Legal AI / participatory designlegal AI / e-discoveryParticipation documented

Legal Track of TREC Participatory Design

Ithaca, United States

"The evaluation goals and design methods of Legal Track were in large part driven by domain practitioners rather than computer scientists, despite it being a computational research venue. Further, we find that by structuring design and evaluation activities through the format of an extended role-play simulation, Legal

participatory designhuman-centered AIproblem formulationdesignevaluation

Project documentation

Overview

Region: North America
Lead organization: Cornell University
Organization type: nonprofit + university collaboration
Technology group: information retrieval and legal document review
Activity status: completed
Start year: Not documented
Last updated: Mar 25, 2026

Participation documentation

Participation mode: co-design with affected users and domain experts
Participants: lawyers; computer scientists; litigation practitioners
Methods: role-play simulation; interviews; co-design
AI lifecycle stages: problem formulation; design; evaluation
Evidence summary: "The evaluation goals and design methods of Legal Track were in large part driven by domain practitioners rather than computer scientists, despite it being a computational research venue. Further, we find that by structuring design and evaluation activities through the format of an extended role-play simulation, Legal
Atlas assessment: Core · High confidence · evidence grade A
Uncertainty: Public evidence indicates that stakeholders directly shaped AI design, data, deployment, evaluation, or oversight; residual uncertainty concerns the exact degree of shared decision power rather than project existence.