The project proposal explicitly and somewhat outrageously promised to ‘achieve a pivotal innovation of legal method’. Though legal scholars are divided on the nature of legal method, with some focusing on formal validity (a positivist approach) and others highlighting the pragmatic nature of legal judgment as an interpretive practice (the hermeneutical approach), practising lawyers are more inclined to take a hermeneutical and pragmatic approach, because they are bound to build on the multi-interpretability that is key to the contestability of written and unwritten legal norms and the ensuing argumentative nature of modern positive law. 

The research has produced four types of innovation as to legal methodology (some theoretical, some hands-on) that I will briefly discuss here, complemented by the fifth type of innovation that ties them all together and enabled them in the first place:

1. Framing concepts

The proposal foregrounded three interrelated concepts as framing instruments to rethink the nature of law and the rule of law: (1) the concept of ‘affordances’ that highlights the relational nature of law and the rule of law and the relational nature of foundational legal concepts such as ‘legal subjects’ and ‘legal effect’; at the same time the concept of affordances is a conceptual tool to acknowledge the fact that modern positive law is embedded in a specific information and communication infrastructure (ICI), i.e. that of the script and the printing press, (2) the concept of the ‘mode of existence’ of modern positive law that highlights that and how law-as-we-know-it has thrived on the ICI of the script and the printing press, thus signalling that the transformation of this ICI will also transform the law itself and (3) the concept of ‘legal protection by design’ that highlights the need to deliberately design the emerging computational ICI in ways that safeguard and enhance legal protection rather than diminishing it, calling for close collaboration between lawyers and computers scientists. 

In the Research Study on Text-Driven Law, the COHUBICOL Team has explored the relevance of these framing concepts in the context of rethinking the foundational concepts of modern positive law, thus helping to demonstrate their added value. In the final section of the Study, the interaction between the framing concepts and the foundational legal concepts has been summarised, offering the basics for a new methodology.

2. Vocabulary of foundational legal concepts and key computer science concepts

The Project proposal promised that as part of a new hermeneutics for computational law, a dedicated vocabulary of computer science terms would be developed, next to a dedicated vocabulary of foundational legal concepts. The goal would be to investigate how the introduction of a new vocabulary with different assumptions and different implications might impact the meaning of the foundational legal concepts. 

Foundational legal concepts

During the first two years, the legal COHUBICOL Team explored and discussed the meaning of a limited set of constitutive legal concepts: jurisdiction, legal effect, legal norms, legal powers, legal reasoning and interpretation, legal subject, positive law, Rule of Law, positive law, sources of law and subjective rights. As the Team consists of lawyers who were trained in different legal traditions it soon became clear that most of these concepts are contested and mean different things to different people. Taking a normative position, grounded in a formal, substantive and procedural conception of the Rule of Law, we defined them in a way that honours their open texture while nevertheless staying tuned to achieving legal protection. 

It turned out that the core concept, distinguishing law from both morality and politics, is that of ‘legal effect’, which is best explained in terms of speech act theory, foregrounding the performative nature of legal decision making. By focusing on legal effect as a sui generis effect it became easier to explain the difference between logical and statistical inferences on the one hand and legal effect on the other, while also pinpointing that legal effect is not a matter of causality. These distinctions are also key to explaining legal effect to computer scientists, who are used to logical and statistical inferences (from which they may in turn infer causal relations) that have a different ‘mode of existence’ compared to performative effects. 

In the course of the project we figured out that and how legal technologies may generate major effects on legal effect, thus indirectly transforming the course of the law. The notion of ‘effect on legal effect’ became core to the project, requiring an investigation into the difference between ‘influence’ and ‘performative effect’ as key to a better understanding of computational law. The third Philosophers’ Seminar was dedicated to this issue under the heading of ‘The legal effect of code-driven law’.

The LAW vocab can be found here:

Vocabulary of computer science terms of trade

In the course of the second to the fourth year, the computer scientists developed a similar set of interrelated concepts that must be understood to grasp the assumptions and methods that inform legal technologies. In the context of code-driven legal technologies, the following terms were explained: Domain Specific Language, Markup Language, Declarative programming, Logic programming, Low-code and no-code development, Syntax and semantics, Formal verification, Compilation and interpretation, Application Programming Interface, Literate Programming, Legal Expert System. In the context of data-driven legal technologies, the following terms were explained: Natural language processing, Machine learning, Supervised machine learning, Unsupervised machine learning, Features, N-grams, Tf-idf, Support Vector Machines, Naive Bayes, Neural Networks, Language model, Attention, Transformers, F1-score, Elasticsearch, Decision tree, Regular expressions, Named Entity Recognition, Network Analysis. 

