not fixed, but are themselves defeasibly derived as conclusions within the system. Thus debates on the choice between conflicting arguments can also be modelled.The proof theory of the system is stated in dialectical style, where a proof takes the form of a dialogue between a proponent and an opponent of an argument. An argument is shown to be justified if the proponent can make the opponent run out of moves in whatever way the opponent attacks. Despite this dialectical form, the system reflects a declarative, or relational approach to modelling legal argument. A basic assumption of this paper is that this approach complements two other lines of research in AI and Law, investigations of precedent-based reasoning and the development of procedural, or dialectical models of legal argument.

This article proposes a formal analysis of a fundamental aspect of legal reasoning: dealing with normative conflicts. Firstly, examples are illustrated concerning the dynamics of legal systems, the application of rules and exceptions, and the semantic indeterminacy of legal sources. Then two approaches to cope with conflicting information are presented: the preferred theories of Brewka, and the belief change functions of Alchourrón, Gärdenfors, and Makinson. The relations between those approaches are closely examined, and some aspects of a model of reasoning with normative conflicts are outlined. Since this model takes into account an ordering of the involved regulations, criteria to order legal norms are finally specified.

Abstract. In this paper I shall take an inferential approach to legality (legal validity), and consider how the legality of a norm can be inferred, and what can be inferred from it. In particular, I shall analyse legality policies, namely, conditionals conferring the quality of legality upon norms having certain properties, and I shall examine to what extent such conditionals need to be positivistic, so that legality is only dependant on social facts. Finally, I shall consider how legality is transmitted from norm to norm and whether the ultimate legality policies (the rules of recognition) of a legal system need to be constituted by social facts.

Beginning with a concise reflection on the significance of digital sovereignty, this chapter describes and problematizes the Chinese social credit system in connection with the notion of digital sovereignty. It provides a synthetic illustration of the origin, evolution, essential characteristics, structure, explicit aims, and main limitations of the Chinese social credit system, taking into account the specific historical and ideological context in which the system has been designed and developed. On the basis of this description, the potentialities, still to be realised, of the use of artificial intelligence techniques in the different applications related to surveillance, evaluation, and sanctioning are assessed and discussed, in relation to the Chinese discourse on digital sovereignty. The last part of the chapter discusses the potential consequences of such a system in terms of individual rights, individual self-determination, and autonomy, which in turn influence the construction of the entire social fabric.

This chapter introduces a conceptual framework for qualitative risk assessment of AI, particularly in the context of the EU AI Act. The framework addresses the complexities of legal compliance and fundamental rights protection by integrating definitional balancing and defeasible reasoning. Definitional balancing employs proportionality analysis to resolve conflicts between competing rights, while defeasible reasoning accommodates the dynamic nature of legal decision-making. Our approach stresses the need for an analysis of AI deployment scenarios and for identifying potential legal violations and multi-layered impacts on fundamental rights. On the basis of this analysis, we provide philosophical foundations for a logical account of AI risk analysis. In particular, we consider the basic building blocks for conceptually grasping the interaction between AI deployment scenarios and fundamental rights, incorporating in defeasible reasoning definitional balancing and arguments about the contextual promotion or demotion of rights. This layered approach allows for more operative models of assessment of both high-risk AI systems and General Purpose AI (GPAI) systems, emphasizing the broader applicability of the latter. Future work aims to develop a formal model and effective algorithms to enhance AI risk assessment, bridging theoretical insights with practical applications to support responsible AI governance.

This article advances an explanatory model of the academic and policy debate on AI as legal persons. It argues that the scientific and regulatory debate on AI as legal persons undergoes periods of relative stability interrupted by rapid paradigm shifts. Three interrelated factors primarily influence these oscillations: (1) competing theories of legal personhood (clustered versus singularist), (2) capability, embodiment, and commercial reach of AI technology, and (3) AI's integration within socio-digital institutions. Two additional forces modulate the depth and durability of any shift: (4) the overlap and internal consistency of cross‑domain legal regimes (data protection, agency, liability, cybersecurity); and (5) the body of judicial precedents (or their scarcity) on extending personhood. Specific convergences of these factors create pressure for status reconsideration; absent them, equilibrium returns. Current EU law and case practice are explained by partial convergence—capability/diffusion surges and some cluster‑theory uptake without sufficiently embedding AI in rights‑bearing institutional roles—and by the damping effect of dense, interlocking legal regimes, plus scarce and negative judicial precedents. Case law suggests that, without explicit legislation, courts are unlikely to grant AI legal personhood on their own. Prospective movement would require demonstrable de facto social personhood: sustained, role‑based participation within socio-digital institutions that creates governance gaps that existing doctrines cannot internalize. The article concludes by assessing near- and long‑term prospects — from generative AI, autonomous agents, social robots, and digital twins to brain–machine interface — highlighting conditional pathways rather than deterministic trajectories.

In this paper we provide a formal analysis of the idea of normative co-ordination. We argue that this idea is based on the assumption that agents can achieve flexible co-ordination by conferring normative positions to other agents. These positions include duties, permissions, and powers. In particular, we explain the idea of declarative power, which consists in the capacity of the power-holder of creating normative positions, involving other agents, simply by proclaiming such positions. In addition, we account also for the concepts of representation, namely the representatives capacity of acting in the name of his principal, and of mandate, which is the mandatees duty to act as the mandator has requested. Finally, we show how the framework can be applied to represent the contract-net protocol. Some brief remarks on future research and applications conclude this contribution.

