PARIS — On the left-hand side of the blue screen, court decisions appear at the speed of light. Some words are highlighted. Simultaneously, on the right-hand side, a map of France takes shape. Graphs form. "So, you see," Louis Garrett-Chahine starts, "if you've been fired for slander and if you want your dismissal to be reconsidered as being without real and serious grounds, it's best to say that you were a little tipsy, and it's best that your company is based in Brittany ... Ha! but careful, the trend is headed downwards — did you see that?"
Yes, we saw that. Moments before, Garrett-Chahine, the young co-founder of Predictice, entered the terms "dismissal," "after 2013," and "slander" into the platform's search engine. Immediately the powerful algorithm began scanning more than 1.5 million court decisions to analyze and present the chances of success and estimate a likely compensation. All it took was a few seconds. "Soon we'll also be able to identify the best arguments to put forward, allowing lawyers to adapt their strategy," he claims.
These kinds of scans are one of the latest technological innovations of Law 2.0, what "legal geeks" have named "predictive justice" — algorithms that, in record time, can analyze a huge mass of legal precedents to anticipate the outcome of a case, choose the most relevant arguments, or assess the potential amount of compensation. These tools can't turn lawyers and judges into fortune tellers. But they can assist the decision-making process.
Not surprisingly, they're drawing the interest of more and more legal experts and companies. "We'll save time," admits Benjamin Pitcho, an attorney in Paris. "It's an opportunity for the trade." But he also worries about the "ethical challenges" being raised, and about what kinds of limits should be placed on predictive justice.
Saving time and energy
Until recently, this kind of information was off limits, in practice if not in theory. But recent legislation in France makes it possible for the public to have free access to all court decisions, which must be made anonymous beforehand. A gargantuan task considering that French civil and commercial courts reach more than 2.5 million verdicts given every year.
Most experts reckon it will take at least five years to put everything into the online database. Predictive justice startups are champing at the bit. For the French market is divided up between five of them: Predictice, Case Law Analytics, Doctrine.fr (for general law), Supra Legem (for administrative law) and Tyr Legal (for labor law). But the field is still very much in its infancy, and SupraLegem is the only one that is open-source.
A very clear operational gain.
Some of these startups are still in beta version being co-developed with lawyers or companies. Predictice, for instance, has already signed partnership deals with legal offices such as Taylor Wessing, but also with the telecommunications group Orange and the legal protection branch of insurance company AXA.
"The use of predictive justice platforms enable us to strengthen and fine-tune our counseling and mediation strategy depending on our client, who will thus be in a position to make the best possible decision in an objective way," says Jean Manuel Caparros, a member of AXA's legal protection team.
It's a promising market, as demonstrated by the 2 million euros raised by the one-year-old Doctrine.fr, which aims to become no less than the "Google for law." The platform claims to be working from a database of 3.5 million court decisions. "We've taught bots to look for court decisions online," explains co-founder Nicolas Busmante. More than 7,000 lawyers are said to be using this platform, which lets them, among other things, assess how long a procedure might take and the risks involved. For now, it's also the only one that displays its subscription price: 159 euros per month.
The legal arm of SNCF Transilien — the train company in the Paris region — sees it as "a very clear operational gain."
Predictive justice makes the myth of one impartial and blind justice collapse.
"When they have to work on a case, jurists or advisers no longer have the excuse of needing time for research, which used to be enormous," says Alexandre Mornay, SNCF Transilien's legal director. Stéphane Dhonte, an attorney in Lille, agrees. "The lawyer's new role, combined with the diversion program, will change how we work. There's no point in fighting against the obvious tide."
But some judges are concerned about the development, as evidenced by the fact that two of France's leading magistrates dedicated their new year's solemn addresses to artificial intelligence in the legal system. "The 21st century must prepare for a new revolution: open data," Bertrand Louvel, first president of the Court of Cassation (the highest court in the French judiciary), warned.
The stakes are high, for one thing because predictive justice makes "the myth of one impartial and blind justice, of judges as the 'mouth of law,' collapse by revealing its caprices," says Antoine Garapon, a magistrate and secretary general of the Institut des Hautes Études sur la Justice (Institute of Advanced Studies on Justice). Some experts also warn that lawyers might find it more difficult to impose innovative arguments, or arguments that go against an entrenched precedent.
There's also a real danger of a performative effect, a standardization of legal practices that could be seriously prejudicial to the people. One extreme example of this danger is the system called "evidence-based sentencing" used in some American states, and which is based on an algorithm to assess the duration of a sentence so as to minimize risks of recidivism. "A few months ago, however, the Obama administration released a report to calm things down because the results were too discriminatory," says Florence G'sell, Professor of Law at the University of Lorraine.
The experiment looks chillingly like the science-fiction movie Minority Report, in which the future society is imagined as one that would have eradicated murder by adopting a prevention system hidden inside the Department of Justice, whose goal is to detect warning signs of homicidal violence and to arrest "pre-culprits."
We're not there yet. For now, French predictive justice platforms have deliberately chosen not to develop algorithms on criminal court decisions, and a white paper on good practice regarding that issue is said to be on the drawing board.