By Dr. Ken Broda Bahm:
Chances are good you’ve heard about “ChatGPT” in the past week or two. Just speaking for myself, my newsfeed has included many articles, including some with ominous titles like, “Will ChatGPT Make Lawyers Obsolete? (Hint, Be Afraid).” If the topic is new to you, a quick definition is that ChatGPT is an Artificial Intelligence assistant that was made open and free for the public at the end of this past November, and it is way beyond Alexa, Siri, and the online “help” bots you’ve frustratingly interacted with in the past. ChatGPT provides an open service where you can make just about any request (e.g., “write an essay on the French Revolution,” or “summarize the legal precedent on statute of limitations in Massachusetts”) at the bot responds with a specific, organized, cogent, and in most cases accurate answer. The “GPT” part of the name stands for “Generative Pre-trained Transformer,” which breaks down what it does: It was “trained” on a massive volume of online text, and is able to generatively apply more than 175 billion parameters in order to generate clear and detailed human-like text that answers a question or a “prompt.” And, of course, as in all dystopian scenarios, it learns and gets better the more it is used. The “Turing Test” has been the long-term gold standard on whether we’ve reached the point that technology could deceive a human into thinking that it is also human, and reports at this stage is that ChatGPT has passed.
Name notwithstanding, the uniqueness of ChatGPT isn’t its ability to “chat,” but rather its ability to generate detailed answers applying the full spectrum of knowledge it can access. Based on my limited prompts, this seems to be true Artificial Intelligence that is able to respond to just about any query with a researched, logical, structured, and coherently written answers. In fact, it has already written published articles focusing on its own likely effects on the field of law. One article, “The Implications of OpenAI’s Assistant for Legal Services and Society” (Open AI’s Assistant & Perlman, 2022) was written entirely by ChatGPT, other than the headings, prompts, and the abstract (which were added by the second author, a human, Andrew Perlman who is Dean and Professor of law at Suffolk University Law School). That AI system will also be serving as the first “robot lawyer” who will actually represent a defendant in court next month. In short, there seems to be a very high chance that AI is quickly ushering in a change that could be every bit as profound as was the popularization of the Internet in the early 90’s — perhaps even more so, because we won’t just have connected networks of human information, we will have a robust and arguably creative source of information that is, literally speaking, super-human. I know, it brings up some existential questions about where our species is headed, but the more immediate practical question is what it will do to a field uniquely grounded in analysis and the creation of advice, the law. In the bot’s own words (yes, that’s a little unnerving), it “has the potential to revolutionize the way legal work is done, from legal research and document generation to providing general legal information to the public.” In this post, I’ll look at a few implications of this.
Changes to Expect in Law
Legal Research
Of course, algorithms have already been used for some time in e-discovery, but with the further development of true AI tools like ChatGPT, it is possible that we will be able to very quickly analyze a huge volume of past cases and provide an actual analysis — not just a summary or a count — but written conclusions that come really close (and likely closer with time) to what we value in a human legal researcher. As AI is able to complete these tasks in vastly shorter periods of time, my prediction is that it will add to the current pressure to re-think how we value work and move away from basing it on the amount of time it takes (what happens when the “billable hour” becomes milliseconds?) and toward the value that the end result has for a given client.
Legal Analysis and Advice
We’ve long comforted ourselves with the idea that judgment is that last realm that machines will never effectively take over. But based on my own limited experiments, the AI seems able to analyze quite effectively. Plus, it helps to remember that at just about everything outside of the top levels of litigation, the public’s needs for legal help are based on a known and predictable set of very common problems (debt, contracts, wills, transferring wealth, taxes, etc.), and ChatGPT seems already able generate specific answers to many of those prompts with ease. And it is likely to become much more effective over time. If the content now gives you the sense that you’re talking with a creative and very well-informed high school student, we can expect more after ChatGPT, in effect, goes to law school, and then absorbs the entirety of the law and court opinions. Already in the article that the AI wrote with Andrew Perlman’s prompts, the ChatGPT created a contract for sale of property, drafted a last will and testament, and wrote a legal complaint for a car accident injury, all of which are included in the article. With a little more nuance, the bot also provided advice on ways to use the law to encourage a school district to create an Individualized Education Program (IEP) for one’s child.
