Dan Anderson was my best friend growing up. He now lives in Burlington, Vt., and is a medical doctor. Two years ago, he decided to leave his practice and go to work for a software company named PKC. I wasn’t surprised. I remember Dan building logic circuits in his basement about the same time Steve Jobs was building the first Apple computers in his parents’ garage.
PKC designs and sells medical decision support software. The PKC system consists of two basic components: a knowledge base of medical information developed during many years and continuously expanded, and algorithms that recognize patterns within the knowledge base. When a doctor sees a patient, the system starts off by asking basic questions about the patient’s symptoms, such as, “Where does it hurt?” Based on the answers, the system intelligently comes up with further questions according to relevancy.
When the questionnaire is completed, the system takes the patient’s information, searches for patterns in the knowledge base, and returns a differential diagnosis describing the patient’s most likely conditions. The system can also lay out a course of treatment based not just on medical necessity but other factors like cost effectiveness. The key to the system providing accurate diagnoses is the quality of the knowledge base, which comes from medical journals and from organizations like the National Institutes of Health.
Given that one of us is a doctor and the other a lawyer, we naturally take shots at each other’s profession. It’s not hard to see how Dan’s work at PKC provides me with great fodder for doing this: “Doctor on the golf course while receptionist is diagnosing patients”; “Outsourcing the kids’ annual checkup to a call center in Kazakhstan.”
In response, Dan says a couple of things. A decision support system never gets tired and neglects to ask questions. No one doctor can ever read enough medical journals or attend enough conferences to approach anything close to the amount of medical knowledge contained in a computer database. A decision support system can see an infinite number of patients in a day without scheduling conflicts and the technology can provide diagnoses and treatment plans that may never have occurred to a doctor.
When you boil down any profession, it becomes a matter of making informed decisions. We in-house lawyers aren’t that much different from physicians in an HMO. People come to us and present us with problems or business opportunities. We ask questions, identify legal issues and address them. While we don’t write prescriptions, schedule tests or perform surgery, we give advice, create documents and prepare for litigation.
The idea of using artificial intelligence in the practice of law isn’t a new one. Applying artificial intelligence to lawyering can be traced as far back as a 1977 Harvard Law Review article by L. Thorne McCarty entitled “Reflections on Taxman: An Experiment in Artificial Intelligence and Legal Reasoning.” Currently, there exists an International Association for Artificial Intelligence and Law. But, according to Dan, the support technologies used by lawyers—mostly e-discovery software and document assembly engines—are the simplest kinds of decision-making technologies and don’t harness the real promise of artificial intelligence.
My biggest challenge as an in-house lawyer is responding to internal client requests and maintaining client satisfaction. Invariably, my clients will contact me as if I am sitting at my desk, twiddling my thumbs. Even if I do have time, they don’t appreciate that sometimes responding requires many hours of research and review.
The industry I work in, health care information technology, is the subject of ever-expanding regulation, and I live in constant fear that I have not kept up. At any given time, I am working with clients located in all four U.S. time zones and potentially in all four corners of the world.
Like all human beings, I have my limitations. I am a morning person. I only speak English and a little Spanish. I tend to get impatient. I am biased towards people that are cooperative and appreciative. I have found that the cardinal sin in my job is to be inconsistent, which is easy to do.
Technology systems, on the other hand, will always produce the same output given the same input. Systems don’t sleep and don’t get tired. They are unbiased and infinitely patient.
The process of analyzing legal issues, applying rules and reaching conclusions (IRAC), while it certainly requires professional training and knowledge, is not as complex as diagnosing diseases in the human body. So why hasn’t decision support technology been put to work in the legal domain?
I predict we will begin to see modest forms of decision support systems make their way into corporate legal departments in the next five to 10 years. Decision support technology will initially be used for responding to everyday requests for legal support, such as handling and disclosing confidential information, managing IP portfolios and complying with policies. The systems will be self-service.
In the future, I see the technology helping lawyers with more complex decision making, such as identifying and analyzing compliance and liability issues in complex transactions and in company systems and operations.
While large sophisticated databases already exist in the legal profession and metadata can easily be extracted from them to build decision support knowledge bases, in-house lawyers relying on these systems to tell them what to do is trickier.
For its product, PKC only incorporates treatment plans for which consensus is firmly established within the medical community, such as management of heart failure approved by the Heart Failure Society of America.
Similarly, in guiding the lawyer’s response, a legal decision support system would need to incorporate published guidelines and best practices from authoritative sources such as the American Bar Association, the American Law Institute and law reviews in the United States. Internationally, sources might include the United Nations Commission on International Trade Law (UNCITRAL) and the European Commission.
Will computers ever replace lawyers? Dan agrees that artificial intelligence will never account for the strategy and the intuition that goes into lawyering. IBM Watson might be good at “Jeopardy!”, but it can never represent a client at the negotiation table.
If you are in-house counsel in your early 30s or younger, I would suggest you seriously begin to familiarize yourself with decision support products on the market and understand how these technologies can be used in the legal department. The general counsels of tomorrow will need to have legal technology skills to effectively serve their companies. Good legal department management will include optimally integrating and balancing use of human lawyers with legal decision support technology.