Seeking Solutions for Aggregating Classifier Outputs — LessWrong