Legal Writing and other Advocacy
The big apparent step forward for ChatGPT is that it doesn’t just find things on the internet, it writes in a way that is clear, readable, grammatically correct, and yes, “human.” Given that a great deal of billable time is focused on creating drafts of letters, reports, motions and filings, I could see ChatGPT being used at all those “first draft” stages. In the article, for example, the bot drafts a brief on why the US Supreme Court should not overturn the Obergefell decision on same-sex marriage. The result is basic, but does cover the surface arguments, and it is easy to see it becoming more specific in response to more detailed prompts. Now some might say that this ghostwriting isn’t actually new, with associates and paralegals already doing a lot of drafting, and the need for senior counsel to review and sign off remaining the same. But as attorneys and other legal actors augment and integrate their own writing with AI-generated content, I could see a more general blurring of what “authorship” comes to mean in this context. And while we naturally think of ChatGPT being used in legal writing, it extends to the verbal parts of advocacy as well. In the article, the bot generated a pretty good list in response to a prompt to “list deposition questions for a plaintiff in a routine motor vehicle accident case in Massachusetts.” I also asked the tool to “write an opening statement for a plaintiff in a personal injury case,” and the result was short and simple (befitting the prompt) but from a clarity standpoint, it wasn’t bad.
Articles About Litigation
Now, I want to be clear that I enjoy the human act of writing. I have written this blog for about 12 years, at most points producing two posts a week. Over all those posts with my byline, it has been all me: no associate, research assistant, or marketing person has ever written a word of it. But when I heard of ChatGPT, I couldn’t resist, and typed, “Write a blog post about legal persuasion.” No, the result wasn’t the post you’re reading, but it was a roughly 300 word essay that was clear and organized, covering the need to analyze one’s audience, combine an understanding of facts and law, and integrate those with emotional appeals that resonate with a judge or jury. Again, it was basic, but it was also pretty comprehensive, clear, and accurate. A sample:
Legal persuasion is an art that requires a combination of logical reasoning, emotional appeals, and an understanding of the law. By understanding your audience, using logical reasoning, and being well-prepared, you can increase your chances of winning your case.
Again, the bot created that sentence, it didn’t find it on the internet. Not bad, but for now I will stick with the feeling that I don’t think I’ll be replaced anytime soon. But I do think that a lot of the writing that gets churned out for internet consumption is going to start being created by ChatGPT. I’d wager a fair amount is already. And given the way AI builds its knowledge base, the truly grim possibility is that AI will become both the main producer and the main consumer of online content, with humans increasingly sitting at the edge of that circle.
The law is, of course, only one realm that will be affected by ChatGPT. Teachers are already scrambling to either step up their prohibitions on plagiarism or to find ways to integrate the technologies in ways that retain a space for human invention. Journals are also updating their editorial policies to clarify what “authorship” means. While a lot of this commentary has a tone of “let’s work with this…” there is no mistaking the amount of disruption that we are likely to face as AI becomes more available and more comprehensive.
In the field of the law, of course, there are unique disruptions. As the bot writes in the article, “One challenge of using ChatGPT in the legal field is ensuring that it produces accurate and reliable results. Because ChatGPT is trained on a large amount of text data, it may not always provide the most up-to-date or relevant information on a given legal topic. This can lead to potential errors or misunderstandings, which could have serious consequences in a legal context.” It continues, acknowledging that it is a tool that “may not always be able to account for the nuances and complexities of the law,” and adding, “It may not have the same level of understanding and judgment as a human lawyer when it comes to interpreting legal principles and precedent.” Naturally, there are both ethical and regulatory considerations, and our systems will be hard pressed to address those in a timely fashion.
More broadly, is this ushering in a apocalyptic future that can only end with humans battling the Terminators sent by SkyNet? Can’t rule it out. But what is more sure is that, within the next couple of years if not sooner, it will add some fundamental changes to the ways that any work that rests on words and knowledge is done, including law.
____________________
Other Posts on Legal Writing:
- For Immediacy, Use Active Voice (but for Abstraction, Passive Voice Can Be Used)
- Avoid Hyperbole
- Cite Social Science to the Court
____________________
Open AI’s Assistant, Perlman, A. M. (2022). The Implications of OpenAI’s Assistant for Legal Services and Society. Available at SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4294197