Senza giudice: are computers fit to decide?
Marcello Gaboardi is Associate Professor of Law at Bocconi University of Milan. He teaches civil procedure law, insolvency law, and arbitration. He was also admitted to the Italian Bar in 2009.
During the French Revolution, the reaction to Absolutism led Baron de Montesquieu to hope for judges who would simply become “the mouthpieces of the law”. The myth of the judge who mechanically applies the law without employing any discretionary power of interpretation still endures. Over two centuries later, the new automaton seems to be the “judicial algorithm”. An algorithm can simply be described as a set of mathematical instructions that, especially if given to a computer, will help us to solve a problem. The promise of a judicial algorithm, therefore, is that in the near future a case will be decided by an automatic process.
The idea of a judge who simply applies the law immediately seduced French Revolutionaries and their followers. Likewise, the idea of an entirely digitized decision-making process seduces modern lawyers and scholars. It promises to reduce the risk of human errors and prevent legal uncertainty. But history teaches that the French Revolutionaries failed in that regard. Their mistake was to believe that improvement of human judgment depended on its automation.
The human factor
On the contrary, automated judgment simply becomes inhuman. It can save time, money, and even avoid errors but it lacks ability to adapt to social changes. For example, where would racial equality be today without the U.S. Supreme Court’s unanimous opinion in Brown v. Board of Education? The decision that U.S. state laws establishing racial segregation in public schools were unconstitutional implied not only legal but also social, ethical, and political considerations. Is an algorithm capable of considering these factors? Can it make, for example, innovative decisions by overruling existing precedents on the basis of new factors emerging in the society?
Artificial intelligence raises serious challenges in several areas of social and individual life. Access to justice is one of them. Several countries have long since digitized civil proceedings. Under Italian law, for example, any civil lawsuit can be brought before a first instance court or an appellate court electronically. Summons and pleadings are generally digital. During the pandemic, the use of modern communication technologies, such as videoconferencing, has significantly increased. Notwithstanding these changes, until recently no one dared to seriously think that a case could be decided totally or partially by artificial intelligence. Increasing computing power and complex algorithms make it possible to imagine a decision-making process guided or, at least, supported by artificial intelligence. The human responsibility for decisions remains crucial while judicial activities such as analyzing electronic documents or calculating economic losses and damages can be performed by an algorithm.
Trials in Italy
Some Italian civil courts (such as the first instance tribunals of Turin and Genoa) have recently explored how algorithms can gather complex information and support the decision-making process. Court personnel worked with computer experts to develop algorithms which can predict litigation outcomes and potential remedies. The experts focused their attention on certain types of standardized procedures such as family disputes and damages lawsuits. These cases present several features that remain constant from case to case.
Consider, for example, the reasons why people apply for a divorce. First, the experts developed a comprehensive classification of relevant claims such as divorce petitions, petitions for child custody and support, applications for legal separation or annulment of the marriage, and so forth. Then, they established the more frequent variables for each type of claim such as the number and age of any children or the spouses’ financial situation. Finally, they collected relevant precedents on more frequently debated matters. By following the reasoning behind previous decisions, the algorithm is able to provide guidance in future when cases to be decided fit the models created by experts.
Although the algorithm is still being written, this early experience demonstrates the constraints of digitizing the decision-making process. The algorithm seems to be unable to bear the burden of deciding on its own. Rather, it is used to supply the judge with a proposal for a decision that human judges are asked to take on their own. While the decision-making process is supported by the algorithm, a form of human scrutiny remains necessary. The decision proposed by the algorithm should be scrutinized by the same court that benefits from the algorithm, or by the appellate court reviewing the decision made by the inferior court with the support of the algorithm. This human scrutiny ensures that the decision takes into account an array of non-standardizable factors such as the specific circumstances of the case or the need to overrule a precedent.
A digitized decision-making process is likely to become a potentially valuable support for an inevitably human judge. Civil cases will be decided with the support of a judicial algorithm. The algorithm increases the efficiency of the decision-making process by reducing the costs and time involved in human decisions. Nevertheless, justice will need to maintain a human form.