The ecology of conviction — LessWrong