The medical and legal professions might seem like entirely disparate disciplines. But in the words ofGeorge Triantis, JSD ’89, andLloyd Minor, MD, deans of the Stanford Law School and School of Medicine, respectively, they are like-minded service professions facing new opportunities and challenges in the era of artificial intelligence.
Both were speaking at the inaugural Future of the Legal and Medical Professions Symposium held Oct. 26 on the Stanford campus. Co-hosted by Stanford Law School, Stanford School of Medicine and the Deborah L. Rhode Center on the Legal Profession, the public event brought together physicians, lawyers, scholars and technology leaders to compare notes on the pressure points the two fields share.
“There are so many similarities between how our professions interact with society and how they’re structured,” said Minor, the Carl and Elizabeth Naumann Dean of the School of Medicine. “Both law and medicine are knowledge-based professions — meaning their practice is based on acquiring a body of knowledge, becoming fluent with that knowledge and then applying it.”
George Triantis and Lloyd Minor speak at the symposium. Photo by Steve Fisch
Whether that knowledge lends itself to the governance of society or the human body, the professions share several key areas of overlap impacted by AI: how to apply it to daily practices, how to train the next generation to use it, how to use it to expand access to health and legal services, and — a central focal point during the event — how to apply it in ways that assuages rather than exacerbates waning public trust in both fields.
“One of our obligations as leaders of schools at Stanford is to do what we can to rebuild trust in our educational programs and in our professions,” said Minor, who is also vice president for medical affairs at Stanford University.
In tandem with that reparation is preparation, said Triantis, the Richard E. Lang Professor and Dean at Stanford Law School.
In a landscape continually shaped and reshaped by AI, it’s not enough to teach students what AI is, or how to use it. The most promising uses of the technology today might be surpassed by new models tomorrow. Students need to learn how to navigate a profession impacted by ever-changing technologies.
“One of the puzzling things for many of us is to try to explain where AI is strong and where it’s weak — where the holes are,” Triantis said. “We’ve all experienced how tremendous it can be with complex tasks, and then it falls short on something extremely simple, and vice versa.”
Why AI excels at some things and fails at others is hard to describe, and it comes down to judgment, which is innately human.
“The key will be to teach adaptability and how to move from one tool to another, how to build tools and how to evaluate them, knowing when it’s important to have a human in the loop.”
“The key will be to teach adaptability and how to move from one tool to another,” Triantis said, “how to build tools and how to evaluate them, knowing when it’s important to have a human in the loop.”
Throughout the day, panelists discussed how AI might improve access to services, how it can help and hinder an incoming generation of clinicians and lawyers, what it will take to adequately train students in both professions to effectively capitalize on AI, and how crucial it is to cultivate trust — between doctors and patients, lawyers and clients, and society and the medical and legal fields at large.
Teaching critical thinking
Education was a central theme of the day as speakers raised questions about whether AI can improve the way subject matter is taught, and about how to identify and sidestep the pitfalls of incorporating AI into early training.
Dong-han Yao, MD, associate director of AI in medical education and a clinical assistant professor of emergency medicine, said medical and law students are expected to master four to five times more content than earlier generations — and it’s not necessarily a good use of time and brainpower. “I spent less time in the emergency department than I did learning about the Krebs cycle,” Yao said of his early training, describing a series of chemical reactions that supports food-to-energy conversion. AI, he posited, could support the information-retrieval side of learning and free students’ time to hone more specialized skills that require critical thinking.
Used incorrectly, however, AI can blunt our ability to apply knowledge and discern good medical or legal practice from bad. Yao described three scenarios educators are seeing when students misuse AI to expedite learning: de-skilling, which is when someone loses a skill they once had; mis-skilling, when AI steers a learner wrong; and never-skilling, in which a skill is never acquired. “How, for instance, will a student who always relies on ambient listening learn to distill a patient history? Or one who always uses an AI tool to listen to the heart learn to diagnose a murmur?” asked John Whyte, MD, CEO of the American Medical Association.
Daniel Ho, William Benjamin Scott and Luna M. Scott Professor in Law and director of the Stanford RegLab, suggested there’s an inherent teaching opportunity in the flaws of current AI models. His team analyzed legal AI tools that are marketed as free of hallucinations or AI-generated mistakes and found that the tools hallucinated 17% to 33% of the time. (More than 2,000 real legal cases have involved hallucinations, he said.)
One possible teaching exercise, suggested Ho, is to assign first-year students a case built on a false AI-built premise and allow the student to discover and analyze the model’s errors.
