Building up to an Internal Family Systems model — LessWrong