AI in Legal Research: Casetext and LexisNexis Battle It Out.
Searching was limited to terms and connectors (Boolean searching). We were taught to identify key terms, then connect them into a search using AND, OR, and NOT, along with some unique (at the time) proximity connectors. Lexis would then show you the search results, including its famous KWIC (Key Words in Context), so you could scan your results quickly. By my last year of law school, we had also obtained a Westlaw terminal with similar features. I had a part-time job as a reference desk attendant at the Idaho Law Library, which gave me more than the normal exposure to Lexis and Westlaw, in turn leading me into my profession.
The AI Revolution
Cut to 34 years later, and the AI revolution is sweeping through a wildly expanded and wildly competitive legal research database marketplace. In addition to LexisNexis, Westlaw, and Bloomberg Law (the largest providers), which are now owned by multinational conglomerates RELX Group, Thomson Reuters, and Bloomberg, L.R, respectively, the market is filling with a growing number of startup AI-driven entries, including ROSS Intelligence, Fastcase, Casemaker, and Casetext.
Casetext (casetext.com) was founded in 2013 by a group of former litigators working with coders to develop "new approaches to legal research." Its platform, launched earlier this year, is CARA A.I., an "artificial intelligence driven technology" that takes a different approach to legal research.
Context Is King
In legal research, context is king, and it is often the major challenge when searching online. In order for a particular precedent to be valuable to the researcher, it not only has to address the legal issues involved in the matter, it also must be contextually sound. It must have relevance to the facts and circumstances of the matter that the lawyer is researching. CARA A.I. addresses this through a unique process--separate from its primary search engine--that uses AI to review legal briefs and other documents uploaded by the researcher to evaluate them for context and identify the most relevant results based on that context. Casetext asserts that this will allow the researcher to not only save time, but also to have confidence that the results are relevant both substantively as well as contextually.
A recent study by the National Legal Research Group, Inc. (NLRG) would seem to endorse that position. The report, "The Real Impact of Using Artificial Intelligence in Legal Research" (law sitesblog.com/wp-content/uploads/ sites/509/2018/09/The-Real-Impact-of-Using-Artificial -Intelligence-in-Legal-Research-FINAL2.pdf), compares search results using Casetext's CARA A.I. and the LexisNexis legal research platform. The study indicates that attorneys using CARA A. I. completed their research projects 24.5% faster than those using LexisNexis. In addition, CARA A.I. located 21% more relevant results.
The study was carried out using research attorneys who had experience with LexisNexis and, if needed, training on Casetext. Each at torney conducted three research exercises, with some being done on Casetext and others on LexisNexis, so that an equal number were conducted with each service. In addition, the attorneys completed a survey addressing their research impressions.
Reliability and Bias
The study was released in September 2018 and was promptly challenged by LexisNexis. A law blog by Jean O'Grady published NLRG's acknowledgement that Casetext had commissioned and initially designed the study and compiled the report, raising concerns about reliability and bias. It also features a statement from Jeff Pfeifer, LexisNexis' VP of product management. He says that "actual search methods used by study participants do not appear to be in line with [LexisNexis'] user activity" and similarly challenges the study's transparency. Jake Heller, Casetext's CEO, responded by standing by the report and encouraging LexisNexis to "conduct their own study."
Through my employer, I have access to Casetext, and I coincidentally did some testing of CARA A.I. earlier this summer for unrelated reasons. I uploaded a brief from my files from an active lawsuit dealing with copyright's Fair Use doctrine. It's a legal area I'm fairly familiar with, having researched and taught it on and off throughout my career. My completely unscientific impression was that the CARAA.I. results were generally relevant, but also the usual suspects, which means they were cases that anyone experienced in that area should find familiar. Yes, they were relevant. However, an attorney doing even routine research on this topic would likely either know them already or encounter them without much difficulty. In fairness, others among my law library colleagues described stronger CARAA.I. results.
LexisNexis and the other largest platforms, Westlaw and Bloomberg Law, are all actively developing their own AI technologies and are incorporating data analytics into a number of areas. LexisNexis purchased the Lex Machina and Ravel search and analytics platforms, and Westlaw recently released its next-generation research platform, Westlaw Edge. In addition, all three platforms benefit from their ownership by large data publishers and are able to use those relationships to provide exclusive content that is unavailable to Casetext or other databases.
These ownership circumstances have led to LexisNexis being in the news for a separate and similarly troubling issue. In June 2018, the American Association of Law Libraries (AALL) complained about LexisNexis' practice of "bundling" unrelated products (aallnet .org/wp-content/uploads/2018/ 06/AALL-2018.06.07-Letter-toM.-Walsh-23527158-vl.pdf). (Disclosure: I am an AALL member.) As noted, LexisNexis is part of a large publishing company and controls a number of print and electronic treatises. Most notable are the Matthew Bender publications, which include leading treatises in areas such as copyright, patent law, bankruptcy, and federal practice. Many law firms were reporting that LexisNexis would not let them subscribe to Bender publications without obtaining a license to its flagship Lexis Advance product. AALL asserted that those practices violate AALL standards and may violate antitrust laws.
The legal database world is changing dramatically, and the flexibility that startup companies have may be a big factor in their marketplace battle with the legacy companies. Whether or not it's actually faster or actually provides more relevant results, Casetext is bringing something new to an increasingly AI-focused research marketplace.
George H. Pike is the director of the Pritzker Legal Research Center at Northwestern University School of Law. Send your comments about this column to email@example.com.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||LEGAL ISSUES|
|Author:||Pike, George H.|
|Date:||Nov 1, 2018|
|Previous Article:||New European Copyright Directive Courts Controversy.|
|Next Article:||Focus on the Customer: Aiding--And Accidentally Abetting?|