Humans Beat Robots at SCOTUS Predictions

Reports that an algorithm can outperform legal scholars at accurately predicting the outcome of Supreme Court cases are largely exaggerated, FiveThirtyEight’s Oliver Roeder writes. Countering an article in Vox, Roeder says the paper that makes the claim that algorithms win covered only the 2002 Supreme Court term, and the algorithm in question outperformed a single polled group of legal academics, and still only managed to get 75 percent of the 68 cases it predicted correct. Linda Greenhouse, a reporter who covered the Court for 30 years, matched that percentage in her predictions in articles. A new algorithm, {Marshall}+, is being routinely trounced by human predictors.