The papers.
Five on SSRN, one of them under review. Publication status is stated in each paper’s own words; nothing is described as published until it is accepted.
Method, taxonomy, and the open TAR methodology.
The Promise Fulfilled: How Proposition-Level Legal Knowledge Representation at Corpus Scale Realizes Three Decades of Computational Jurisprudence
Document-level legal AI fails in five structural ways that agentic retrieval cannot fix. Proposition-level extraction across the published corpus with typed metadata makes structural reliability achievable, validated on a stratified 700-case sample across 22 active knowledge dimensions.
A system can hallucinate zero citations and still serve dead law as live law.
Detecting Genuine Doctrinal Ambiguity: A Multi-Layer Framework for Identifying Structural Indeterminacy in Judicial Reasoning
Shallow ambiguity is resolvable by more research; genuine ambiguity is a structural feature of the law. A five-type taxonomy, an eighteen-test framework across extraction-time, aggregation-time and structural detection, a pre-registered validation protocol, and a promised open benchmark.
A tool that erases contestedness erases the very property that determines how much independent judgment its answer requires.
isResponsive: An Open TAR Methodology for Court-Defensible Responsiveness Review
Court-defensible technology-assisted review does not require a proprietary black box. Four controls (hash-locked rubrics, a calibrated cascade under recall constraints, a deterministic privilege screen, and a formal validation stopping rule) end in a counsel-signed Defensibility Certificate, measured axis by axis against the leading validation order.
The system does not ask whether the point estimate looks good. It asks whether the lower bound clears the floor.
Good Law for What? A Proposition-Usability Model for Verification in AI-Assisted Legal Research
“Verify the case” understates the task. Four proposition-level questions replace it (still good law; governs here at this posture; whom it helps; already rejected here), with seven demonstration-based procurement questions any vendor, including the author’s, can fail.
A case is not a unit of law.
Ariadne’s Thread: Measuring Legal Openness as Artificial Intelligence Measurement Infrastructure
Hold the verified law fixed, vary only the judicial posture, replicate to estimate instrument noise, and report the noise-corrected between-posture dispersion at a procedural gate as a contestability artifact for human review. The instrument is interpretable because no generative model writes a citation or a validity verdict.