GenAI as an Alternative Approach to Technology-Assisted Review

The emergence of Generative Artificial Intelligence (GenAI) has triggered discussions within the legal community about its potential as an alternative to traditional Technology-Assisted Review (TAR) for document review in discovery, as reported in a blog post by eDiscovery Today. 

Author Doug Austin cites a recent Sedona Conference Journal article titled “TAR 1 Reference Model: An Established Framework Unifying Traditional and GenAI Approaches to Technology-Assisted Review,” explaining the integration of GenAI into established TAR workflows.

Rather than being in conflict, the article suggests that GenAI and TAR are complementary and propose a GenAI TAR 1 approach, which integrates GenAI prediction algorithms into the TAR workflow, following the TAR 1 Reference Model. These are the steps included in the model: Scope, Label Control Set, Iterate Model, Classify, and Validate. 

Although the TAR 1 and GenAI TAR 1 approaches are very similar, a key distinction lies in the Iterate Model steps (Evolve, Encode, Apply, and Evaluate), which can be repeated until completion of the assessment. Both approaches share similar steps and sub-steps, but GenAI replaces traditional steps such as selection and review of additional training documents with the creation, application, and iteration of GenAI prompts.

Other important considerations for the GenAI TAR 1 approach compared to that of TAR 1 include precision and recall, risk of sensitive information disclosure, project cost and time, and ease of use. There is also a potential for hybrid workflows and integration of GenAI with other eDiscovery tools.

As GenAI capabilities improve and associated costs decrease, GenAI has the potential to become a preferred approach to TAR 1. In the interim, GenAI can be aligned with established TAR workflows that involve sampling and statistics and will promote successful outcomes for first-pass review projects.