These terms are explained by computer scientists in a way that affords lawyers a taste of how these terms are used, offering a short working definition, a link to related terms in the same vocabulary, a link to relevant legal technologies in the Typology (see next item below), and further reading (offering both scientific papers and more accessible postings). The explanations were tested against the lawyers who worked on finalising the Typology, refining them until sufficiently clear without betraying their salience as computer science terms. 

Developing this vocabulary has been a substantial challenge, requiring and generating a genuine cross-disciplinary conversation, contributing to mutual understanding without resorting to infantilisation of one’s own discipline. As computer science – compared to legal scholarship – is a young discipline, these concepts have not been the object of longstanding scientific debate with little reflection about their precise meaning, the assumptions that must be made to make them ‘work’ and the implications of their use. By forcing computer scientists to explain their meaning with an eye to estimating their impact on law and legal protection, their conceptualisation resulted in a more incisive awareness of different ways of understanding the same word. Our postdoctoral researcher Masha Medvedeva made quite a difference here as she had already demonstrated the different and often misleading uses of the term ‘prediction’ in the context of predictions of judgements with the help of natural language processing. 

Developing, sharing and discussing this type of vocabularies between computer scientists and lawyers will be key to the development and deployment of reliable legal technologies, as this enables an appropriate understanding of the methodology and underlying epistemology that informs the other discipline. It thus exemplifies the cross-disciplinary methodology of the project while contributing to the new hermeneutics of computational law. 

The CS vocab can be found here:

3. The Typology of Legal Technologies

The computer science vocabularies of code- and data-driven legal technologies were developed while constructing and configuring the Typology of Legal Technologies. The Typology has been introduced and explained under Goals and Objectives as a method, a mindset and a resource, meant to inform lawyers and developers of legal technologies about claimed and actual functionality. The Typology was not foreseen in the project’s proposal. It was an initiative of the PI to ensure that the more theoretical approaches that flourished in the project could be tested against and benefit from hands-on engagement with existing ‘legal techs’. In the process of developing the typology, however, it became a two-year project within the project that spurned new directions (such as key attention to the difference between claimed functionality and its substantiation), confronted splendid challenges (as it forced computer scientists and lawyers to join forces in describing functionalities of computational systems with regard to legal tasks), provided grounded incentives to make the idea of cross-disciplinary collaboration work (especially in the context of the vocabularies) and offered a way to share the output of our endeavour with lawyers and developers (the target audience of the Typology).

Building a typology rather than a taxonomy, the idea was not to select and define mutually exclusive concepts but to make relevant distinctions between different types, acknowledging that the relevant tokens would often integrate different types, as types would overlap or be integrated (such as, for instance, in the case of data- and code-driven approaches). 

As explained under Goals and Objectives the Typology became much more than a repository or a database; it became a method and a mindset, ready for dissemination and training of law students and computer scientists working on the cusp of law and tech. When deployed as such in a master course in law and technology in Tel Aviv, it turned out to be an excellent educational instrument, helping students to focus, engage with another discipline on its own terms and confront the potential implications for law and the Rule of Law. The Typology was selected as a tutorial at ICAIL, one of the main conferences in the domain of law and AI, attracting keen interest as an educational tool to train computer scientists working in this domain. 

Laurence Diver built the Typology in the form of an online web tool, eliciting the filters and settings described under Goals and Objectives– in close collaboration with the PI. The Typology thus affords its ‘users’ a process of learning by doing, by (1) inviting various levels of engagement (a generic level, a technical level, a legal level and a theoretical level) and by (2) offering incentives to study and apply the computer science vocabulary (which is hyperlinked in the Typology). 

Besides being a methodology, the Typology also provokes a critical mindset, as the user finds out that functionality claims cannot be taken for granted, while contributing to the ability to obtain a detailed understanding of the design decisions that make a difference to law, the Rule of Law and legal protection. We believe that the Typology is a major output of the project and offers a hands-on demonstration of the new hermeneutics promised in the proposal. 

