What if judges used mathematics better to decide compensation for victims?

Imagine a traffic accident in which a motorcyclist would lose his life, and his passenger, who was his spouse, would be injured. This passenger is the victim of bodily injury, on two counts: as a direct victim, for her own injuries, and as an indirect victim, the death of her spouse having an impact on her life. It has a right to compensation, to be asserted against the insurer of the other vehicle involved in the accident. The latter has the obligation to make him, within a certain period of time, a complete and quantified offer of compensation. It will be up to the victim to accept this offer, or to take legal action. In practice, more than 98% of victims transact with the insurer ; the more serious the accident, the greater the sequelae, and the more frequent recourse to the judge.

The quantification of the compensation for certain items of damage is a pure calculation: add up the health expenses, reconstitute the loss of income… For other items, the operation is more delicate. How to assess the loss of affection of the passenger, for the death of her spouse? What hourly rate should be retained for the salary of the third party who must now help him in his daily life? What price to give to the alteration of his physical appearance due to his scars?

Find “fair” compensation

It is essential that each party, for each of these positions, be able to predict what would be compensation which could be paid by a judge. In fact, the victim, in order to make an informed decision to accept or refuse the insurer’s offer of payment, must have an idea of ​​what “fair” compensation would be. The insurer, for his part, has an interest in avoiding the lawsuit, and therefore in proposing a sufficient offer, knowing moreover that he is exposed to a significant penalty in the event of an offer. considered manifestly insufficient.

To this end there are indicative benchmarks for judges, the most used before judicial courts being that of Monsieur Mornet, designed with the aim of harmonizing the amounts allocated on the territory. It contains a calculation method including legal elements reflecting the applicable case law, but above all figures on the amounts that can be allocated. However, these figures are in no way binding on the judge, who is free to rule below or above.

[Près de 80 000 lecteurs font confiance à la newsletter de The Conversation pour mieux comprendre les grands enjeux du monde. Abonnez-vous aujourd’hui]

Relying on open data to improve knowledge

The generalization of open data of court decisionsi.e. easy access to all court decisions rendered in France, calls into question question the hegemony of repositories to achieve this predictability of the outcome of a dispute.

The mass of decisions made available to researchers and legaltech entrepreneurs (name given to companies applying computer technologies to the law) makes it possible to bring out the jurimetryin other words the application of mathematical methods to the study of legal phenomena.

Statistics applied to court decisions make it possible to bring out information that was not accessible with traditional methods, which are essentially based on taking into account a decision, or a few decisions, but not a mass. Thus, it turns out that regional disparities are emerging, whether on the type of litigation (traffic accidents occupy a lot of judges in the south-east, less elsewhere), or on certain compensation amounts. All other things being equal, and even setting aside sociological disparities in income, women seem to be less well compensated than menand the reason for this discrepancy remains unclear.

It also appears that the references, including that of Mr. Mornet, are used by the parties to formulate their requests or their offers, but that the judges also know how to extract themselves from them, to propose amounts of compensation. sometimes very far from the numerical indications of these documents. An example: for so-called average suffering, Mr. Mornet’s reference system recommends compensation between €8,000 and €20,000. However, the aforementioned study finds compensation ranging from €4,000 to €50,000.

For the victims, and the people who advise them, access to this information is essential, because it is likely to influence the interest of accepting or not an offer of compensation. Several companies are aware of this, and market online solutions allowing access to such statistics according to certain parameters, from a collection of manually analyzed decisions, such as Lexis360or automatically by natural language processing, like predictive.

A civil liability reform projectcurrently dormant, provided in its article 1271 that a large public database on compensation data be used to build a compensation reference system. This project was launched, under the name of Datajust ; experimentation, decried by some fearing that the tool is poorly designed and will lead to perverse effectswas completed without the results being able to be published, the task being apparently immense in relation to the means mobilized, which leaves the field open for private initiatives.

Optimization of strategies

Beyond the knowledge provided by statistical studies, jurimetry makes it possible to optimize the strategies of those involved in compensation in a finer way, especially when they find themselves before a judge. It was thus possible to demonstrate a “smile effect” compensation: when a payer makes a reasonable offer, the judge tends to rule in the interests of the latter.

If he makes a high offer, the judge is obliged to compensate at least up to this offer; but if he makes an offer that is too low, and this is the whole point of jurimetric analysis, the judge will tend to award a higher indemnity than if the offer had been reasonable. Conversely, legal claims are also experiencing a optimum for each position of damage : for example, for damage to affection linked to the death of a spouse, it is counterproductive to ask for more than €60,000, under penalty of seeing your compensation reduced. Other studies have shown that, before appellate courts, this effect is not always found and depends in particular on the procedural status of the victim (appellant or defendant before the court of appeal).

Companies specialize in providing advice so that strategies, both out of court and in litigation, are optimized. Case Law Analytics uses neural networks to simulate the judge’s reasoning to anticipate the risk represented by an order for compensation; i.praedico determines using artificial intelligence algorithms the most adequate amount of compensation, taking into account the propensity of the actors to go to litigation according to the compensation offers. No doubt they will be joined by other legaltechs in the near future.

Whether it is based on simple statistics or analyzes carried out by artificial intelligence, jurimetry is on the way to revolutionizing the way of carrying out research on the law, or of practicing it.


The authors founded a completely new publication, Jurimetry, review of the measurement of legal phenomena to stimulate reflection on this method, and compile works applying it to the different branches of law. The first number illustrates this diversity.

We would love to say thanks to the author of this short article for this incredible material

What if judges used mathematics better to decide compensation for victims?


Check out our social media accounts and other pages related to themhttps://www.ai-magazine.com/related-pages/