AI on trial – will the backlog of cases finally be cleared?
Munich, 19 December 2023
Like many other areas of our society, artificial intelligence (AI) will transform the field of justice. In many respects, AI applications open up new possibilities of freeing up the justice system. For example, AI-supported processing will be able to shorten the duration of proceedings in similar kinds of cases, such as airline passenger compensation or minor motor vehicle damage. However, the use of AI in the sensitive area of law also raises ethical questions. Experts and guests discussed the instances in which the use of AI is wise and its place in our legal system at acatech am Dienstag on 12 December.
In his welcome address, acatech President Jan Wörner went into the public’s hopes and concerns in relation to AI. He mentioned the political agreement that has just been reached by the EU institutions on the Artificial Intelligence Act (AI Act), which will specifically govern the reliable use of AI in Europe. Mr Wörner emphasised that the dynamic advancement of the technology should also be taken into consideration. The important thing, he said, is to create trust in institutions through transparency.
Applications for AI technologies in legal practice
Frauke Rostalski, Chair of Criminal Law, Criminal Procedure, Philosophy of Law and Comparative Law at the University of Cologne (in German) and member of Plattform Lernende Systeme (PLS), gave an introduction to the legal ethics. She gave three examples of possible applications of AI in the justice system: an electronic (criminal) courtroom, a sentencing database and a “iudex ex machina”; that is, a sort of robot judge. The applications differ greatly in terms of autonomy: ranging from decision support software to systems that replace members of the legal profession, said Frauke Rostalski. AI decision support tools can be used directly in existing processes: for example, an electronic (criminal) courtroom automatically produces a transcript of the trial from audio recordings. A sentencing database can help ensure more consistent sentencing for similar offences, by providing an insight into the sentences passed in other regions. Up until now, sentences handed down in Germany have varied very considerably from region to region, which, according to Frauke Rostalski, could potentially jeopardise the acceptance of sentences. However, not enough personnel with the ability to classify the judgements and record the data were available. It is now hoped that the creation of such sentencing databases can be automated with the help of AI.
Frauke Rostalski concluded her talk by presenting the concept of a robot judge, “iudex ex machina”. This decision-making software could effectively pass sentence itself. However, AI systems lack a key feature of the legal system: the ability to weigh up reasons. It is also crucial that changes in case law and legislation be fully taken into consideration and older, superseded principles be disregarded. She summed up by saying that accountability remains paramount in the justice system and that humans must always make the final decision. However, AI applications should be available as a helpful tool.
Is the justice system ready for AI? And what support is needed?
Stefan Schifferdecker, German Judges Association (DRB) – Berlin State Association (in German), confirmed: There are no plans to install a robot judge in Germany; accountable decisions will continue to be made exclusively by people. That said, the justice system will not be able to do without AI in the future. The main question is not whether but how soon we can expect AI support to come along. Despite a decline in the number of proceedings, the workload is increasing as the cases are getting more extensive, the duration therefore longer and the volume of data to be considered ever larger. AI applications could take the pressure off judges on a lasting basis. However, the justice system also needs better technology skills to decide which products to use and which ones they can trust.
What he would like to see is anonymisation and translation software as well as an AI that evaluates and maps data. It would then fall to the judiciary to come up with an ethical compass for its use. The aim is to deepen trust in decisions and to ensure that fair judgements are passed despite the use of AI. The efficiency gains that could be achieved would be considerable and occasional mistakes that may arise could be caught by the established control mechanisms, said Stefan Schifferdecker, in response to a question asked by moderator Linda Treugut, Plattform Lernende Systeme, about how judicial control of AI recommendations would work in practice.
How can AI provide us as consumers with legal support?
In his talk, Jan-Frederik Arnold, CEO and Managing Director of Flightright GmbH, spoke about the use and function of artificial intelligence in companies. The first step Fightright GmbH took in this direction was to increase investment in digitalising and automating stages of its processes. For around six years now the company has been using machine learning to calculate the probability of extraordinary circumstances that exclude compensation payments by airlines. For this, an array of structured data from the company’s own database is used: for example, flight, weather and air traffic control data. In addition, pilot projects are being run with large language models to filter relevant data out of the documents submitted. At the moment, all of the AI’s results are being rechecked by a human, said Jan-Frederik Arnold on recent developments with regard to the use of AI.
The final topic of discussion was how to achieve the safe and reliable future use of AI in our legal system. Frauke Rostalski reiterated how important it is that judgements be published in anonymised form, so that they can be analysed to ensure comparability. In addition, judges and public prosecutors have to learn the necessary skills. Stefan Schifferdecker added that it will cost a lot to develop such tools, for which all German states should join forces to ensure that the endeavour does not fail for lack of resources. Jan-Frederik Arnold emphasised that the justice system and legal technology companies must work together to develop and put in place interfaces and common standards.