Umeå University's logo

The chapter discusses the creation and maintenance of databases offering accurate, research-ready data for multidisciplinary use. It draws on the experience with the IUROPA CJEU Database Project (IUROPA), which has collected data about the decision-makers and the decisions of the Court of Justice of the European Union (CJEU). IUROPA and similar multi-user databases must live up to four criteria for databases, as proposed by Weinshall and Epstein. First, they must address real-world problems. Second, they must be open and accessible. Third, they must deliver reliable and reproducible data. Fourth, they must be ageless and easily calibrated to research purposes unknown at the time of data collection and cleaning. These criteria involve trade-offs: the quest for reliability may, first, precipitate difficult choices such as whether to discard or improve upon ‘imperfect’ data or tempt creators to endlessly postpone publication of ‘incomplete’ data; second, sustainability and human intervention are inversely proportionate when it comes to database maintenance; finally, a fledgling discipline like empirical legal studies in EU law imposes a disproportionate time commitment and financial responsibility on a small group of researchers.

Political scientists have discovered the European Court of Justice, but has it discovered law? We address this enduring question, first posed by Armstrong in 1998, by tracing the evolution of law in political science work on the CJEU, from a concept understood in rudimentary terms as an external constraint on judicial behaviour to more recent nuanced accounts of legal concepts, doctrine, and judicial practices. While political science has come closer to the nuts and bolts of CJEU decision-making, we argue that there is untapped potential in exploring the multidimensionality of legal cases and the micro-level details of legal interpretation and adjudication.

Identifying case law and other legal resources that substantiate legal propositions is a fundamental aspect of legal research and decision-making. Existing legal information retrieval systems assist this task by recommending relevant legal documents. However, these documents are often lengthly, and users are primarily interested in accessing and referencing specific, directly relevant sections. Augmenting recommendation at the document level with suggestions at the paragraph level, here referred to as ‘pincite recommendations’, could significantly enhance efficiency. This paper presents, tests, and proposes an approach for such a pincite recommendation model. Using the case law of the Court of Justice of the European Union as a test case, we demonstrate that a language embeddings model can predict citations with a high degree of accuracy, providing users precise and pertinent pincites for legal propositions.

Why do judges disagree about judicial reasoning – i.e., the use of legal reasoning in adjudication – and how can such disagreement be studied empirically? We propose a general theory of variation in judicial reasoning that focuses on a core element of legal analysis – the choice of authoritative sources. In particular, we argue that source-based disagreement is likely to manifest itself on two dimensions: The degree to which extra-national versus national and extra-legislative versus legislative sources should be relied on. Where judges place themselves on these dimensions have important normative implications for the origin of and power over law, and in particular the status and power of the national legislator. We demonstrate the strength and usefulness of our theory with original data from more than 3400 judicial opinions of the Swedish Supreme Court over a 40-year period. We find that Justices’ backgrounds, in terms of when they went to law school and professional pre-appointment experience, are correlated with their placement on these two dimensions in theoretically expected ways.

Party-appointed arbitrators (PAAs) are a common, but controversial feature of many arbitral systems. The main point of contention is whether parties can and do use their power to appoint one of the arbitrators in a way that undermines impartial, equal, and fair arbitration. While the scholarly debate on PAAs is old and vast, claims made have been subject to limited empirical testing. Using the Court of Arbitration for Sport (CAS) as a case study, this article empirically explores the e!ects of arbitrator appointments on arbitration outcomes and the relevance of arbitrator experience, arbitrator attitudes, and party capability on party success. In doing so, this study finds that arbitrator experience and attitudes a!ect outcomes and that this can be exploited by parties when selecting arbitrators. Importantly, the study also finds that the ability to make use of this strategic opportunity depends on parties’ capability. While PAAs may be less problematic in arbitration between parties of equal capability, this finding calls into serious question the fairness and legitimacy of using PAAs in arbitration where there is a disparity in capability between the parties.

This report outlines the foundational components necessary to develop EurLexGPT - a modular, domain-specific AI-based system built on EU legislative data - to serve as a core infrastructure for the European RegTech sector, enabling flexible, AI-supported access, analysis, and integration of legal information to support compliance, oversight, and policymaking across diverse regulatory contexts. We apply a multidisciplinary methodology that combines legal analysis, data evaluation,and technical design to define the requirements for building EurLexGPT. We begin by analyzing the structure of EU law to determine which types of legal content the system must accurately represent. We then assess existing EU legal data sources for their suitability as training material, and explore the technical architecture needed to support a modular, verifiable AI system. This integrated approach ensures that EurLexGPT is both legally grounded and technically robust, positioning it as a foundational tool for the European RegTech sector. Several strategic choices must be made. Selectinga suitable open-source base LLM is a crucial first step. This model will be at the heart of the system’snatural language generation ability. Since development in the field is fast paced, we recommend using a recently released open-source model with demonstrated high general performance. Secondly a choice must be made about fine-tuning the selected model to better perform on the targeted functionof EurLexGPT. Recent research shows that even light finetuning can enhance performance of LLM’sin annotation tasks. We therefore describe fine-tuning in some detail, analyze the benefits and drawbacks of using fine-tuning. We also provide some final recommendations about how such tuning can be implemented to a selected base model. Thirdly, we discuss retrieval augmented generation(RAG). RAG is today an industry standard in most use cases for language generation tasks and in thefield of law serve the important purpose of highlighting legal sources used for generation. We recommend building EurLexGPT as a RAG system, with options to select which data sources shouldbe included. Then, fourthly, we shortly point out the option of setting up EurLexGPT to deliver Chainof-Thought generation (using the RAG database as sources of intermediary reasoning), before finallypointing to the importance of producing a user guide to EurLexGPT highlighting both the genericissue of prompt sensitivity, that applies to all LLM generation and the specific risks, in a legal context,of over-relying on LLMgenerated text.

I turbulenta tider, då grundläggande rättigheter och begränsningar för maktutövning utmanas, ökar betydelsen av de gränser som den konstitutionella rätten ställer upp för utövandet av offentlig makt. Samtidigt blir det konstitutionellrättsliga läget allt mer komplext, inte minst på grund av europarättens inflytande.

Konstitutionell rätt behandlar den konstitutionella rättens tre centrala teman: maktdelning och andra förutsättningar för utövande av offentlig makt, skyddet för grundläggande rättigheter samt rättsligt genomdrivande av konstitutionellrättsliga normer. Utifrån dessa tre teman behandlar författarna konstitutionella bestämmelser med ursprung i svensk nationell rätt, EU-rätten och Europakonventionen om mänskliga rättigheter.

Boken är anpassad för att kunna användas som lärobok på juristutbildningen, men riktar sig också till en bredare publik. Konstitutionell rätt är relevant läsning för såväl erfarna jurister som läsare som är intresserade av konstitutionella frågor mer allmänt. Denna tredje upplaga är påtagligt omarbetad och utvidgad.

We argue that negative references to amicus curiae briefs in high court judgments - instances where a court explicitly signals disagreement with the legal arguments in such briefs - are a significant and understudied feature of judicial reasoning. We theorize that such references may provide courts with a tool for increasing the precision of its case law, fostering its legitimacy, and increasing compliance pressure. Our empirical analysis of the Court of Justice of the European Union indicates that negative references are used both to boost its legitimacy and to specify not only what the law is, but what it is not.