Algorithm developed to predict Supreme Court decisions
DOI: 10.1063/PT.5.028139
Ars Technica : A new algorithm can predict the voting behavior of the US Supreme Court with roughly 70% accuracy, according to a study published on the Social Science Research Network. “Our model has proven consistently accurate at predicting six decades of behavior of thirty Justices appointed by thirteen Presidents,” says one of the developers, Josh Blackman, on his blog . The algorithm uses the extremely randomized trees method to weight some 90 variables, including decisions in individual cases and background information on the justices and the court. The algorithm will be tested against humans in an upcoming tournament, in which it will be pitted against the players of Blackman’s FantasySCOTUS .
