Will 'life of the law' be experience ... or AI?

Robert J. Brink, BridgeTower Media Newswires

In his seminal 1897 article, “The Path of the Law,” Oliver Wendell Holmes coined one of his many memorable epigrams: “The prophecies of what the courts will do in fact, and nothing more pretentious, are what I mean by law.”

What “we call the law,” he explained, is “systemized prediction.”

Had artificial intelligence been around when Holmes propounded his predictive theory of law, would he have characterized law more as “systemized prediction” and less as “prophecy,” two terms the Supreme Judicial Court justice used interchangeably?

Systemized prediction connotes a level of reliability associated today with AI’s predictive analytics, whereas prophecy sounds more akin to the superstitions of soothsayers centuries ago.

Holmes intended his predictive theory to dispel confusion between morality and law. Legal advice is not to help bad people be good, but to counsel clients just how far they can go before getting into trouble.

For Holmes, the pragmatist, a legal duty “is nothing but a prediction that if a man does or omits certain things he will be made to suffer in this or that way by judgment of the court.”

—————

Holmes hedged

So, is an understanding of how courts decide cases based on “systemized prediction” or “prophecy”? Holmes hedged.

He made determining what advice lawyers give clients sound formulaic. Its method is the study of court decisions that are “generalized and reduced to a system”— a system, Holmes implied, that renders reliable predictions with methodic precision.

But advising clients on what courts will likely do has always involved guesswork and gut instinct as well. So, as a fail-safe to systemized analysis, Holmes compared law books to “oracles.”

“In these sibylline leaves,” he mused, “are gathered the scattered prophecies of the past on which the ax will fall.”

Holmes’ allusion to ancient Rome’s “Sibylline Books” of rhyming prophecies was a clever admission that advising clients combines learning with something similar to reading tea leaves. Both require an interpretation “of reports, of treatises, and of statutes in this county and in England extending back six hundred years.”

—————

Mastered within a reasonable time

Recognizing that no human could synthesize six centuries of law, Holmes told the bar not to be “frightened,” because every generation restates law into a “finite body of dogma which may be mastered within a reasonable time.”

Really? It’s hard enough for lawyers to keep up with daily advance sheets.

But Holmes was a workaholic and apparently a dreamer, too, reportedly saying after a long evening at the Social Law Library that people could accomplish anything if they wished “hard enough, continuously, morning, noon and night, and perhaps subconsciously while sleeping.”

—————

Deep Blue played like it had a brain

Holmes envisioned that systemized prediction would eventually supplant speculation. “Far the most important and pretty nearly the whole meaning of every new effort in legal thought,” he foresaw, “is to make these prophecies more precise, and to generalize them into a thoroughly connected system.”

Now that legal research increasingly utilizes AI, Holmes’ vision of systemized predictive precision is coming true, although in ways he could never have dreamed.

Computers can now “outthink” the most accomplished humans on some of the most complex human tasks. Perhaps the most unforgettable contest pitting machine versus man was 23 years ago when IBM’s Deep Blue dethroned Garry Kasparov, the reigning world chess champion.

Capable of considering 200 million moves per second, it seemed that Deep Blue — a computer — could think for itself as it skillfully countered Kasparov’s every move.

As Charles Krauthammer wrote in the Weekly Standard: “Deep Blue won. Brilliantly. Creatively. Humanly. It played with — forgive me — nuance and subtlety.” Krauthammer forewarned of the “terrors to come.”

—————

IT integral to law practice

Indeed, AI is wreaking a disruptive and terrifying JOLT — an apt acronym for Harvard Law School’s decade-old Journal of Law & Technology — to the practice of law.

Thirty-seven states have imposed a new duty of competence in technology since the ABA modified the Model Rules of Professional Conduct in 2012.

Malpractice claims loom for solos and small firms that eschew technology that is suddenly integral to everyday law practice, while “chief technology officers” now populate the largest firms striving to leverage IT as much as possible.

One article in JOLT reports that if the bar adopted technologies available today, lawyers’ hours would be reduced by 13 percent and that 23 percent of a lawyer’s job could soon be automated.

—————

Breakneck advances

Such estimates for reduced hours and automated jobs may not be hyperbole. Follow the money. Investment in legal technology skyrocketed to $1 billion in 2018 from only $233 million the year before.

Although the tally for 2019 is not in, as of last September another $1.2 billion had been invested in companies vying to maximize profits by automating as much of law practice as possible.

