Legal technology interface illustrating predictive coding and review workflows

THE TAR TRILOGY: Three Cases that Transformed TAR in eDiscovery

August 1, 2025

Lucy Robinson · Right Discovery Staff Writer

Hon. Andrew J. Peck, retired SDNY magistrate and patron saint of predictive coding doctrine, sat down with our editors to revisit the "TAR Trilogy"—three decisions reframing how litigators negotiate algorithmic review. The calendar also notes nine years since Hyles v. City of New York recalibrated party autonomy debates.

What "TAR" means here

Technology-assisted review—predictive coding—today spans TAR 1.0 seed-driven workflows and TAR 2.0 continuous active learning (CAL) regimes where models evolve as senior reviewers code on the fly, shrinking rigid starter sets while demanding disciplined QC storytelling.

The TAR guru himself

Judge Peck's Search, Forward scholarship benchmarked algorithmic stacks against keyword purgatory—cheered by some, jeered by others, but impossible to ignore once productions ballooned past human linearity.

1. Da Silva Moore v. Publicis Groupe & MSL Group

February 24, 2012: first federal endorsement of predictive coding inside a sprawling gender-discrimination docket—proof courts could bless statistically grounded sampling when transparency and meet-and-confer hygiene held.

2. Rio Tinto Plc v. Vale S.A.

April 2015: both camps chose TAR; Judge Peck cemented presumptive deference— absent prejudice, producing parties may deploy predictive workflows without opponents vetoing math they dislike. TAR stops being exotic liability; it joins keyword and linear methods as responsible options when proportionality demands it.

3. Hyles v. City of New York et al.

Plaintiffs sought to compel defendants onto TAR; Judge Peck refused. Parties choose their discovery modality absent abuse—opening space for good-faith negotiation instead of judicial coding mandates—even when he personally favored algorithmic efficiency.

Together the trilogy normalized transparent statistical review, empowered negotiated adoption, and preserved autonomy—heralding today's GenAI disputes about disclosure, validation, and judge-facing narratives.

Topics: predictive coding, TAR, CAL, Da Silva Moore, Rio Tinto, Hyles, Judge Peck, Right Discovery, eDiscovery precedent