Will generative AI replace technology assisted review (TAR)? Here are six reasons why gen AI may replace TAR…someday.

But first: We Need Your Help! The eDiscovery Today State of the Industry Report survey is currently live, with 12 questions about the state of the eDiscovery industry, including questions on generative AI, technology assisted review (TAR), mobile device and collaboration app discovery, eDiscovery use cases, hyperlinked files, and more! Takes two minutes to fill out. Please check it out here – doing so will get you a FREE copy of the report when it’s published in January.

While TAR/Continuous Active Learning (CAL) remains valuable for targeted, large-scale document reviews, generative AI opens up new possibilities for more comprehensive, efficient, and user-friendly approaches to eDiscovery. Here are six reasons why gen AI may someday replace TAR (at least in my opinion):

1. Enhanced Contextual Understanding: Gen AI models, like ChatGPT, excel at understanding language nuances, including contextual subtleties. This means they can identify relevant information even when it’s expressed in indirect or complex language. TAR often relies on keywords and pattern recognition, which can miss nuances that generative AI can interpret more accurately.
2. Improved User Experience: With its natural language capabilities, gen AI can provide a more user-friendly and interactive experience, allowing reviewers to query data, generate summaries, and gain insights conversationally. This reduces the learning curve compared to the use of TAR tools, which may have complex workflows. Gen AI can answer specific, open-ended questions, enabling a more conversational approach to information retrieval. This makes it easier to refine searches dynamically and explore emerging questions without needing to re-train or configure models specifically, as is often needed with TAR.
3. Improved Adaptability and Learning: Gen AI models are more adaptive to new types of data and information as they can handle multiple domains of knowledge and learn from broader data sets. Unlike TAR, which may require separate training sets for different topics or document sets, generative AI models can generalize more effectively across various legal and factual domains.
4. Potential for Predictive Insights: By analyzing patterns and language, gen AI can provide predictive insights, like estimating the likelihood of relevance or predicting future queries based on ongoing review outcomes. This predictive capability enhances strategic planning, helping legal teams prioritize and streamline review.
5. Broader Range of Applications: While TAR focuses specifically on document review and relevance scoring, gen AI can be applied across a wider array of eDiscovery use cases. For example, it can assist with document summarization, drafting review protocols, generating potential queries for custodians, and even synthesizing complex sets of documents to aid case strategy. Let’s see TAR do that! 😉
6. Greater Transparency for Document Classifications: When applied to review, gen AI provides greater transparency for document classifications, as most implementations of gen AI for document review provide a description for how it classified the documents, as well as a confidence level for its prediction. With TAR/CAL, you get a ranking of documents in terms of their likely responsiveness, but no description of why the algorithm ranked the documents the way it did.

Does that mean it’s a done deal that gen AI will replace TAR? Not necessarily. TAR is built on established processes, which gives legal teams more control over the accuracy and relevancy of the results – essential for maintaining compliance with legal standards and defensibility in court. TAR is highly effective at handling well-defined, repetitive tasks that are common in eDiscovery, such as relevance and privilege filtering. TAR systems are typically designed and optimized specifically for document review tasks, allowing them to perform with high efficiency and precision once they are properly trained.

Probably the biggest question about gen AI is its reliability. We’ve seen instances where gen AI models have hallucinated or generated different responses to prompts that are identical or only vary slightly. Unpredictability is a huge barrier when you have opposing counsel pushing you to provide transparency about your process.

There are six reasons why gen AI may replace TAR…someday. However, today is not that day. 🙂

So, what do you think? Do you think that gen AI will someday replace TAR? Please share any comments you might have or if you’d like to know more about a particular topic.

Image created using Bing Image Creator Powered by DALL-E, using the term “robot holding up six fingers”. Hey, it’s a robot, you can give it as many fingers as you want! 😀

Disclaimer: The views represented herein are exclusively the views of the author, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.