Explore the vast range of titles available.

Governments around the world are increasingly turning to algorithms to support or fulfil administrative tasks – from automating social benefit regulation to the automatic determination of the merit of court cases. With this global trend came along an increasing number of scandals that often resulted from a mismatch between implementation of an algorithmic process by the government and public expectation. These (violated) expectations concern several factors of the deployment, which are currently understudied. This article analyses how six of these factors – institution, aim, legal field, data provenance, trigger, and level of automation – influence social acceptance of automating administrative and legal processes in France. From our factorial vignette study with 6 variables across 540 vignettes and N=7,560 assessments, we draw three main conclusions. First, we find that the institution, aim, data provenance, trigger and level of automation are statistically significant in driving the acceptance of automatically processable regulations. Second, we find that acceptance across all conditions is only slightly positive on average, when expressed on a scale from −5 to +5. Overall, our study confirms that the social acceptability of the implementation of algorithms in administrative and legal processes is highly dependent on specific implementation and contextual details; public discourse on the implementation of such algorithms hence requires information about specifics of their deployment.

Artificial Intelligence (AI) is increasingly used worldwide to make decisions, be it by public administrations, industry, banks, or insurers. One area with a particularly high impact on citizens is its use in judiciary processes. To this day, there has been little investigation into why countries decide to adopt AI in this area. In this article, we show that the mere promise of efficiency gains is not enough of an explanation for why countries adopt AI within judiciary processes. Instead, we show that the administrative burden (as measured by the number of days to trial), a government’s leaning within the political spectrum (namely towards the left), and the level of adoption of technology by governments in neighboring countries predict countries’ announcement to adopt AI in their judiciary. Taken together, our findings derived from a data analysis of 28 advanced democracies constitute an important step towards a better understanding of the factors shaping AI’s integration into judiciary processes.

The field of computational law has increasingly moved into the focus of the scientific community, with recent research analysing its issues and risks. In this article, we seek to draw a structured and comprehensive list of societal issues that the deployment of automatically processable regulation could entail. We do this by systematically exploring attributes of the law that are being challenged through its encoding and by taking stock of what issues current projects in this field raise. This article adds to the current literature not only by providing a needed framework to structure arising issues of computational law but also by bridging the gap between theoretical literature and practical implementation. Key findings of this article are: (1) The primary benefit (efficiency vs. accessibility) sought after when encoding law matters with respect to the issues such an endeavor triggers; (2) Specific characteristics of a project—project type, degree of mediation by computers, and potential for divergence of interests—each impact the overall number of societal issues arising from the implementation of automatically processable regulation.

An important challenge when creating automatically processable laws concerns open-textured terms. The ability to measure open-texture can assist in determining the feasibility of encoding regulation and where additional legal information is required to properly assess a legal issue or dispute. In this article, we propose a novel conceptualisation of open-texture with the aim of determining the extent of open-textured terms in legal documents. We conceptualise open-texture as a lever whose state is impacted by three types of forces: internal forces (the words within the text themselves), external forces (the resources brought to challenge the definition of words), and lateral forces (the merit of such challenges). We tested part of this conceptualisation with 26 participants by investigating agreement in paired annotators. Five key findings emerged. First, agreement on which words are open-texture within a legal text is possible and statistically significant. Second, agreement is even high at an average inter-rater reliability of 0.7 (Cohen’s kappa). Third, when there is agreement on the words, agreement on the Open-Texture Value is high. Fourth, there is a dependence between the Open-Texture Value and reasons invoked behind open-texture. Fifth, involving only four annotators can yield similar results compared to involving twenty more when it comes to only flagging clauses containing open-texture. We conclude the article by discussing limitations of our experiment and which remaining questions in real life cases are still outstanding.

This paper revisits the claim that the state capacity of attribution works as a deterrent for criminals to launch cyber attacks. Motivated by other empirical evidence for other types of crimes that do not support the claim, this research designed two quantitative analyses to test it. The first experiment looked at macro-level variables at the unit of the state and found that attribution can act as a deterrent. However, a second experiment looking at individual cases distinguished between three types of population and identified only one population for which the attribution-deterrence nexus is valid. Grounded in control theory, the claim is valid for individuals with a sufficient knowledge about the attribution process, who act rationally, and who are concerned about the socio-economic cost of the punishment. Enhancing attribution mechanisms is unlikely to result in any change of behaviours for criminals who act without knowledge or only with a limited perception of the attribution mechanisms, or for individuals who do not fear punishments as society praises their technological skills despite their anti-social and unethical behaviours. [PUBLICATION ABSTRACT]

