Selected Publications

The landmark empirical comparison of technology-assisted and manual review.

A shared vocabulary for lawyers, judges, vendors, and researchers.

A near-turnkey Auto TAR process designed to work reliably across topics and collections.

Reliable high recall with a fixed amount of training effort, regardless of corpus size.

Independent assessment of an unequal-inclusion-probability sample as the basis for fair comparison.

A stress test of whether observed differences survive chance, bias, protocol variation, and assessment disagreement.

Why a particular assessor’s judgments cannot serve as an infallible gold standard.

Statistically sound comparison of review effectiveness without assuming perfect relevance assessments.

A pointed argument against untested tools, inherited practice, and pseudo-scientific validation.

A practical framework for testing the validity, reliability, and admissibility of AI-based evidence.

Practical guidance for courts confronting authentic, disputed, and deepfake AI-generated evidence.

A framework for authentic AI output and disputed deepfakes.

A rapid review that used CAL to accelerate evidence synthesis during the pandemic.