Pseudolikelihood as a source of cognitive bias — LessWrong