Integrating Three Models of (Human) Cognition — LessWrong