Artificial intelligence predicts lawsuit outcome with high certainty

British and American researchers have developed an artificial intelligence that can predict the outcome of human rights cases with an average certainty of 79 percent. The system does this by analyzing texts of statements.

The system had to predict whether or not a particular case constituted a violation of three different articles of the ECHR. The scientists found that the most important factors for predicting the outcome are the language used and the subjects and circumstances of the text. For example, the judges’ decisions were more related to the facts of the case than to the legal arguments. The 584 cases analyzed all come from the European Court of Human Rights in Strasbourg, the scientists write. These can be found in a public database. By analyzing the texts with a machine learning algorithm, they were able to identify patterns and make predictions based on that.

According to one of the scientists, the system is not capable of replacing lawyers or judges, but it is valuable in identifying patterns that indicate a particular outcome. In addition, the system is useful for identifying cases that are likely to lead to a violation of human rights. Until now, research on the prediction of court decisions has mainly focused on predicting the decisions of individual judges, without using textual information.

The ECtHR cases are particularly suitable for text-based analysis, according to the researchers, because they have a clear division. For example, a judgment must contain a section in which the facts are described and which procedure preceded the case. The researchers derived different properties of the text from each relevant part, such as so-called n-grams. Scientists from the British universities of Sheffield and London and the American University of Pennsylvania will present their research on Monday.

Illustration of the classification of an ECtHR case