Evaluating AI Risk -- A Legal Perspective
Does relying on technology make our interactions less human?
image by Dall-e with prompt two humanoid robots on a date
The Atlantic published an article last week with the breathless title “The Big AI Risk Not Enough People are Seeing.” Ever curious, I poured over the article by Tyler Austin Harper. There are a number of interesting points that I agree with.
First, Harper focuses on the impacts of AI on human interactions, using online dating as a focus point. He describes Bumble cofounder Whitney Wolfe Herd’s vision of AI “dating concierges” who “date” other people’s concierges and find the perfect romantic match. Harper’s concern with this approach is the dehumanizing consequences of outsourcing fundamental human interactions to artificial intelligence.
Let’s view this point in the context of our profession. We should be wary of AI that seeks to replace or overly augment the human interactions of the legal profession. The core premise of the legal profession and the underlying point of English Common Law is that we serve the interests of ordinary people in the face of complex laws, rules, and practices. While AI Positivism argues for an LLM-powered Super Lawyer, the risk is ending up with an LLM powered Inhuman Lawyer.
Consider an example from customer service uses of AI, and extended from Wolfe Herd’s vision: An AI powered “Legal Client Services” agent that handles new client intake. Possible with today’s technology, your firm could send new clients to an AI portal where they describe their legal needs and concerns. The AI system could do a quick conflicts check, and then evaluate legal claims with the AI providing analysis, options, and estimated costs. Together, the AI system and the client could discuss how to proceed. We will soon see variations of this across all customer-facing industries, especially those with heavy call center operations.
The downside risk of such a system in legal practice is the dehumanizing aspects of such an interaction. Allow me to appeal to Hollywood to explain. Recall the example of Erin Brockovich, where a compassionate legal assistant went against the her firm’s expectations and industry norms and fiercely advocated for a group of underserved, overlooked people. Now, presume that firm had an AI-powered first interaction where the AI couldn’t find sufficient precedent, compassion, or motivation to recommend representation. Imagine the impression of the client having to explain their pain and suffering to a machine. Would the AI system be able to appropriately consider the emotional side of the claim? Would the AI system be able to offer an understanding ear so that the potential client feels heard and understood? Would the AI system be a good judge of whether the client will make a good witness or not?
The Atlantic’s article moves on to an even more serious concern by citing the Austrian philosopher Ivan Illich’s 1977 thesis in a book chapter he wrote on “Disabling Professions” where technology leads to “a disabling standardization and professionalization of everyday life.” We clearly see this in our profession. We are increasingly specialized in the practice of law, with the generalist “country lawyer” becoming evermore rare. This transformation from generalist to specialist empowers increased complexity, which then increases the formulaic or algorithmic identification of “best practices”. The concern I share with Harper is twofold: First, that AI-dependent lawyers will be reduced to rehashing common algorithms learned by AI. Second, that AI assistants will diminish the value proposition of our profession.
As I have explained in prior articles, AI is trained by reading a vast corpus of material, including case law, statues, and legal texts. Over the course of all this reading, AI models have created a conceptual understanding of legal principles and can cite (with varying degrees of accuracy) cases and precedents. The downside of this approach is that it is entirely trained on precedent. The ability for AI to compose a novel approach is limited. Current AI models have been compared to “a precocious adolescent” in terms of output quality.1 We should be wary of over-reliance on AI products to craft our work. The individual consequences include the possibility that AI “dumbs down” our work, robbing us of the differentiating qualities that define us from the next lawyer. Additionally, the premise that AI can articulate a legal thesis is certainly welcome news to those inclined to self-lawyer, including the inevitable use of AI as “co-counsel” to pro se litigants. The balance our profession needs is AI that educates, explains, and summarizes without diminishing the value of legal counsel that is a combination of compassion, creativity, and expertise.
I encourage you to read “The Big AI Risk Not Enough People Are Seeing” and consider the broad societal implications and the specific ramifications of AI on your profession. My focus in my practice and in this newsletter is to help you navigate the many advantages and challenges of Artificial Intelligence.
Here are my recommendations:
1- Avoid the use of AI-powered interactions in situations that require sensitivity to the needs and emotions of your clients. Be very judicious in your use of all AI that is client-facing. I would go so far as to argue that chatbot style technology (for example where you have a typed conversation with an obvious “robot” when you need help or need to make a claim online) has no place in the practice of law. From all aspects—technical, ethical, compassionate, and financial—such technology will alienate and disassociate you from your clients, and likely offend them as well. Another name for a lawyer is counselor. The practice of law involves being a counselor to our clients. Along with advocating for them, we must educate them and counsel with them. We lose the benefit of acting in that role if we are attempting to counsel through an AI system.
2- Use AI to summarize, compare, and critique your sources and your work, but avoid using AI to craft the approach or thesis of your cases. This is true for lawyers, and is especially true for judges as they deliberate on the merits of cases before them. Perhaps lawyers and judges can brainstorm legal theories with AI, ask AI to evaluate the flow of their legal briefs or decisions, and ask AI if our arguments sound complete and effective. But there is some loss to our professional credibility if AI is actually deciding cases in lieu of a human judge, or is the only entity crafting legal arguments in lieu of a creative, thinking, human lawyer.
As I have shared here, here, and here, AI is great at drafting general language of agreements and forms precisely because this type of document is rooted in standardization. Very few lease agreements need to introduce innovative legal concepts. While using AI systems to draft legal documents such as motions, responses, and demand letters, you should be guiding the AI tool, and not the other way around.
In other words, focus your AI adoption on automating the standard, consistent work that depends on alignment to standards and formality.
3- Use AI to empower you and your firm’s differentiating features. You can train GPTs that use your prior work as a reference for future work. Use interactive tools like ChatGPT and Claude to respond to and critique your work, but ensure that you—not your AI—is the primary author of your work product. This is important for several reasons. First, you can think through your arguments, style, and the facts of the case as you write. This often helps to refine our thoughts, and makes certain connections in our thinking. Second, you are much less likely to sound stiff and robotic than an AI tool.
4- Tolerate (what other choice do we have?) our clients’ use of AI to understand complex legal topics. We have behind us two plus decades of clients sharing their Google search results; now we will have clients asking us to carry forward their AI-generated understanding of their legal issue. Being able to navigate the increased sophistication of arguments created by clients using AI, or being able to explain why the arguments or issues put forward by AI tools are not in the best interest of the client, will be a new skill for many of us, but learn this skill we must so that we maintain our clients’ trust and respect.
5- Stay current with the evolution of case law, regulations, and technology related to these topics. That is the focus of this newsletter, so you are off to a good start. Expect that legislative bodies, legal decisions and perhaps even specialized practice sections of various bars will evolve the practice and limitations of AI use.
I am a zealous advocate for lawyers and legal professionals using AI in their practice. It’s what I have currently devoted my career to because I am confident that the use of AI can make our professional lives easier, more efficient, and perhaps may legal services more accessible. I am equally fiercely devoted to the concept that AI should strengthen us, assist us, and be our tool, rather than the other way around.
https://thecripplegate.com/ai-is-here-the-brooding-adolescence-of-artificial-intelligence/