Metikoš, a postdoctoral researcher at the UvA’s Institute for Information Law, examined the legal and ethical questions that arise when AI is used in judicial decision-making procedures. His dissertation, titled ‘Enabling Contestation: The Right to an Explanation of Judicial AI’, makes the case for a legal right to explanation for citizens who encounter AI-assisted justice.

The applications of AI in the courtroom vary considerably. On one end of the spectrum are relatively neutral tasks such as sorting documents or searching for case law. On the other end are systems that profile individuals and calculate, for example, that a person has a seventy percent probability of reoffending, or that actually draft judicial rulings. Metikoš emphasises that such systems can have far-reaching consequences in cases involving custody, prison sentences, and welfare benefits.

In the Netherlands, pilot projects have explored the use of AI to draft rulings in straightforward cases, but broader application remains limited. Profiling systems, however, are already used extensively, for example by the probation service. Internationally, China and Brazil are at the forefront of AI use in the justice system. Brazil also has extensive and detailed legislation specifically addressing this use, which according to Metikoš makes it a more instructive example than China, where regulation lags behind.

One of the central problems Metikoš describes is that citizens may have many questions about how AI has operated in their case: why did a system produce a particular ruling, why was a specific recidivism percentage assigned, or which cases were used for comparison and which were not. Without insight into the underlying logic, citizens are unable to identify errors or contest decisions. This directly affects procedural justice, meaning the extent to which people experience the process and its outcomes as fair.

Metikoš also notes that it can sometimes be technically difficult or even impossible to fully understand how an AI system arrived at a particular outcome. The largest AI models can behave in ways that are illogical and untraceable to human observers. Nevertheless, he considers transparency indispensable. When citizens are reduced to a set of factors they are unaware of and cannot respond to, they are, in Metikoš’s view, not being treated as full participants in the procedure. The researcher states that failing to provide this insight reflects a profound lack of respect for the people involved.