From beating some of the best players in complex board games to controlling self-driving cars, significant advances in artificial intelligence software have been made and implemented in various fields such as health-care, finance and transportation. Now, a group of UK scientists has developed an AI "judge" that can predict the outcome of human rights trials.
According to a study, conducted by researchers at University College London, the University of Sheffield and the University of Pennsylvania, the AI system has been able to accurately predict the outcomes of hundreds of cases heard at the European Court of Human Rights- involving torture, degrading treatment and privacy - with 79% accuracy.
"We don't see AI replacing judges or lawyers, but we think they'd be useful for rapidly identifying patterns in cases that lead to certain outcomes," said Dr. Nikolaos Aletras, the lead researcher from UCL's department of computer science. "It could also be a valuable tool for highlighting which cases are most likely to be violations of the European Convention on Human Rights."
While developing the program, researchers found that judgments by the European Court of Human Rights depended more on non-legal facts rather than purely legal arguments, suggesting that the judges were more "realists" rather than "formalists."
After identifying English language datasets for 584 cases, the researchers applied an AI algorithm to look for patterns in the text and label each case as a "violation" or "non-violation." The researchers selected an equal number of violation and non-violation cases for the study.
They found that the most reliable factors for predicting the courts' verdicts were the language used as well as the topics and circumstances stated in the case texts.
"Previous studies have predicted outcomes based on the nature of the crime or the policy position of each judge, so this is the first time judgments have been predicted using analysis of text prepared by the court," UCL computer scientist Dr. Vasileios Lampos said.
However, the team did note that the algorithm was prone to making mistakes in determining verdicts when there were two similar cases, one of which was a violation and the other a non-violation, suggesting that the software still needs to be further developed to identify subtle differences and nuance in the legal data.
"We expect this sort of tool would improve efficiencies of high level, in demand courts, but to become a reality, we need to test it against more articles and the case data submitted to the court," Lampos said. They also add that the tool could help improve the significant delay imposed by the Court and possibly encourage more applicants who may have been previously discouraged by the expected delays.
Their findings have been published in the journal PeerJ Computer Science.