Research
Understanding law in practice and helping people navigate legal institutions.
My research combines legal analysis with computational modelling and the development and evaluation of working systems. It follows two connected lines: investigating legal reasoning across large collections of decisions, and using those insights to support access to justice.
Computational analysis of legal data
I develop methods for empirical and qualitative research across legal corpora. An early study combined legal analysis, annotation and computational modelling to examine how legally relevant circumstances relate to outcomes in landlord–tenant decisions. I subsequently developed CAESAR, an environment using sentence embeddings and similarity search to help researchers explore and annotate legal texts. This work received the JURIX 2020 Best Paper Award.
My collaborative work investigates how language models can support annotation requiring specialist legal knowledge and thematic analysis in empirical legal studies. The latter moves beyond predefined categories to help researchers discover themes across large collections of factual descriptions, while retaining the human investigator’s role in setting priorities and refining research questions.
I also investigate connections between legal rules and their application in case law. From Text to Structure, which received the JURIX 2023 Best Paper Award, studies how language models can extract structured legal pathways from legislation. Automated Mapping of Legal Criteria examines how such representations can identify whether decision-makers found particular criteria satisfied. Collaborative work on the European Court of Human Rights investigates the Court’s use of external sources of legal authority.
Across these projects, the aim is to make patterns in facts, criteria, themes and sources of authority available for systematic inquiry. Computational representations need to be validated against the legal material so that they support, rather than replace, substantive interpretation.
Computational insights for access to justice
I conceptualised and led the development of JusticeBot, which combines legal rules and information from decisions into guided pathways for people without legal training. Its authoring environment, JusticeCreator, allows legal experts to develop and revise those pathways. A version deployed with the Quebec Housing Tribunal supports people facing landlord–tenant disputes. These systems provide a practical setting for studying how legal information can be made usable outside professional practice.
I study how AI can lower barriers to obtaining this support, including mapping layperson narratives to legal issues, extracting information from images of legal documents, and analysing videos of legally relevant situations.
Other projects support tasks within legal processes. I developed DALLMA to investigate semi-structured legal reasoning and drafting, and LLMediator as an experimental environment for AI-assisted communication and dispute resolution. Related studies evaluate language models as mediators and investigate intake in civil legal services.
My aim is to connect advances in computational capability to evidence about what helps people navigate the legal system. Developing working systems makes it possible to examine whether people understand the guidance, recognise its limits and can act on it appropriately.
Connecting the two research lines
My research programme links the analysis of institutional reasoning with the evaluation of practical legal support. I am interested in how relationships between facts, rules and reasons can be reconstructed across collections of decisions, and how those findings can inform explanations, guided pathways and document support. This calls for both expert review of computational analyses and research into users’ understanding and experiences.