Machine learning classifiers are increasingly used to inform, or even make, decisions significantly affecting human lives. Fairness concerns have spawned a number of contributions aimed at both identifying and addressing unfairness in algorithmic decision-making. This paper critically discusses the adoption of group-parity criteria (e.g., demographic parity, equality of opportunity, treatment equality) as fairness standards. To this end, we evaluate the use of machine learning methods relative to different steps of the decision-making process: assigning a predictive score, linking a classification to the score, and adopting decisions based on the classification. Throughout our inquiry we use the COMPAS system, complemented by a radical simplification of it (our SAPMOC I and SAPMOC II models), as our running examples. Through these examples, we show how a system that is equally accurate for different groups may fail to comply with group-parity standards, owing to different base rates in the population. We discuss the general properties of the statistics determining the satisfaction of group-parity criteria and levels of accuracy. Using the distinction between scoring, classifying, and deciding, we argue that equalisation of classifications/decisions between groups can be achieved thorough group-dependent thresholding. We discuss contexts in which this approach may be meaningful and useful in pursuing policy objectives. We claim that the implementation of group-parity standards should be left to competent human decision-makers, under appropriate scrutiny, since it involves discretionary value-based political choices. Accordingly, predictive systems should be designed in such a way that relevant policy goals can be transparently implemented. Our paper presents three main contributions: (1) it addresses a complex predictive system through the lens of simplified toy models; (2) it argues for selective policy interventions on the different steps of automated decision-making; (3) it points to the limited significance of statistical notions of fairness to achieve social goals.

Reasoning with cases has been a primary focus of those working in AI and law who have attempted to model legal reasoning. In this paper we put forward a formal model of reasoning with cases which captures many of the insights from that previous work. We begin by stating our view of reasoning with cases as a process of constructing, evaluating and applying a theory. Central to our model is a view of the relationship between cases, rules based on cases, and the social values which justify those rules. Having given our view of these relationships, we present our formal model of them, and explain how theories can be constructed, compared and evaluated. We then show how previous work can be described in terms of our model, and discuss extensions to the basic model to accommodate particular features of previous work. We conclude by identifying some directions for future work.

In a world of human-only readers, a trade-off persists between comprehensiveness and comprehensibility: only privacy policies too long to be humanly readable can precisely describe the intended data processing. We argue that this trade-off no longer exists where LLMs are able to extract tailored information from clearly-drafted fully-comprehensive privacy policies. To substantiate this claim, we provide a methodology for drafting comprehensive non-ambiguous privacy policies and for querying them using LLMs prompts. Our methodology is tested with an experiment aimed at determining to what extent GPT-4 and Llama2 are able to answer questions regarding the content of privacy policies designed in the format we propose. We further support this claim by analyzing real privacy policies in the chosen market sectors through two experiments (one with legal experts, and another by using LLMs). Based on the success of our experiments, we submit that data protection law should change: it must require controllers to provide clearly drafted, fully comprehensive privacy policies from which data subjects and other actors can extract the needed information, with the help of LLMs.

Criminal liability for acts committed by AI systems has recently become a hot legal topic. This paper includes three different contributions. The first contribution is an analysis of the extent to which an AI system can satisfy the requirements for criminal liability: accomplishing an actus reus, having the corresponding mens rea, possessing the cognitive capacities needed for responsibility. The second contribution is a discussion of criminal activity accomplished by an AI entity, with reference to a recent case involving an online bot, the Random Darknet Shopper. This discussion will provide the context for the analysis of commonalities and differences between criminal activities by humans and by artificial systems. The third contribution concerns the evaluation of different ways of addressing criminal activities by AI systems in a regulatory perspective.

Accidents involving autonomous vehicles raise difficult ethical dilemmas and legal issues. It has been argued that self-driving cars should be programmed to kill, that is, they should be equipped with pre-programmed approaches to the choice of what lives to sacrifice when losses are inevitable. Here we shall explore a different approach, namely, giving the user/passenger the task of deciding what ethical approach should be taken by AVs in unavoidable accident scenarios. We thus assume that AVs are equipped with what we call an “Ethical Knob”, a device enabling passengers to ethically customise their AVs, namely, to choose between different settings corresponding to different moral approaches or principles. Accordingly, AVs would be entrusted with implementing users’ ethical choices, while manufacturers/programmers would be tasked with enabling the user’s choice and ensuring implementation by the AV.

The fields of linguistic pragmatics and legal interpretation are deeply interrelated. The purpose of this paper is to show how pragmatics and the developments in argumentation theory can contribute to the debate on legal interpretation. The relation between the pragmatic maxims and the presumptions underlying the legal canons are brought to light, unveiling the principles that underlie the types of argument usually used to justify a construction. The Gricean maxims and the arguments of legal interpretation are regarded as presumptions subject to default used to justify an interpretation. This approach can allow one to trace the different legal interpretive arguments back to their basic underlying presumptions, so that they can be compared, ordered, and assessed according to their defeasibility conditions. This approach allows one to understand the difference between various types of interpretive canons, and their strength in justifying an interpretation.