From persistence infrastructure to legal knowledge
Legal AI is often described as a retrieval problem. That is too narrow. A legal system is not merely a collection of documents: it is a network of authorities, propositions, holdings, exceptions, standards, burdens and relationships.
The design challenge is therefore architectural. Reliable legal AI needs to preserve authority, provenance, structure and contestability as information moves from source material into machine-assisted analysis.
The architectural question
A useful legal knowledge system should be able to distinguish the text of an authority from the proposition extracted from it, the issue to which that proposition relates, and the later authorities that support, limit, distinguish or reject it.
That is the direction I am exploring through LegisAlta.