There’s an avalanche of new products based on litigation analytics designed to help attorneys advise clients on what courts and judges will likely do — exactly the “specialized prediction” Holmes had envisioned.
Algorithms can promptly analyze readily available court decisions at a granular level. Publishers reportedly all now have access to all federal court dockets, and state dockets are increasingly available for AI’s analysis.

In the blink of an eye, AI-informed lawyers can see how individual judges have typically ruled on scores of different types of motions. They can learn, for instance, how often particular judges rule on summary judgment, and then what specific precedents — and the exact language — they most often cite.

AI can promptly provide analytics on a lawyer’s or law firm’s performance before individual judges, or whether a judge statistically favors plaintiffs or defendants, as well as how often the judge is reversed or affirmed.

Jury trials? One app searches backgrounds of potential jurors, such as voter registration, criminal history, financial and real estate transactions, and social media. Its algorithm correlates behavioral and personality traits to different types of cases and then ranks whether potential jurors would likely favor plaintiffs or defendants, all purportedly within 90 seconds.

There’s a wave of AI-empowered products specifically designed to make logical, evidenced-based predictions of how judges and juries will probabilistically decide cases.

—————

Brain power cannot compete

Artificial “intelligence” is not a misnomer. The MIT Technology Review confirms that algorithms “can learn, reason and act for themselves,” as well as “make their own decisions when faced with new situations in the same way as humans... .”

The more people read, the smarter they become. The more big data AI analyzes, the smarter its statistically probable predictions become. We humans are handicapped because, however sublime, our brain power simply cannot compete with modern computing power.

Holmes idealized working morning, noon and night, as well as subconsciously while sleeping. Even so, it was inconceivable that anyone could ever assimilate 600 years of written law.

But computers never sleep. And they can do in milliseconds what might take humans months, years and even lifetimes to find, read and digest. Intel chips now process over 10 trillion calculations per second compared to Deep Blue’s paltry 200 million chess moves per second two decades ago. The contrast boggles the mind — a trillion equals a million million.

Holmes once thought that six centuries of law was too dense to decipher. Yet, his once “sibylline leaves” have been converted into massive datasets for AI to mine in minutes if not seconds, such as the 6.7 million cases spanning the history of American law digitized by Harvard Law School.

—————

Not a pretty picture for humans

By now it’s probably no surprise that when an AI algorithm was tasked with predicting the outcomes from every U.S. Supreme Court case between 1816 and 2015, not only did it correctly predict 70.2 percent of the court’s decisions but also 71.9 percent of each justice’s respective votes.

The best result among living and breathing Supreme Court scholars in previous such experiments was 66 percent, prompting the lead author to quip that “[e]very time we keep score, it hasn’t been a pretty picture for humans.”

Indeed, retrospective study after retrospective study — such as one involving a bail algorithm, the recommendations of which for over 550,000 past cases would have resulted in a whopping 25 percent fewer recidivists than the actual decisions of judges in New York City — demonstrates that machine beats man again and again.

—————

Robot judges?

Given that AI is superior to law professors and judges when predicting outcomes of past cases, some scholars think that algorithms might soon be better than humans at charting the future direction of Holmes’ “Path of the Law.”

An article in the Stanford Tech Law Review states that “the prospect of ‘robot judges’ suddenly seems plausible — even imminent.”

And a Duke Law Journal article opines that once there is “software that can create persuasive opinions, capable of regularly winning opinion-writing competitions against human judges … we should in principle accept it as a judge, even if the opinions do not stem from human judgment.”

Farfetched? Cutting-edge legal research systems already utilize natural-language algorithms that intelligently respond to search queries with AI-generated legal memos. A 2017 New York Times article quoted two people familiar with the then-novel legal-memo feature. One said that the AI-generated memo identified relevant cases and was “indistinguishable from a memo written by a lawyer.” The other deadpanned that the computer is “not much of a writer.”

At least AI still can’t turn a phrase like Holmes.

Yet, the Duke article foresees the possibility of AI Supreme Court justices if the day ever comes when AI-authored decisions are truly learned and persuasive.

—————

Human experience or computer logic

Perhaps Holmes’ most enduring epigram is that the “life of the law has not been logic, it has been experience.”

If his vision that law may one day be reduced to a system of “systemized prediction,” could it be that the life of the law will be far less human experience and much more the logic of artificial intelligence?

Since that terrifying possibility is hard to predict with confidence, we can always, as Holmes advised, consult sibylline books for the answer.

—————

Attorney Robert J. Brink is executive director of the Social Law Library.