Machine Intelligence and Legal Services

The level of chatter about machine intelligence and how it can impact law practice seems to be growing.

That makes sense – what with IBM’s Big Blue wiping the floor first with chess grandmasters and then with Jeopardy champions, it seems appropriate to wonder how ever smarter machines can impact legal services.

Watson today is all grown up compared to the baby versions that trumped chess players and game show champions. Today’s Watson has moved on to less structured and more difficult chores such as developing debate arguments, and IBM’s General Counsel claims it could easily pass a bar exam.

 Machine intelligence is already transforming other knowledge industries. One example would be journalism, where some routine forms of news are now entirely created by computers from sources such as securities filings, and where human curated periodicals have been significantly supplanted by machine curated services such as Google News.

LegalZoom and its competitors such as RocketLaw are already very much on the radar screens of consumers and small firm lawyers. Documents that have been staples of small firm practice – simple wills, simple contracts, or residential and commercial leases – can be personalized and printed without a lawyer involved. Even simpler, smart phone apps such as Shake can create simple contracts such as sales agreements or confidentiality agreements, obtaining signatures online and e-mailing copies to all the parties. On the commercial side, most big cases now begin with software driven review of documents via predictive coding, not human review.

John McGinnis and Russell Pearce recently published a piece in a symposium issue of the Fordham Law Review that predicts a bullish future for machine generated legal services. They see benefits for some lawyers – exceptional lawyers will be easier to spot in the machine age, they believe, and consumers now priced out of legal services might find solutions through practitioners who harness machine intelligence. That aside, they see downsides for meat and potato lawyers as machines take on ever fatter slices of the work of document review, document and form creation, better and faster legal research, and outcome prediction based on big data rather than intuition.

More recently, Silicon Valley veteran Paul Lippe and law and data master Daniel Katz have a column in the ABA journal with ten predictions about how technology like Watson will impact the law. Like McGinnis and Pearce, they see an ever growing role for machine intelligence in law.

 One prediction they make that I hope is right is that machine intelligence will drive demands for more readily discoverable law, as opposed to the murky dogs’ porridge of statutes, cases, regulations, opinion letters and more that currently obscure legal duties. As machines do more of the grunt work of digging out rules, pressure will build to make the rules conform to rational and predictable data structures.

Another has to do with the classroom – once all law students have the equivalent of a Watson on their smart phones, the power of the Socratic professor will be undermined. I think Wikipedia already does some of that, but I think the ultimate point is a bit deeper – as it becomes trivial to discover applicable legal rules, law schools will have to rethink exactly what they can teach that is of value and how they need to teach it.

So long as regulatory barriers don’t block progress, I think legal machine intelligence will have another impact – enabling non-lawyers to play ever bigger roles in the delivery of routine legal services. Technology can become a force multiplier for lightly trained technicians, providing not just answers but a vehicle for monitoring outcomes.

MinuteClinic allows non-doctors to provide drugs and other treatment for a range of common diseases. They can do that because technology allows them to reliably diagnosis a defined range of common ailments and then proceed straight to the standard treatment. Whereas back in the day a doctor would use training and judgment to decide what ailed a patient, today a blood or urine sample plugged into a machine provides a faster and more reliable answer, enabling nurses to read out the results and hand over the correct medication. When machine intelligence allows computers, perhaps working off client entered forms, to provide the same kinds of answers that machines analyzing blood or saliva samples can in the medical settings, non-lawyers can help provide and follow through on the standard solutions at a much higher level than now possible with online form services. 

These are fundamental historical and societal changes in play, as significant in their own way as the rise of factories at the dawn of the industrial revolution. Technology now impacts the intellectual professions the way mechanized factories impacted hand crafts. The changes, even for organizations that survive and prosper, will be profound.

Comments

[Adam]

Thanks, Ray:

Two of these points caught my attention:

"One prediction they make that I hope is right is that machine intelligence will drive demands for more readily discoverable law, as opposed to the murky dogs’ porridge of statutes, cases, regulations, opinion letters and more that currently obscure legal duties. As machines do more of the grunt work of digging out rules, pressure will build to make the rules conform to rational and predictable data structures."

