Computational signatures of psychopathy — LessWrong