Even after law school matriculation, AI continues to complicate early career training for lawyers, as firms increasingly turn to the technology to handle entry-level tasks, saidNora Freeman Engstrom, Ernest W. McFarland Professor in Law and co-director of the Deborah L. Rhode Center on the Legal Profession. If technology takes over the entry-level tasks that have traditionally been training grounds for young lawyers, she said, how will they develop their instincts, hone their judgment and learn to tell a winning argument from a losing one?
“The interests of the patient and the client are first and foremost … that’s the part that shouldn’t adapt too quickly or too strongly.”
Regardless of how AI might shift the structure of early-career training, panelists agreed that educating a sharp lawyer or clinician comes down to a commitment to teach students how to be discerning and to retain cognitive agency — that is, to think for themselves, and not outsource it to AI.Jenny Martinez, Stanford University provost, professor of law and the Frederick Emmons Terman Professor, summed it up: In a world where intelligence is abundant, cultivating habits of rigorous thinking is paramount.
Boosting access, capacity
Legal and medical fields face longstanding issues when it comes to access to services, often because too few professionals are practicing, or because the services are cost prohibitive.
For instance in about 75% of U.S. civil cases at least one side lacks a lawyer. She calls it the two-door problem: People facing legal challenges can either hire expensive counsel or represent themselves. There aren’t any other options. That’s because there are restrictive rules in nearly every state that make it a crime to practice law or provide legal assistance without a law license
Medicine, however, has a larger network of care professionals — patients can be treated by nurse practitioners or physician assistants instead of MDs — and there are debates if such a model might be feasible for the legal field.
While AI, conceptually, may be a resource to expand professional capacity for medicine and law, full AI doctors or AI lawyers are still just an idea. But that doesn’t mean AI can’t be a boon to capacity.David Freeman Engstrom, LSVF Professor and co-director of the Rhode Center, discussed one effort to use AI to streamline a work stream too time-consuming to be effectively handled only by humans.
In debt collection cases in the Los Angeles County Superior Court, he said, nearly half of default judgments (rulings issued when a defendant doesn’t show up in court) contain errors, and court staffers can’t keep up with reviewing all of them. His team built an AI assistant that flags likely errors for attorneys to review. It proved faster and more accurate in testing with law students, and a randomized controlled trial is underway. He framed the tool as a way to make the court “fairer and more accurate,” not just more efficient.
Trust: Earned in drops, lost in buckets
In a conversation about the state of public trust, California Supreme Court Justice Goodwin Liu and former White House science and technology adviser Arati Prabhakar described a decades-long erosion of confidence in both professions, exacerbated by today’s political and technological pressures. (Only 35% of Americans expressed confidence in the justice system in a poll taken in December 2024.)
During a symposium session focused on resilience during political change,David Studdert, professor of law, vice provost and dean of research at Stanford University, said professions must adapt to the world around them, but not everything should bend.
“The interests of the patient and the client are first and foremost … that’s the part that shouldn’t adapt too quickly or too strongly,” Studdert said. “I’m a bit of an optimist and I have to believe that if the average professional in one of these fields performs well and honestly for the average patient or client, that gives significant insulation against the turbulence of political times.”
Hypothetically, Liu said, AI could help build confidence in the justice system by speeding up a slow court. But AI can only expedite menial tasks for now, and its accuracy and judgment are still woefully lacking. “I can persuade my model to give me just about anything I want because it is designed to please me,” Liu said, adding that he’s wary of outsourcing independent judgment to a product built by a company that wants customers.
Prabhakar said a bigger problem is that the science of validating AI isn’t up to snuff.
“We do not have a scientific and technological foundation for the kind of thorough evaluation and validation of these models and their applications,” she said, stressing that more research is needed.
Both speakers urged trainees to treat listening and communication — between doctors and patients, lawyers and clients, institutions and the public — as the core to building trust. Vivian Lee, MD, PhD, an executive fellow at Harvard Business School, described trust as multidimensional but said that by and large, “people want to feel like you understand them and that you care about them. The ability to express some level of empathy with people who are different from you is something that we really try to build into our medical education,” she said.
Still, AI can play an important role for some patients.Alyce Adams, PhD, the Stanford Medicine Innovation Professor and a professor of health policy and of epidemiology, said her team works with patients who have diabetic neuropathy, a condition caused by damage of nerves outside of the brain. There is no cure, it comes with highly variable symptoms and there’s often stigma attached, she said.
“Patients again and again have told us that, in that situation, they would rather tell the AI agent about their symptoms than tell their doctor,” Adams said. She was clear that the doctor must still be part of the picture, adding that the doctor should remain engaged and be the final decision maker.
Still, the question remains: Can AI help build back public trust? According to Liu, the California Supreme Court justice, we’ll know in time: “The jury’s still out.”