This strikes me as a lot more revolutionary than the authors credit, and thus much less likely to actually happen. I actually think that the "dogs' porridge" is not particularly murky most of the time, but we can assume that lawyers are need in the ten percent (say) of situations where it is. The problem is that the porridge looks the way it does because dozens of different centers of legal gravity have emerged at different times, to solve various problems, and do so in different ways. Crucially, each of these complex institutional histories makes sense to each of these institution, and merely identifying an inconsistency with another is not very likely to result in them abandoning their respective missions and modalities. Even my use of the term "institutions" is misleading, as many sources of law are themselves plural in nature, and thus intra-instuitional consistency is very difficult, to say nothing of trans-institutional inconsistencies. I agree that Watson can reveal that some problems are worse than we suspect, but rationalization across sources of law is not a purely technocratic enterprise.

"Another has to do with the classroom – once all law students have the equivalent of a Watson on their smart phones, the power of the Socratic professor will be undermined. I think Wikipedia already does some of that, but I think the ultimate point is a bit deeper – as it becomes trivial to discover applicable legal rules, law schools will have to rethink exactly what they can teach that is of value and how they need to teach it."

The authors suggest that this is akin to the introduction of calculators in math classes. In a very general way, I understand that schools have made their peace with calculators in some subjects. I could certainly imagine a Watson-like technology aiding in advanced law classes (simulations, etc.) But, I have difficulty understanding what the point would be of asking a student a question in English, watching her "Watson" the query (maybe Siri will do it directly?) and listening to her read the answer off the screen. The likelihood that this scenario will produce what is quaintly known as "learning" seems quite low. (I'm not defending the Kingsfeldian notion of Socratic learning here, assuming anyone still uses it that way.)

This does lead to your next point, about non-lawyers being able to deliver more legal services. The student in the prior example, after a year of these rather strange lectures, wouldn't be a "lawyer", but could be a "legal search professional" of some kind. That's probably useful. But, we will need to completely revisit (and adjust downward) our conception of "lawyering" for me to see the role for Watson in the legal academy suggested here.

[Pasquale Research]

1) MERS might be a good example of machines leading to "demands for more readily discoverable law, as opposed to the murky dogs’ porridge of statutes, cases, regulations, opinion letters and more that currently obscure legal duties:"
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1684729

2) Indeed, MBS's drove quite a demand for legal simplification:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2427531
http://www.nakedcapitalism.com/2010/09/floridas-kangaroo-foreclosure-courts-judges-denying-due-process-on-behalf-of-banks.html

3) I have critiqued the cited McGinnis/Pearce piece here:
http://www.concurringopinions.com/archives/2014/06/a-more-nuanced-view-of-legal-automation.html

[Ray Campbell]

Thanks, Adam, for your thoughtful comments.

I think their point on Watson is that it would shift the balance of power in the classroom, which has been based on a professor who knows the doctrine quizzing students who don't quite have it mastered. Back in the day, I lived in fear of not having the answer; I think my daughter who is currently a 1L feels much the same. If Watson ever actually gets good enough to defuse a Socratic inquisition, law school classes will have to proceed in some different format because, as you note, not much useful happens when students merely transmit the answers given by the machine. I tend to think law schools over stress teaching legal rules (I blame Langdell) and don't sufficiently stress teaching how to process and solve actual multi-dimensional problems, so I would hope a world where rules are easily found would lead to more emphasis on the parts of real lawyering that get scant attention in law schools.

On your second point, I think we are well into an era where we need to revisit our conception of what lawyering is. Part of that has to do with specialization, with lawyers today often looking more like technocratic consultants than the fabled statesmen of yore. Part of that has to do with processes of commercialization and deprofessionalization, which have made the field much more about money and much less about service than the ideal, and even, I tend to believe, than lawyers in the past. Technology is just another pressure on the idea of what it means to be a lawyer. I tend to think we are moving toward a world much like the modern hospital full of health care professionals, where we have many legal service professionals, with lawyers needing to define and justify their role in the mix.

Professor Pasquale, thanks for the links you provide. I look forward to reading them.

[anon]

"Part of that has to do with specialization, with lawyers today often looking more like technocratic consultants than the fabled statesmen of yore."

In what world does this hold true? For what percent of the people of the US? And when were lawyers primarily "statesman"?

[Marc Lauritsen]

Some more history and context - Learning Law by Teaching Machines How to Think like Lawyers - http://youtu.be/C01legdVziA?t=27m14s (short presentation at Stanford Future Law conference this spring)