About the Institute

Mission

The Institute's charter, in full:

The Computational Law Institute exists to turn fifty years of legal reasoning theory into standards the profession can test, rerun, and rely on. Its mission is to define what "filing-grade" means for machine-assisted legal work and to measure it in the open: not merely whether an AI system fabricates citations, but whether it finds the controlling authority, surfaces the adverse precedent, confirms that a holding is still good law, and shows the exact reasoning path from precedent to conclusion. The Institute pursues this through four commitments. It publishes open, rerunnable benchmarks with public rubrics that grade completeness and omission alongside accuracy. It develops shared formal vocabularies for holdings, treatment, and the recognized modes of legal inference, so that reasoning can be audited rather than trusted. It advances the principle that reliability in legal AI must be designed into a system's architecture, not inspected in after the fact. And it maintains a public record of AI citation failures in court, so that bar associations, judges, and malpractice carriers can set policy on evidence rather than anecdote. The Institute builds on open law, publishes its methods, and invites replication by any vendor, court, or scholar, because in a profession that answers to judges, a standard only matters if anyone can check it.

Where the Institute came from

The Institute grew out of the research program that Ross Brodskiy built around Legawrite.AI, the legal technology company he founded in Aliso Viejo, California. Building a system that lawyers could file from kept raising questions that were not engineering questions at all. When is the law genuinely unsettled, and how would anyone know? What makes a real, accurately quoted case unusable in a particular motion? Which inferences from precedent would a court accept, and which only sound as if it would? Who carries the burden, and what happens to a model when the Supreme Court moves it?

The answers accumulated as working papers on SSRN, essays on Medium and a larger body of working drafts. The Institute publishes that program openly and in one place, with stable addresses, version histories and citations, so that the ideas can be read, tested and argued with on their merits. Legawrite.AI remains the founding sponsor and the applied testbed. What that relationship means, and how it is disclosed, is set out in Independence and Disclosure.

What "applied computational law" means here

Computational law is not new, and the Institute does not pretend otherwise. In 1957 Layman Allen proposed normalizing legal text into logical form, so that one could retrieve the structure of a rule rather than every document that mentioned it. In 1977 L. Thorne McCarty's TAXMAN reasoned over the conceptual structure of tax law. Kevin Ashley's HYPO (1990) modeled the way lawyers cite a case, distinguish it and rebut the distinction, and the CATO work organized the facts of cases into side-relative factor hierarchies. Henry Prakken and Giovanni Sartor formalized defeasible argument, Trevor Bench-Capon developed argument schemes with the critical questions that attack them, Branting treated the ratio decidendi as a structured justification, and the Carneades model took on burdens of proof in legal dialogue.

That tradition was rigorous, and it was largely right about what legal reasoning requires. Its limits were practical. Systems reached production where the law was rule-like, as in statutory domains, and stayed in the laboratory where it was not. Case-based systems were deep but narrow, each resting on a hand-built model of a single domain. Argumentation frameworks worked beautifully once their input had been structured, and assumed that someone else would do the structuring. The hardest step, extracting holdings, treatment and inferences from unstructured opinions reliably and at scale, was the one left undone. The longer history is told in a Library essay on the road to applied computational law.

"Applied computational law", as the Institute uses the term, starts at that missing step.

The last row matters most. When the Institute's founder ran a blind eDiscovery review with no humans in the loop, the write-up reported the gates the run failed alongside the ones it passed, and the engine was released under an open-source license so that others could rerun it. That is the standard the Institute sets for its own work and for everyone else's.

The four commitments

1. Open, rerunnable benchmarks. Public rubrics that grade completeness and omission alongside accuracy, because a motion is lost on the controlling case nobody pulled as often as on a bad citation. See what a filing-grade benchmark must measure and The Docket Test.
2. Shared formal vocabularies. Named, versioned definitions for holdings, treatment and the recognized modes of legal inference, so that reasoning can be audited rather than trusted. See the Lexicon, the Case Treatment Taxonomy and the Twelve Bridges.
3. Reliability by design. Guarantees that live in a system's architecture rather than in a checker bolted on afterwards. See the Four Laws and The Verification Layer Is a Smoke Alarm, Not a Building Code.
4. A public record of AI citation failures in court. An evidence base for the bar associations, judges and malpractice carriers who have to set policy, maintained so that policy can rest on evidence rather than anecdote.

What the Institute does not do

The Institute is a research publisher and nothing else. It runs no events and takes no speaking engagements. It offers no consulting and no other services. It has no membership, no summits and no sales funnel.

There are three things a visitor can do here:

• Read. Everything the Institute publishes is in the Library, organized by research program.
• Cite. Every piece has a stable address and a citation block. See How to Cite.
• Subscribe. New research by email or by feed, with no marketing. See Subscribe.

Who writes here

The Institute's founder and director is Ross Brodskiy. Nathan Pokov is co-author of Ariadne's Thread. The Library also publishes pieces under ten named house voices. These are editorial personas, not real people and not outside contributors, and the Institute explains how and why it uses them in its Editorial Standards.