By Jennifer Mendez (Follow us on LinkedIn)
In the May/June 2022 issue of AALL Spectrum, I had the honor of co-authoring an article titled, Innovation Is Changing the Role of Law Librarians—And They’re Ready for It. In that article, my coauthors, G. Patrick Flanagan and Jill L. Kilgore, and I argued that, “Because law librarians are content specialists, they have a thorough understanding of all the data available via research platforms. They understand the coverage, reliability, and currency of the data offered by the myriad of research platforms and can make recommendations based on the question or use case” (emphasis my own). With the introduction of ChatGPT and GPT-4 powered solutions, we need librarians, now more than ever, to help users “understand the coverage, reliability, and currency of the data.”
Law librarians would be among the first to tell users that one of the biggest challenges posed by ChatGPT, especially in the legal research context, is that it has limited knowledge and is not connected to a live repository of current data. In fact, it does not have data beyond 2021. Like the technology field, (nearly) two years in legal years is an eternity. In that time, plenty of new legislation has been introduced and passed, new precedents have been established, and completely new legal issues have arisen (i.e. how many firms do you know that had lawyers dedicated to COVID-19, blockchain, cryptocurrency, NFTs, or A.I. five years ago?). The law is always changing and tools relied upon by attorneys need to be current and accurate; attorneys want to be sure they are citing to the latest and greatest laws, statutes, regulations, etc., which brings up an even bigger problem with ChatGPT: hallucinations. When prompted, ChatGPT will provide citations. However, the citations provided may not be where the model actually pulled the information for its response, and, more importantly, it may not refer to real or accurate sources. ChatGPT can and will hallucinate citations. Who better to verify citations than law librarians?
That’s not to say that GPT-4 powered technologies are not reliable. Casetext’s CoCounsel, for instance, has been trained on legal projects, and also leverages its up-to-date collection of caselaw, statutes, and secondary sources to respond to questions, and it even includes hyperlinked citations to those sources. And law librarians are among the best-suited to assist attorneys in deploying such responsibly-trained GPT-4 powered resources, which can help streamline various tasks, such as legal research, document review, and contract drafting. In the same breath, they’d also be able to recommend users not rely on certain A.I.-powered tools for skills they do not sufficiently possess (for now), such as legislative tracking.
Whether it was the introduction of electronic research databases, eBooks, or data analytics tools, law librarians have a history of embracing change and helping their users along in that journey. The overarching message my coauthors and I were hoping to convey back in 2022, still holds true in our post-ChatGPT world: As AI and other innovative technologies continue to disrupt the practice of law, law librarians will continue to be at the forefront of adoption, training, and increasing productivity. While GPT-4 and other artificial intelligence tools are transforming the practice of law, law librarians still have a valuable, exciting role to play.
Notes Between Us (NBU) is a blog about conversations and topics of interest to the writers. The writers are expressing their personal opinions solely. The essays represent their personal beliefs and not that of their workplaces or any organization they are associated with.