What Does Metacognition Do For Us?

Philosophy and Phenomenological Research Philosophy and Phenomenological Research Vol. LXXXIX No. 3, November 2014 doi: 10.1111/phpr.12148 © 2014 Philosophy and Phenomenological Research, LLC What Does Metacognition Do For Us? BARRY C. SMITH University of London Commonsense psychology is framed in terms of the concepts we use to characterize our own and others’ mental states. These include the concepts of belief, desire, intention, perception, action, memory, imagination and emotion. In addition, we deploy concepts of mental processes and activities such as acting, thinking, feeling, remembering, perceiving; and to these we can add further categories of diffuse mental phenomena such as experience, consciousness and self-consciousness, awareness and self-awareness. These concepts pick out, in ourselves and others, what are taken to be the key features of being minded. So to attempt to revise this list, either by addition or subtraction, marks a significant step in the philosophy of mind and to take that step requires a substantial argument or theory. This is what Jo€elle Proust attempts in the first, comprehensive book-length treatment of metacognition, which she takes to be a capacity to monitor and control one’s mental performance in perception, memory, learning, reasoning, and communicating. For Proust, metacognition also plays a fundamental role in constituting oneself as a person, and so, on Proust’s recommended understanding of the notion, metacognition is an indispensible part of the toolkit for the human mind. In defense of her preferred, evaluative and procedural conception of metacognition, she provides an impressively comprehensive account of the reach and workings of metacognition informed by relevant findings from primatology, psychology and neuroscience. The range of application is striking, though it is here that one begins to worry about whether there is a single notion or distinctive set of processes picked out by the term metacognition. This is not because of any looseness of definition on Proust’s part. On the contrary she goes into considerable and convincing detail to distinguish her brand of evaluative metacognition from metarepresentation on the one had, and merely associative processes on the other. These are important BOOK SYMPOSIUM 727 clarifications, but still, the rather demanding notion of regulative, norm-sensitive metacognitive processing that emerges finds an application everywhere from tuning up perceptual discriminative capacities, arriving at learning heuristics, selecting appropriate epistemic norms, gauging the likelihood of one’s remembering, assessing performances in reasoning, and restraining impulsive behavior. Can it be one capacity—one kind of process—that does all of these things? Or is the term metacognition just a place-holder for a diverse set of regulatory systems for monitoring, evaluating and guiding first order mental states and processes? I shall return to this question below. I At this stage, it is worth reflecting on how surprising it seems that such a fundamental feature of the mind should have been overlooked until now in philosophical accounts of the mind. This may mean that philosophers thought they could deal with the tasks metacognition is said to perform without introducing a new level into the cognitive architecture of the mind; usually by introducing thought about thought, conceived as involving representation of the contents of first-order states. Proust usefully shows why such a move will not do, and why it remains too static a conception of the dynamic mind. But perhaps the reason why metacognition has appeared on the philosophical scene now is due to the relatively recent trend for philosophers of mind to be suitably informed about relevant empirical work in psychology and neuroscience. This trend is leading to revisions in traditional conceptions of mind that have held sway for over two thousand years. For example, until recently philosophers took it for granted that we had just five senses that each operated independently of one another: a view hardly sustainable in the light of neuroscientific research into multisensory perception. In the case of metacognition, however, the emergence of the phenomenon had rather humble experimental roots based on cases like the tip of the tongue phenomenon, when we try to recall a name; or feelings of knowing when we try to remember a number, or solve a puzzle. The cases of metamemory, where we are able to assess our own resources and likelihood of success in remembering an item were seen as interesting but peripheral phenomena in the study of cognition. What is, therefore, so impressive about Proust’s treatment of metacognition is how far she has developed the notion and how central she takes its place to be in normal, day to day, mental functioning. There are detailed discussions of early stage and later development of mentalizing capacities in children, personal identity, neuropathologies—in particular, schizophrenia—linguistic communication and primate cognition; each of which is valuable in its own right. But what remains 728 BARRY C. SMITH open is whether the notion of metacognition being called on to do explanatory work in all these cases is just one thing or many. II According to Proust, ‘metacognition is a set of processes whose function is to regulate mental or cognitive actions’ (149), and therefore any creature with metacognition has greater flexibility to adjust its performance dynamically in a given context than creatures at the mercy of first-order states or routines. At the same time, this potential for evaluating and refining one’s approach to a first-order task does not require full-blown thought about thought: ‘the cues for cognitive success. . .will be properties of processing, not properties of content’ (56). This is an important distinction because it can help in drawing the boundary in the right way between the minds of human and non-human animals. A convenient way of distinguishing between the mental capacities of human and non-human animals is to point to point to the achievements of the former due their ability to use language and make the contents of their minds available to one another. This capacity for language not only provides an interface between minds; arguably, it extends the capacities of individual minds considerably. But what exactly is it that language does for us? What changes in cognition are wrought by the advent