From a Crowdsourced Prediction Market to an Intelligent Litigation Assistant

November 3rd, 2011

Following up from my post about Harlan–the virtual litigation assistant of the future–here is a section I wrote for my forthcoming article on FantasySCOTUS:

Admittedly, in its present form, FantasySCOTUS 1.0 is not particularly reliable for making important legal decisions. Further, while, the eighty or so cases the Supreme Court decides each year are no doubt quite significant, and of broad interest, the 280,000 civil cases commenced in District Courts and the 57,000 appeals commenced in Courts of Appeals in 2010 affect far more people. A prediction market that can provide accurate predictions for the vast number of cases filed, and appealed, in federal courts each year could prove invaluable to lawyers and non-lawyers alike.

Building on an idea developed by Professors Kobayashi and Ribstein in Law’s Information Revolution, a future version of FantasySCOTUS could shift from using a crowdsourced model—it is not likely that enough people will be intimately familiar with the thousands of cases decided in the inferior courts—to a super cruncher model with an improved decision engine, that could analyze data from previously decided cases to offer predictions for cases not yet filed.

It would be quite conceivable for a bot to crawl through all of the filings in PACER —which stores every brief, opinion, and order filed in the federal courts, reportedly around 500 million documents —and develop a comprehensive database of all aspects of how each court works. Using sophisticated text-recognition and natural language searches, a database could automatically index all of the cases—no need for fallible research assistants to laboriously tag cases—noting the parties who filed briefs, the courts they filed in, the judges who decided the cases, what type of case it was, what were the damages or relief sought, what were the merits of the case, how long it took to decide, what types of briefs were filed, how the case was resolved, etc. This process would be instantly performed with every new filing, so the database would constantly be up to date with the latest jurisprudential and litigation trends—no need to resort to outdated data sets from the past.

With this data, a prediction engine could determine the various traits of successful and unsuccessful actions of various types, in various courts, under various circumstances. With enough data the prediction engine could provide, ex ante, a prognosis of how a case will likely proceed. Telling a client how a case will turn out—usually any client’s main concern—is something that attorneys, no matter how well qualified, can only do imprecisely. As Professor Ayres remarked, “[t]rolling through databases can reveal underlying causes that traditional experts”—even pricey, experienced lawyers—“never even considered.” If lawyers can ascertain in advance what the likely results of litigation would be, they could “avoid[] disputes altogether” and settle out of court. Even if the dispute cannot be avoided, a realistic prediction of probable damages could yield ways “to contain disagreements amicably and to avoid unnecessary escalation”

But what if the engine could tell an attorney not only what will happen, but also how it should be accomplished? Imagine a program similar to the iPhone’s Siri app. Call it Harlan. A would-be litigator could tell Harlan the relevant parties, the facts, the merits, and the remdy sought; any relevant documents would be shared. Harlan could generate a roadmap of how the case would be resolved with different judges in different courts, and perhaps even recommend an ideal forum (call it fantasy-forum-shopping). Harlan could explain how best to structure the litigation, what types of motions would be most successful, and how to arrange arguments. With advances in artificial intelligence—Google has developed cars that drive themselves, and IBM’s Watson defeated the Jeopardy world champion —it is not much of a stretch to suggest that Harlan could even draft the briefs (many sections of briefs today are copied from boilerplate anyway), or at least check the persuasiveness of the arguments against other successful arguments already accepted by courts. Harlan would also work wonders for non-lawyers. A person could download the app, talk to Harlan in plain language, explain the problem, and listen to possible remedies—that may or may not involve paying a lawyer. Harlan would improve access to justice, at little to no cost.

Such a product would transform the legal profession, and our society. This change would require a fundamental rethinking of approaches to legal education, the practice of law, and broadly speaking, our system of justice. It will likely first be first met with doubt—computers can’t replace human lawyers! This technology would not be about replacing lawyers—at least lawyers who adapt —but rather providing advocates with information and knowledge in order to serve clients more effectively at a lower cost. Next, there will be fierce resistance to change from entrenched interests in the form of ethical and regulatory challenges —computers can’t follow the rules of ethics, and they will provide ineffective assistance! These criticisms are fair, but they serve as opportunities to improve the quality of representation all around, rather than instinctively oppose any change that upsets the status quo. Reforms to the regulatory regime will come, followed by gradual acceptance of this technology. We hope that FantasySCOTUS will serve as a first step in the evolution from today’s time-consuming, customized labor-intensive legal market to tomorrow’s on-demand, commoditized law’s information revolution

I see the last paragraph as basically a roadmap for my career over the next decade or so. I’m sure at some point I, or others, will go back to this post and say something either (a) Damn, he was right; or (b) Damn he was a fool. I’m hoping for the former, but I wouldn’t be surprised if it’s the latter.