New research shows that people recognize more of their biases in algorithms' decisions than they do in their own -- even when those decisions are the same.