Predictive justice, used in the United States to calculate, thanks to big data, the chances of winning a lawsuit, lands in France. Legal publishers are in ambush, and a few start-ups. By Thierry Kirat, Paris Dauphine University – PSL
Big data is making a significant entry into the world of law: witnesses, the growing number of press articles, colloquia and commissions devoted to “predictive justice”; the development of start-ups, called “legaltechs”; the diversification of the legal publishers towards the “encrypted jurisprudence”, as many evolutions which raise legitimate questions. Do the artificial intelligence algorithms that treat these mega-data substitute intelligent machine justice for the justice of men?
The attention of lawyers and analysts focuses on “predictive justice”. What is it about ? Computer systems, calculation algorithms, with a great power of extraction of data on the web, which tend to establish an estimate of chances to win a trial, to succeed in a legal proceeding, to obtain a certain monetary amount (damages, alimony …), in various fields: divorce, dismissal, neighborhood troubles, etc. While in the United States, applications exist in criminal matters to predict the risk of re-offending, the field of French law is restricted to civil or administrative justice.
Start-ups and legal publishers
In France, two types of actors are present in the predictive justice market: classic legal publishers and start-ups. The first, familiar with the production of case law databases, have developed an enhanced offer of computerized exploitation of these data. For example, Dalloz and Éditions Francis Lefebvre publish a “case law” online service. LexisNexis does the same, on the litigation of compensation in different areas of private law. These tools make it possible to identify, with great precision, the amounts of compensation received by a certain profile of litigants before all the courts of appeal and, sometimes, the courts of the first degree. As for start-ups, they are currently few in France. The two most active are Case-Law Analytics, an offshoot of INRIA, and Predictice. They include, among their founders, a lawyer. Predictice is experienced by two courts of appeal (Rennes and Douai) and the Lille Bar.
The faces of Legaltechs
In fact, the activities of the legaltechs are diverse:
Predictive justice: calculating probabilities from court decisions, intelligent systems and artificial intelligence, visualization of data from big data.
Information management tools: automated production of legal documents, certification of documents, completion of formalities and online filings (eg introduction of an instance), programming of intelligent contracts (updated automatically in case of change legislative or regulatory), document review.
Online conflict resolution tools.
LegalTechs are probably more “legal” than “tech”. Few of them are present in predictive justice. A survey of the audit firm Day One conducted on 140 start-ups in the world shows that their number has increased considerably since 2013 (figure below), but also that less than 11% of them only fall into the category ” predictive justice “.
The world of Legaltechs is not homogeneous. Some of them are limited to providing information (to find a lawyer or notary), others offer useful services to streamline the work of legal professionals (for example: conducting an automatic legal smart contacts). Still others focus on online conflict resolution.
The study finds that the most promising future prospects are in predictive justice. They are conditioned by future technological developments (in particular the transition from machine learning to deep learning). Currently, the algorithms use existing legal databases, which include the decisions of superior courts and courts of appeal. Few first level jurisdiction decisions are available. The law on the Digital Republic of October 7, 2016 (Lemaire law) provides in article 21 that “decisions rendered by the judicial courts are made available to the public free of charge with respect for the privacy of individuals. concerned. If this provision is implemented, it will have the effect of considerably enriching the data usable by producers of predictive algorithms.
However, value creation comes more from data than algorithms, often developed in open source. A report from France Stratégie estimates that “the value […] comes from the data needed for learning rather than the algorithm”. The enrichment of jurisprudential bases with the decisions of the courts of first degree will give more raw material to the legaltechs, perhaps also more precision in the results.