(Adapted from chapter 1, pp. 2-8)
Most writers on belief assume that it is a unitary psychological kind – that whenever we ascribe a belief to a person, creature, or system, we ascribe essentially the same kind of state. Of course, no one denies that belief has varied aspects and manifestations – it is widely accepted that beliefs can be both occurrent and standing-state, explicit and tacit, conscious and non-conscious, and so on. But it is generally assumed that these are different aspects or variants of the same underlying state. So occurrent beliefs can be thought of as activations of standing-state beliefs, tacit beliefs as dispositions to form explicit beliefs, conscious beliefs as beliefs that are the object of higher-order beliefs, and so on. We might call this as the unity of belief assumption. A similar assumption is often made about reasoning. It is assumed that this, too, has a uniform character, and that thought-processes involve a single, generic kind of activity – computational operations in a mental language, say, or associative processes of some kind. This assumption – the unity of processing assumption – is not quite so pervasive as the parallel one about belief, and has occasionally come under challenge from psychologists. It is common in the philosophical literature, however, and shapes many of the debates there.
There have, it is true, been dissenting voices, suggesting that the apparent uniformity of folk-psychological discourse masks important psychological distinctions. Some writers distinguish passive belief formation from active judgement. Philosophers of science, too, commonly mark a distinction between partial and flat-out belief (sometimes called ‘acceptance’). And Daniel Dennett has argued that we must distinguish non-verbal beliefs from a class of language-involving cognitive states which he calls opinions. The failure to distinguish these states, Dennett suggests, lies at the root of many philosophical misconceptions about belief.
It would not be too surprising if something like this were true. Everyday users of folk psychology are interested primarily in behavioural prediction and explanation, not precise psychological taxonomy. If two psychological states or processes were similar enough to be lumped together for everyday purposes, then we should not expect folk psychology to make a sharp distinction between them – though it might register their distinctness in indirect ways. The states and processes in question might nonetheless differ significantly, and it might be important for a developed psychology to distinguish them – even if we continued to conflate them for everyday purposes. The introduction of new distinctions like this is common in science. Consider, for example, how psychological theory has adopted the common-sense concept of memory, while at the same time distinguishing various kinds of it – long-term, short-term, episodic, procedural, semantic – each with different functions and properties.
I believe that something similar will happen with the folk concepts of belief and reasoning. These, I suspect, conflate two different types of mental state and two different kinds of mental processing, which form two distinct levels of cognition. In short, we need a two-strand theory of mind. Only by developing such a theory, I believe, can we resolve some deep disputes about the mind and provide a sound basis for integrating folk psychology into science.
There have, however, been few sustained attempts to develop two-strand theories of belief. Few theorists have sought to link up the various distinctions that have been proposed or to explore their implications for issues in philosophy of mind. And while some psychologists have advanced ‘dual-process’ theories of reasoning, there have been few attempts to integrate these theories with two-strand theories of belief or to consider their philosophical consequences.
This book aims to remedy these omissions. Chapter 2 begins by highlighting some divisions in the folk notion of belief – divisions relating to consciousness, activation level, degree, method of formation, and relation to language. These divisions, I argue, are real and run deep and link together in a natural way to yield a tentative two-strand theory of belief – the first strand non-conscious, partial, passive, and non-verbal, the second conscious, flat-out, active, and often language-involving. I then move on to look at similar divisions in our view of reasoning. Again, I argue that these run deep and indicate the need for a two-strand theory – the strands corresponding closely to the two strands of belief. The final section of the chapter looks at some further divisions in folk psychology, concerning the ontological status of belief and the function of psychological explanation. I identify two broad interpretations of folk psychology, which I call austere and rich, and which correspond roughly to the views of philosophical behaviourists and functionalists respectively. On the austere interpretation, folk psychology is a shallow theory, which picks out behavioural dispositions and offers explanations that are causal only in a weak sense. On the rich interpretation, it is a deep theory, which aims to identify functional sub-states of the cognitive system and to offer causal explanations of a more robust kind. I suggest that these two interpretations each have a firm basis in the folk outlook and that a reconstructed folk psychology needs to admit both. The two interpretations, I argue, indicate the need for two theories, corresponding to the two strands of mentality identified earlier: an austere theory for the non-conscious strand, and a rich theory for the conscious one.
It is one thing to identify two strands of mentality, of course, another to construct a substantive two-strand theory of mind. A developed theory will need to explain how the two strands are related to each other, what role they play in reasoning and action, and how they combine to form a single intentional agent. Chapters 3, 4, and 5 are devoted to this task.
Chapter 3 begins by setting out some challenges to the proposed two-strand theory. Prominent among these is what I call the ‘Bayesian challenge’ – the challenge of reconciling our common-sense belief in the existence and efficacy of flat-out belief with a Bayesian view of rational decision-making. I then review some precedents for a two-strand theory of mind, seeking hints as to how to develop the theory and respond to the challenges. I focus in particular on possible models for the conscious, flat-out, language-involving strand of belief, and on suggestions as to how this strand might be related to the other, non-conscious strand. Although none of the models examined fits the bill exactly, I identify several promising ideas, including the behavioural view of flat-out belief developed by some Bayesians, Dennett’s picture of the conscious mind as a virtual machine, and Cohen’s account of acceptance as a premising policy. I conclude the chapter by suggesting how elements of these views can be combined to give a picture of the conscious mind as a premising machine, formed by the adoption and execution of premising policies, and driven by non-conscious, partial beliefs and desires.
