LEGAL REASONING RESEARCH
We study the questions that determine a case: how facts, evidence, legal rules, and arguments support or challenge a conclusion. Our aim is to make that reasoning explicit, traceable, and open to examination.
Make the structure of the reasoning explicit.
Our research examines which legal requirements must be met, what the record supports, who bears the relevant burden, and how competing interpretations affect the analysis. We seek representations that preserve the distinctions between evidence, inference, and legal judgment.
We are developing a language and compiler to facilitate accurate, explicit representation of facts, legal rules, arguments, and their relationships for litigation and adjudication. The goal is to make those representations available for structured checks while keeping assumptions, source support, and unresolved questions visible.
Representing the Matter
A factual allegation, a piece of evidence, an inference, and a legal requirement play different roles. We investigate representations that preserve those distinctions while connecting propositions to their sources and to the issues they may support or challenge.
This includes identifying missing information, conflicting accounts, and assumptions that would otherwise disappear inside a fluent narrative.
Procedural Context & Burdens
Legal analysis depends on the decision at issue, its procedural setting, and the applicable standards and burdens. We study how to represent that context explicitly, so a reasoning process can be evaluated against the question it is meant to address.
The focus is on making those dependencies visible: what is assumed, what must be established, and which unresolved questions limit the conclusion.
Arguments & Outcome Dependencies
We investigate how factual premises, legal requirements, exceptions, and competing arguments can be connected in an inspectable structure. This makes it possible to examine where an argument depends on a disputed fact, an interpretation, or an unresolved evidentiary gap.
We are particularly interested in how changes propagate through the analysis: whether new evidence alters a premise, whether contrary authority challenges a rule, and which conclusions must then be reconsidered.
Neurosymbolic Execution & Evaluation
Neural methods can help interpret source material and propose relevant relationships. Symbolic representations and rules can make specified checks explicit and repeatable. Our research explores how these methods can work together while preserving source support and exposing the assumptions introduced along the way.
We distinguish reproducible execution of a defined check from correctness of the legal interpretation it relies on. Deterministic steps remain dependent on the quality of the inputs, representations, and rules.
RESEARCH IN PRACTICE
Test whether the reasoning changes for the right reasons.
Our approach is to develop bounded research exercises in which the inputs, rules, assumptions, and expected checks can be examined. Synthetic matters can isolate a particular reasoning problem without introducing client information.
For example, an exercise might connect a required proposition to supporting and contradictory evidence, then remove the only supporting source. The test would examine whether the system identifies the unsupported premise and revisits the dependent conclusion, rather than preserving an answer the record no longer supports.
These exercises help distinguish a persuasive explanation from reasoning that remains faithful to its stated premises and constraints.
Develop methods. Establish their limits. Evaluate their use.
Our work in dispositive reasoning is an ongoing research effort. We distinguish proposed methods, experimental results, and capabilities evaluated for a particular use. The language and compiler are under development; their utility must be established through testing. Practical implementation helps identify further questions for research.