Open-texture—e.g. vague, ambiguous, under-specified, or abstract terms—in regulatory documents lead to inconsistent interpretation, and are an obstacle to the automatic processing of regulation by computers. Identifying which parts of a legal text fall under open-texture is therefore a necessary requirement to make progress in automating the law. In this paper, we propose that large language models (LLMs) might provide an effective way to automatically detect open-texture in legal texts. We first investigate the obstacles by situating open-texture in the broader literature, and we test the hypothesis using two different LLMs—the proprietary gpt-3.5-turbo and the open-source llama-2-70b-chat—for the task of identifying open-texture in the General Data Protection Regulation. We evaluate their performance by asking 12 annotators to assess their output. We find, overall, that gpt-3.5-turbo overperforms llama-2-70b-chat on F 1 -scores (0.84 vs 0.67), and its high F 1 -score could make it a suitable alternative, or complement, to using human annotators. We also test the sensitivity of the findings against four further LLMs combined with six different prompts, and replicate a finding that there is low agreement between annotators when it comes to the identification of open-texture. We conclude the article by discussing the subjectivity of open-texture, the lessons to draw when testing for open-texture, and the consequences of using LLMs in the legal domain.

Driven by the increasing availability and deployment of ubiquitous computing technologies across our private and professional lives, implementations of automatically processable regulation (APR) have evolved over the past decade from academic projects to real-world implementations by states and companies. There are now pressing issues that such encoded regulation brings about for citizens and society, and strategies to mitigate these issues are required. However, comprehensive yet practically operationalizable frameworks to navigate the complex interactions and evaluate the risks of projects that implement APR are not available today. In this paper, and based on related work as well as our own experiences, we propose a framework to support the conceptualization, implementation, and application of responsible APR. Our contribution is twofold: we provide a holistic characterization of what responsible APR means; and we provide support to operationalize this in concrete projects, in the form of leading questions, examples, and mitigation strategies. We thereby provide a scientifically backed yet practically applicable way to guide researchers, sponsors, implementers, and regulators toward better outcomes of APR for users and society.

Attribution - finding the identity of the actors behind an attack - is of primary importance in order to be able to classify an attack as a criminal act, an act of war, or an act of terrorism. For cyber attacks, three assumptions prevail in the literature: attribution is a technical problem; it is unsolvable; and it is unique. The thesis seeks to examine these assumptions more closely by asking the following two research questions: What constrains attribution? And what does the attribution process entail? It argues that the three prevailing assumptions are misleading. Approaching attribution as a problem forces us to consider it either as solved or unsolved. Yet attribution is far more nuanced than that would suggest: it is better approached as a process in constant evolution, driven by judicial and political pressures. The thesis methodology of detailed examination and comparison of case studies, in addition to interviews with experts, provides evidence to support the arguments. The attribution process arises in two different contexts, with two distinct sets of constraints and goals. In the criminal context, courts must assess the guilt of criminals, mainly based on technical evidence. In the national security context, decision-makers must analyse unreliable and mainly non-technical information in order to identify an enemy of the state. Attribution in both contexts is political: in criminal cases, laws reflect prevailing norms and power in society; in national security cases, attribution is the reflection of a state's will to maintain, increase or assert its power. However, both processes differ on many levels, which are examined in turn throughout the thesis. The constraints, which reflect common aspects of many other political issues, constitute the structure of the thesis: the need for judgement calls, the role of private companies, the standards of evidence, the role of time, and the plausible deniability of attacks.

This article posits that attribution is better approached as a process than as a problem. Departing from models for attribution emphasising its technical constraints, the article distinguishes between two distinct attribution processes operating in two different contexts and answering two different questions. On one hand, in a criminal context, the attribution process seeks to identify individuals who launched cyber attacks. On the other hand, in a national security context, the process seeks to identify adversaries which may have sponsored cyber attacks. Five elements can serve as determinants to distinguish between the two processes: the type of target, the severity and scale of the damage, the apparent origin of the instigator of the attack, the means of the attack, and lastly the claims by political groups. After reviewing these elements, the article analyses how the constraints and characteristics of the two different attribution processes vary in terms of responsible authority, standard of evidence, stake in play and relevance of timing for attribution. In a criminal context, the collection of digital evidence is of primary importance to be able to reach a high level of judicial proof. However, in a national security context, investigators will likely not have access to evidence but only to intelligence, and the decision of attribution will not be taken within the judiciary but within the executive (e.g. a government). As such, attribution of national security incidents requires a form of political judgement to heed the political context and circumstances of the incidents that courts do not need to consider when examining the guilt of criminals. Political judgement is also required to balance the consequences of attribution on a political level, potentially resulting in a straining of relations with other international actors. As the constraints between the two different attribution processes differ, their requirements for easing the constraints also vary. [PUBLICATION ABSTRACT]

Among patients with a high risk of reintubation, spontaneous-breathing trials performed with pressure-support ventilation did not result in significantly more ventilator-free days at day 28 than T-piece trials.