Chapter 4 is devoted to filling in the picture of the premising machine sketched in the previous chapter. I discuss the nature and scope of premising policies and distinguish several varieties of them, including a goal-oriented form. I then look at what is involved in executing these policies and what role natural language plays in the process. Finally, I consider how premising policies are related to other mental states and how they influence action. Crucially, I argue that an agent’s premising policies are realized in their non-conscious, partial beliefs and desires – and thus that the premising machine constitutes a distinct level of mentality which supervenes on the one below it. To emphasize the point, I call premising policies supermental states, and the level of mentality they constitute the supermind. By analogy, I call the non-conscious attitudes in which the supermind is realized the basic mind. Because the supermind is realized in the basic mind, I argue, supermental explanations of action are not in competition with those pitched at the basic level. Rather, each corresponds to a different level of organization within the agent.
Chapter 5 shows how we can use the framework developed in the previous chapter to flesh out the two-strand theory outlined in chapter 2. I begin by arguing that conscious, flat-out beliefs can be identified with a particular subclass of premising policies, and thus that they, too, are supermental states. The upshot of this is that our two-strand theory of mind becomes a two-level one, with conscious, flat-out states realized in non-conscious, partial ones. (In line with the terminology adopted earlier, I call the former superbeliefs and the latter basic beliefs.) I then go on to highlight the attractions of this view, and to show how it can resolve the challenges posed in chapter 3. The chapter concludes with some remarks on the function of the supermind. I argue that supermental capacities carry with them considerable cognitive benefits. The supermind is a slow but highly flexible system, which can kick in whenever faster but less flexible basic processes fail to yield a solution. Moreover, because supermental processes are under personal control, we can reflect on them, refine them, and supplement them. The flexibility, adaptability, and improvability of human cognition flow directly from the supermind.
Having developed the core theory and shown how it can vindicate some important aspects of folk psychology, I then move on to consider some further folk commitments and to show how these, too, can be vindicated by the theory.
There is a case for thinking that folk psychology makes some substantial assumptions about the functional architecture of the cognitive system. In particular, it has been argued that there is a folk commitment to the theses of propositional modularity and conceptual modularity – accounts of how propositional attitudes and their component concepts are stored and processed. The idea that there is a folk commitment to these theses is worrying, since they in turn seem to involve claims about the architecture of the brain, and therefore to run a risk of empirical falsification. If the folk are committed to them, then their conception of the mind may be seriously mistaken and ripe for revision or even elimination.
In chapters 6 and 7 I show how the two-level theory developed in earlier chapters can vindicate folk psychology’s architectural commitments. Chapter 6 deals with propositional modularity and chapter 7 with conceptual modularity. In each case I begin by arguing that there is indeed a folk commitment to the thesis, building on arguments in the literature. I then show that this commitment can be vindicated at the supermental level, without involving claims about the structure of the brain. I show that the supermind exhibits both propositional modularity and conceptual modularity, and thus that the folk assumptions are correct. I argue, however, that there is no inference from this to claims about the structure of the brain. The supermind is implemented, not in the hardware of the brain, but in basic-level intentional states and actions. And the basic mind need not itself exhibit propositional or conceptual modularity in order to support a supermind that does. The upshot of this is that the folk architectural commitments are compatible with any account of the underlying neural architecture. Given this, the threat to folk psychology – in this guise at least – vanishes.
The final chapter outlines some further applications of the proposed theory – starting with a discussion of akrasia and self-deception. These conditions can seem puzzling, and it is sometimes suggested that they reveal the presence of conflicting subagents within the human psyche. Here a two-level theory offers a different and, I think, more attractive perspective. The conditions can be thought of as involving a conflict, not between subagents, but between levels of mentality – the attitudes at one level in tension with those at the other. In each case, I sketch the two-level account and show that it offers an economical way of resolving the associated puzzles. The chapter then moves on to look at first-person authority. Again, this can appear puzzling. How can we be authoritative about our mental states, given that there are independent behavioural criteria for their possession? And, again, the present theory offers a fresh perspective. The thought is that first-person authority proper extends only to supermental states, and is primarily a matter of control. Superbeliefs can be actively formed and processed, and in self-ascribing these states we are not simply reporting that we meet the criteria for their possession, but committing or recommitting ourselves to meeting them. The authority attaching to such an ascription, then, is that of a sincere commitment, rather than that of a reliable report.
In addition to helping to clarify philosophical debates about belief, the theory developed here may have application to issues in developmental and clinical psychology, and the final chapter closes with some brief remarks on this. In particular, I suggest that the theory may be able to shed light on the nature of autism.