The WSJ reports:
In a nationally televised competition, the Watson computer system built by International Business Machines Corp. handily defeated two former “Jeopardy” champions. * * *
To emulate the human mind, and make it competitive on the TV quiz show, Watson was stuffed with millions of documents—including dictionaries, anthologies and the World Book Encyclopedia. After reading a clue, Watson mines the database, poring over 200 million pages of content in less than three seconds. Researchers developed algorithms to measure Watson’s level of confidence in an answer in order to decide whether it should hit the “Jeopardy!” buzzer.
The article notes that one commercial plan for Watson targets the health care industry. Well, what about law?
In my recently posted working paper, Law’s Information Revolution, Bruce Kobayashi and I discuss how developments like this could fundamentally change “law practice” into an information-based industry where law-based information is sold through product and capital markets. For example, we note (footnotes omitted):
There is room for more radical developments in using computers to create legal knowledge. This could involve reengineering the underlying idea of what legal research entails. Instead of the conventional method of relying on courts’ holdings categorized in treatises or “tagged” via West Key Numbers, lawyers might analyze facts in extensive databases of cases or court records available through PACER (Public Access to Court Electronic Records) to predict case results. These predictions might be refined using theories based on economic analysis, psychology, sociology, decision theory and political science to determine relevant variables. Lawyers might collaborate with computer scientists to develop new computer prediction algorithms. This would be analogous to the techniques already used to predict consumers’ tastes in films and music. Computers already can provide the correct Jeopardy question “Who is Eddie Albert Camus” for the answer “A ‘Green Acres’ star goes existential (& French) as the author of ‘The Fall.’” They ought to be able to answer a question like “can a lawyer copyright a complaint?”
I will be discussing the implications of these developments for law teaching at an Iowa symposium next week, and posting my paper on that shortly.
Of course computers won’t replace humans anytime soon. Watson’s creator conceded that “A computer doesn’t know what it means to be human.”
Yes, but do lawyers know that?
RSS - Posts
I think that the the things that Prof. Ribstein talks about, and those that Mr. Levene talks about in comment number 4, are way too hard for any foreseeable AI type program, they are too close to core human cognition too even begin to describe the program field.
OTOH, those functions are maybe ten percent of a lawyers time. Much of the rest is just donkey work and fill in the blanks. Functions like preparing an estate plan for a middle class person, or a retail lease for a small landlord, ought to be much more automated than they have been. A lot of the document work performed by small firms and solos will be replaced by computers.
Heck, lots of big law firms don’t even have standard forms of basic agreements. They have had no incentive to spend the effort. Better to send an associate to comb through the files looking for the last deal somewhat like this one. As clients get more resistant to paying for those snipe hunts, there will be changes.
One other point. In my experience, the very best lawyers – both litigators and corporate lawyers – are excellent judges of human character. They can read people, understand their motivations, fears, and pressure points, and know how to appeal to and manipulate them. I suppose it’s conceivable that someone might someday create a computer algorithm that can duplicate that skill, but we are currently far from that point.
Computer algorithms might well be able to provide low-priced legal “advice” for simple legal questions. In fact, they already do – you can go online and get a will or a tax return, services that used to be provided only by highly trained professionals. For anything that’s not cookie cutter, I think it’s more likely that we’ll end up with something like Walmart providing low cost legal services to the middle class, as now appears to be happening in England. However, computer algorithms won’t ever replace Marty Lipton telling you that a 5% pill is OK at a time when the courts have only upheld 15% pills.
If computers are ever able to provide reliable answers to legal questions, why stop at replacing lawyers? The real application for such a computer is as a judge.