SFU & Caseway use AI to make court decisions searchable

Simon Fraser University has started a legal artificial intelligence research collaboration with Vancouver-based Caseway focused on making court decisions searchable in formats that work for modern AI systems.

The work centres on a research question about people who go through the justice system without legal representation. The collaboration will test whether improved access to primary legal texts changes how those individuals understand legal options and make early decisions.

The partners plan to build infrastructure that indexes more than 100 million Canadian and US court decisions in machine-readable formats. The project will also examine how retrieval systems and large language models use those decisions as sources.

Primary sources

The collaboration comes amid ongoing debate about the reliability of general-purpose AI tools for legal questions. Researchers and legal practitioners have raised concerns that some systems generate inaccurate legal information when they cannot access court decisions directly. In those cases, systems can draw on forums, blogs and social media material.

Caseway's work focuses on structured publishing of judicial decisions. The approach aims to make decisions usable for retrieval systems and language models. It also links AI outputs back to the underlying decision text, so users can check an answer against the judicial record.

Professor Angel Chang of the School of Computing Science leads the academic research effort at Simon Fraser University.

"Our aim is not to replace lawyers or automate legal advice," said Angel Chang, Professor, School of Computing Science, Simon Fraser University. "We are asking a measurable, evidence-driven question. If individuals without lawyers have access to accurate, searchable court decisions through systems that leverage artificial intelligence, does that change how they understand their legal options and make early decisions? That's what we want to evaluate."

Scaling search

The collaboration includes a data engineering programme that aims to index a large volume of judicial decisions. The partners said the work includes creating machine-readable formats and making decisions discoverable by humans and AI systems.

The indexing effort forms part of a Mitacs-supported research project. The organisations said technical and prototyping work has started.

The partners described a focus on retrieval. Retrieval systems typically locate and rank documents for a user query before a language model generates an answer. In a legal context, retrieval can determine which cases and passages a system presents as supporting material.

Caseway framed the issue as a data access problem for legal AI systems. "This research is about fixing the data foundations that legal AI systems rely on," said Alistair Vigier, CEO, Caseway. "Right now, most AI systems answer legal questions by pulling from Reddit threads, blogs and second-hand commentary because the real court decisions simply aren't accessible to them. Our goal is to change that by making official judicial decisions searchable at scale and usable by modern language models, so when artificial intelligence explains the law, it can point directly to the same sources judges rely on."

Measuring outcomes

Rather than assessing whether AI can provide legal advice, the project will measure outcomes for self-represented litigants. The partners said they will look at comprehension of precedent and legal standards. They will also examine confidence in identifying next steps such as deadlines or document requirements.

Other planned measures include the quality of early strategic decisions, such as selecting venue or framing legal issues. The project also plans to assess whether users can cross-check AI explanations against linked primary sources.

Under Chang's supervision, graduate students at Simon Fraser University have begun prototyping parts of the retrieval system. The work includes embedding experiments and evaluation of ranking quality across jurisdictions and legal topics.

That technical work will feed into human-centred studies, according to the partners. Those studies will examine how people interact with search results and AI-assisted explanations when preparing for legal proceedings.

Wider context

The partners linked the initiative to a broader push in Canada for domain-specific AI systems grounded in primary sources. They also pointed to related research involving the University of British Columbia on reducing hallucinations in legal research tools and improving reliability.

Access to justice remains a longstanding issue across many jurisdictions. A significant share of people go through legal processes without representation. Cost pressures and procedural complexity can add risk for individuals who rely on public legal information or online materials.

"Our work is designed to move AI research in law from hype to evidence," said Chang. "We want measurable insights about how better data access impacts real people facing legal uncertainty."

The collaboration will continue with further development of indexing and retrieval infrastructure alongside planned user studies, with the partners focusing on evidence about how people use machine-searchable court decisions during early legal decision-making.