The Typology has been explained in more detail under Goals and Objectives and can be found here:

4. Cross-disciplinary method in the context of the JCRCL

To disseminate, develop and foster the cross-disciplinary method that is key to the new hermeneutics of computational law, the PI founded a new scientific journal, together with postdoctoral researcher Laurence Diver: the Journal of Cross-Disciplinary Research in Computational Law (CRCL). The Journal is a peer reviewed online diamond Open Acces Journal, neither the authors nor the subscribers or visitors pay for access.

Frank Pasquale joined as co-editor in chief with Hildebrandt and over the course of four years the Journal has engaged a fabulous team of handling editors

The Journal has published 15 high quality articles, with 7 or 8 submissions to be published later this year and a pending special issue of 8 high-profile position papers on the ‘future of computational law’.

The goal of the Journal is to publish excellent work in the domains of law and computer science with regard to data- and code-driven legal technologies and their hybrids, including social science and humanities perspectives. The added value of the Journal is to be found in its unique format. After a paper has been accepted following double blind peer review by reviewers from the discipline(s) of the authors, the editors invite a brief cross-disciplinary reply by a scholar or scientist from ‘the other discipline’ that is followed by a response of the author(s). This affords a combination of a main text that is tested against the methodological integrity of the own discipline and a cross-disciplinary conversation that uncovers misunderstandings or misrepresentations of the ‘other discipline’. That combination offers a way forward that avoids interdisciplinary soup while taking both the own and the other discipline(s) seriously. The aim of the Journal is not to compete with existing Journals or Proceedings in the domain of AI and Law, such as ICAIL, JURIX, or the NLLP workshop, but to offer a space where frontline research in law and computer science can meet to investigate and reflect on their integration in ‘legal techs’ and the consequences of their deployment.

In April 2024 we hope to publish the invited position papers of the CRCL23 Conference under the heading of ‘The Future of Computational Law’, written by high profile scholars with relevant expertise: Sarah Lawskey, Sayash Kapoor & Peter Henderson & Arvind Narayanan, Lyria Bennet Moses, Frank Pasquale & Gianclaudio Malgieri, Floris Bex, Natalie Byrom and an introductory paper by Mireille Hildebrandt. In a sense this Journal, funded by the ERC grant, even though not planned in the proposal, testifies to the salience of the project, its methodology and the need to invest in cross-disciplinary research to support law, the Rule of Law and legal protection. This special issue follows the Symposium on ‘The Future of Computational Law’ that took place on the first day of the CRCL23 Conference.

5. The COHUBICOL website as a platform for sharing different types of scientific output

The COHUBICOL website is not merely a platform to post news, publications, blog postings and other output. Instead, it fosters an integrated approach to the affordances of an online web platform that offers both static and dynamic ‘content’ and allows ‘visitors’ to become ‘users’ by navigating a range of interconnected scientific outputs whose ‘mode of existence’ as scientific output is transformed by design choices that allow for new ways of interacting with the information that is shared. This notably concerns the interrelated project outputs of the Typology, the Law and CS vocabs, and the two Research Studies. The Typology enables quick mapping, comparing and in-depth assessment based on filtering and smart navigation settings, while it links to a range of ‘how to use’, ‘methodology’, ‘FAQs’ and other guidance. The Typology also links to the vocabularies that provide short working definitions with – in the case of the CS vocab - hyperlinked references to scientific papers as well as easy to read blog posts and so on; thus, the CS vocab offers different layers of analysis and understanding, depending on one’s background and interest. The Law vocab is deeplinked with chapter 3 of the Research Study on Text-Driven Law, thus offering both relatively short working definitions and in-depth discussion. The CS vocab also deeplinks to the Typology, especially with respect to the substantiation of functionality claims. The Research Studies are offered in both pdf format and in a html format that allows deeplinking to relevant paragraphs within the Study, thus offering a different type of access and dissemination compared to traditional publications of legal text. 

Postdoctoral researcher Laurence Diver, who has a professional background in web design, developed the website, the presentation of the Typology, the vocabs and the Research Studies, in close collaboration with the PI, basically offering a new way of sharing, presenting, navigating and accessing scientific output. Though webtools and html text are not new in themselves, the integrated, dynamic and cross-referential nature of the COHUBICOL website turns it into an interesting example of how to share, present, navigate and access research findings.