AI Predictions of Court Rulings are Surprisingly Accurate

Researchers at the University of Sheffield in collaboration with the University of Pennsylvania have developed AI systems reportedly capable of predicting European Court of Human Rights (ECHR) judicial decisions of the with 79% accuracy. The team has published the results of their study in the journal PeerJ Computer Science.
In conducting the study, a team of legal aids and computer scientists began by extracting case information published by the court from a publicly accessible database.
After sifting through the information, the team selected a large dataset of 584 cases relating to Articles 3, 6 and 8 of the European Convention on Human Rights.
Then, the legal aids selected an equal number of cases in violation and not in violation of the articles of the convention.  Armed with the data, their AI considered the entire dataset to form its predictions.
Armed with the data, the team then fed the entire data set to the AI as the basis on which to form its predictions for case outcomes.
The research team found that the most important factors for the AI’s predictions were: the language used, the topic, and the specific circumstances of the case.
Circumstances refer to the background of the case and topics refer to case relevant articles of the Convention. According to Dr. Nikolaos Aletras who led the study, AI is not likely to replace judges and lawyers.
However, the court itself could use an AI to weed out frivolous cases before they reach the hearing stage – a fact that reinforces AI’s future as a supplementary tool, and not a replacement or threat for humans