0. Introduction/Executive Summary

This paper is dedicated to the memory of David Kittrell, a long-time pioneer and contributor in the field of eDiscovery and an early member of the drafting team for this paper.  David passed away in September 2020 and will be sorely missed and fondly remembered.

Executive Summary

This brief survey provides a description of how artificial intelligence, or “AI,” is currently used in areas related to eDiscovery. Its primary purpose is to provide background information on the use of artificial intelligence in eDiscovery. Artificial intelligence and eDiscovery are both large, complex, rapidly changing fields and it is impossible to do justice to either of them in a paper of reasonable length. There are many books about AI and a few about eDiscovery. Our intention is to provide a preliminary overview of the field.

There is no universally accepted definition of “artificial intelligence.” We adopt a working definition that artificial intelligence refers to the capability of machines to mimic aspects of human intelligence, such as problem solving, reasoning, discovering meaning, generalizing, predicting, or learning from past experience. Of the different types of AI, machine learning is probably the most prominent and most familiar to eDiscovery practitioners.

Although machine learning receives the bulk of attention in the context of AI in eDiscovery, rules-based systems also meet the working definition of artificial intelligence and can play an important role in eDiscovery and other legal activities.

Applications of AI in eDiscovery include document categorization (e.g., technology-assisted review (“TAR”) or predictive coding), identification of personally identifiable information (“PII”), investigations, and some forms of early case assessment. AI can offer significant cost and time savings relative to previous unautomated methods. AI should not be used by persons without an appropriate understanding of its capabilities and limitations. Doing so, or uncritical acceptance of its results, can lead to errors and even potential ethical issues.


Download the complete paper here.