Moral cognition can be fast or slow. But recent research in cognitive science suggests that for moral thinking to be done well, a third type of cognition is needed: critical thinking. Does this mean it is time for triple process moral psychoogy? I think the answer is yes.
Over the past decade, dual-process theories of cognition have become the dominant empirical paradigm for understanding moral judgment and reasoning. These theories aim to understand moral thinking in terms of two separate mental subsystems: one slow and analytic, the other quick and crude. Both types of processing suffer from complementary defects: the former system is adaptive and efficient, but prone to error; the latter is flexible and sophisticated, but its power is limited, and its deployment costly. However, it has proven surprisingly difficult for dual-process theories of moral judgment to deliver on their main promise of bringing the tools of empirical psychology to bear on the question which moral rules and principles make sense, and which would be better given up.
I believe that moral thinking and reasoning are only insufficiently understood when described in terms of a quick but intuitive and a slow but rational type of cognition. This approach severely underestimates the importance and impact of dispositions to initiate and engage in critical thinking – the cognitive resource in charge of counteracting my-side bias, closed-mindedness, dogmatism, and breakdowns of self-control. Moral cognition is based, not on emotion and reason, but on an integrated network of intuitive, algorithmic and reflective thinking.
Let me spell out this idea in a little more detail. Some judgments are made quickly and effortlessly, others require deliberate thought. Dual Process Theories (DPT) in cognitive science capitalize on this observation, and argue that the human mind consists of two subsystems (Evans 2003, 2008): System I processes information automatically, intuitively, and without drawing on scarce cognitive resources. System II works slowly, analytically, and requires controlled conscious processing (Kahneman 2012).
What does this model entail for moral cognition? Are moral judgments based on intuition and emotion (Haidt 2001, Prinz 2007)? Or are they grounded in reason and conscious deliberation (Kennett 2009, Sauer 2012)? Many philosophers and psychologists hope that DPT will help us shed light on these issues, and allow us to address questions such as whether morality is relative, or whether moral norms can be rationally justified.
Existing dual process theories suffer from various problems (Evans and Stanovich 2013): the criteria for distinguishing System I and II are often vague; System I is not a unified system, but really a set of autonomous modules. More importanly, however, DPT overlooks the fact that System II must be further differentiated into an algorithmic and a reflective system (Stanovich 2009). The former is in charge of the tools that allow us to override non-normative default responses generated by the intuitive system. The latter is reponsible for the dispositions of critical thinking that initiate and monitor this override. This distinction is so important that some have proposed a move towards triple process accounts of cognition. But, unlike dual process theory, these developments have not yet been applied to issues in moral psychology. I think this gap should be closed by investigating the prospects of triple process moral psychology (TPMP). Critical thinking is one of the main aspects of competent moral judgment. TPMP will make it possible to conceptualize the significance of reflective thinking for moral agency, and to develop an account of human moral cognition that is both empirically informed and normatively satisfying.
It frequently remains unclear, for instance, when and why System II intervenes upon the intuitive reponse generated by System I, and whether such interventions tend to be detrimental or beneficial. Triple Process Theory is supposed to remedy this problem by arguing that System II is really comprised of two functionally separate systems. This is not just a terminological distinction: for one, the difference between the algorithmic and the reflective mind manifests in different patterns of breakdown (deficits in fluid intelligence vs. cognitive delusions); for another, reflective reasoning and algorithmic processing can be tapped by measuring performance on intelligence and critical thinking tasks, respectively, neither of which neatly predict performance on the other (Stanovich and West 2000).
The main function of System II is System I override: the default response generated by the intuitive system often needs to be intervened upon by conscious processing to arrive at the normative response. Crucially, however, this theory overlooks the fact that a) the need for such override has to be detected, b) the override itself needs to be initiated and c) the result of the override checked for accuracy. This higher level control is exercized on the basis of epistemic dispositions such as critical thinking, need for cognition, or open-mindedness. The initiation of default override is arguably the more fundamental process, yet the distinction between conscious processing and reflective self-criticism is lost on standard dual process models of moral judgment. Critical self-reflection is of central importance for moral psychology, since cognitive errors such as my-side bias, self-deception, or dogmatism are arguably most harmful in the moral domain.
This has far-reaching implications for the empirically informed study of moral cognition in both philosophy and psychology. Existing two-system approaches to moral judgment such as the Social Intuitionist model (Haidt 2001, 2008) or Joshua Greene’s dual process theory (2001, 2004, 2008) are either implausibly pessimistic about the influence of reasoning on moral judgment (Jacobson, Railton, Sauer) or suffer from a variety of empirical and conceptual issues (see Guy Kahane’s work on this; cf. Greene 2014). At the heart of both theories are accounts of intuition override: the main claim of the Social Intuitionist model suggests that such override happens surprisingly rarely (“moral dumbfounding”, Haidt 2001). Greene’s Dual Process account suggests that arriving at consequentialist responses often requires override of an emotionally more salient response. Yet neither of those theories offers a convincing account of when, why, and under what conditions such an override is initiated.
What are possible implications of this approach? For one, the distinction between rationalism and sentimentalism about moral judgment may well turn out to be scientifically obsolete. Another, perhaps more surprising upshot of TPMP may be that the influence of reason on moral judgment needs to be reconsidered as well. Here, two questions need to be addressed: 1) is this influence frequent or rare? 2) Is this influence beneficial or detrimental? One possible answer suggested by TPMP could be that the influence of Type II processing is frequent, but tends to result in harmful rationalizations and moral disengagement; the influence of reflective Type III processing, on the other hand, is beneficial, but tends to be rare. The resulting theory will be a form of rationalism with a pessimistic twist.
