research essay / Legal AI and Computational Law
Law as a chain of events
Why legal analysis needs an event-based ontology
An event model separates documents, facts, actions, and legal qualification instead of treating law as a flat collection of text fragments.
In brief
- A document is not an event.
- Retrieval should support reconstruction, not similarity alone.
- A legal conclusion needs an explicit interpretation path.
The problem is not purely textual
A document describes an event, but it is not the event itself. A filing, decision, payment, and service notice may concern one matter while carrying different dates and procedural effects. Flat passage retrieval can easily merge those layers.
A Legal AI result should identify not only matching text, but also the document's role, participant, time, and relationship to other actions.
Working hypothesis
Analysis becomes more controllable when the corpus is represented as a graph of events and evidence. Nodes stand for actions, documents, parties, and qualifications; edges preserve explicit temporal and procedural relationships.
A minimal model
- Register the source, version, and scope of each document.
- Separate a mention of an event from an established fact.
- Connect participant, time, and procedural role.
- Record the rule that links reconstruction to qualification.
- Preserve contradictions and missing data.
How to test it
A benchmark should test reconstruction of the path, not similarity of one passage. Each scenario includes expected events, sources, relationships, and a failure the system must not accept.
The most important negative case is a convincing chronology assembled from documents concerning the wrong person or the wrong period of applicable law.
Practical conclusion
The ontology is a contract between data, retrieval, and interpretation. If that contract cannot be shown and tested, the final answer remains hard to audit regardless of model quality.
01 / METHOD
The question the test must answer
An event model helps order the path from document to decision: a mention is not a fact, a fact is not a qualification, and a qualification is not yet a professional decision. The value of the material is that a reader can separate the user question from a convenient tool metric and identify the point at which a result is no longer sufficient for the next stage of work.
02 / METHOD
Material and counterexamples
A minimal schema records document, source, event, time, actor, rule, interpretation act, and certainty status. A relation must let a reader return to material, not merely to a generated assertion. Control material is not meant to confirm the author’s thesis; it should expose a case in which a plausible result is dangerously misleading for the workflow at hand.
03 / METHOD
Control method
A working timeline places competing records together, names their versions, and separates observation from interpretation. QA looks for missing dates, confused actors, duplicate events, and a citation without a boundary. The account includes step order, input, and result trace so an independent person can inspect what was measured and what the procedure does not measure at all.
04 / METHOD
Acceptance matrix
A package is ready for review when a professional can move from a decision to a claim, from the claim to an event, and from the event to a document and its version. Criteria are stated before the result is read, so a single successful example cannot substitute for control of a critical failure type.
05 / METHOD
Handoff and rerun
The event ontology provides a working reference for fields and relations; a data-model change should be versioned with an example of its effect on an earlier timeline. Reproducibility also means showing why a subsequent run differs from an earlier one and who decides whether a new version may be used.
06 / METHOD
Boundary of the conclusion
The schema does not determine factual truth or legal qualification. It helps expose material and uncertainty but needs professional interpretation in a particular matter. The material is therefore methodological: it does not pose as a universal recipe or as an independent confirmation that has not been performed.
07 / METHOD
A worked check
A working timeline can begin with two records of the same event. Each record receives source, version, time, actor, and certainty status; only then are interpretation and decision added. QA checks whether a citation actually supports the named event rather than merely discussing a similar topic.
08 / METHOD
Source and review record
The event ontology is a reference point for fields and relations here, not an engine for legal qualification. Handoff should preserve mapping, an uncertainty example, and a model-version change so a later researcher can challenge the chronology instead of inheriting a finished narrative.
LIM
Limitations and scope
This is an analytical model proposal, not a complete ontology for every legal system.
SRC
Sources and external edition
Original or earlier Medium material
AI tools assisted with structure and editing. The final text underwent human editorial review of facts, sources, conclusions, and attribution.
LOG
Change history
- First expanded edition in the owned archive.
- Review of structure, limitations, and evidence links.