of language? Philosophers such as John McDowell and Donald Davidson have argued that language is essential for thinking, and therefore that being a language user is necessary to being minded. Infants and animals, they conclude, lack rational minds. This conclusion, arrived at on the basis of an a priori argument, flies in the face of much empirical research by developmental and comparative psychologists into the cognitive capacities of infants and primates. So are these philosophers simply guilty of refusing to look at the facts? It is more complicated than this quick move suggests. McDowell and Davidson can acknowledge that there something in going on in the heads of these creatures but deny that having informational and affective states of the sort in question amounts to being minded in anything like the way adult humans are. The crucial line McDowell and Davidson would draw is between creatures who are, and creatures who are not, capable of thinking thoughts. But how should we characterize a thought? In the absence of a pure Fregean characterization of thought in general terms, both McDowell and Davidson would say that thoughts are the sorts of things we express and communicate by means of language by using structures with the logical forms of indicative sentences. Psychologists may conclude that other forms of thinking are available to infants and animals, such as spatial and emotional forms of thinking and reasoning, but the structure of such thinking BOOK SYMPOSIUM 729 will not be the structure of our thinking. That structure is analogous to the structure of language: it is logical form. Now it may seem circular to insist that non-linguistic creatures lack thoughts because they lack the means to express them linguistically. So is there some other, non-linguistic way to characterize the type of thinking and thought considered to be so characteristic of adult human cognition? To suppose there is does not answer the question of whether language is necessary for thought or whether thought is prior to language, for it may still be the case that the only way to grasp something like conceptually structured propositional thought is by means of acquiring a language in which such thoughts can be expressed. At this stage, it is often proposed that to be capable of full-blown thought one must be capable of reasoning—of exploiting the logical and rational connections between thoughts—and to do so effectively, one must be able to think about and assess one’s thoughts, to compare them to what one already knows, to determine whether some further thought is compatible or incompatible with what one already accepts. This capacity requires a thinker to have thought about thought. For Tyler Burge, this shows that thinkers must have self-knowledge; while for McDowell and Davidson, shared thoughts or thoughts about other’s thoughts are required to provide an objective background to our own thinking, and for that one requires a language. If one takes thought about thought to be a prerequisite for the thinking of thoughts, and if one takes the latter capacity to be an essential requirement of being minded—and these are big ifs—then the only way that non-linguistic creatures can have minds is if they are capable of thought about thought. And the test case for that would be whether non-linguistic creatures are capable of attributing thoughts to one another; that is, if they have some form of theory of mind. It is not surprising, therefore, that those who have fought to show that the minds of infants and animals are more socially sensitive and involve greater cognitive capacity than was previously supposed have sought to provide conclusive evidence of theory of mind, or mentalizing capacities, in non-human primates and pre-linguistic infants. The later is the subject of continuing controversy and Proust offers a thorough account of the key debates between Renee Baillargeon, Joseph Perner and others in Chapter Three about what goes on in infants before the onset of full-blown theory of mind and explicit false belief attribution, at around three and a half years of age, marking out two quite different levels of sensitivity and awareness of others’ mental states. However, in the case of non-human primates there is near consensus that while chimpanzees and monkeys approximate to human performance on falsebelief task, there is still no need to suppose that they attribute mental states like their own prior epistemic states to others. That is, there is no 730 BARRY C. SMITH reason to suppose they or infants have thoughts about thoughts. This leaves open that tantalizing possibility of explaining the capacity for mental metarepresentation as coming on stream with the advents of a recursive linguistic synax, and therefore as being the preserve of human language users. This would make the capacity for self-knowledge and precise knowledge of other minds to be an exclusively human capacity; and while I think this is right, and that that we are the only creatures with minds that know they have minds, it need not relegate non-human primates or infants to the class of creatures with an exclusively first-order, flat cognitive architecture, who are incapable of reflecting on, or responding to their own current mental states, or being suitably responsive to the effect others’ mental states have on them. And it is Proust’s account of procedural and evaluative metacognition that makes this insight clear. Metacognitive processes of self-monitoring and control to ensure flexible and improved performance can be detached from the attributive view of higher cognition that requires that one be able to represent one’s own and others’ first-order cognitive states. Unlike metarepresentation, which according to Proust ‘involves an ability to evaluate a first-order embedded content from a shifted viewpoint’ (97) bringing in reference to another, metacognition, for Proust, is focused exclusively on self-evaluation. III With the distinction between meta-cognition and meta-representation in place, Proust is able to provide philosophical support for the hypothesis that monkeys are capable of assessing their chances of success in a choice task by evaluating their own state of certainty, or uncertainty, in making a decision. The empirical work is that of David Smith and colleagues who train monkeys to operate a joystick to indicate whether a picture they are shown has high density or low density patterns on it. If they guess correctly they are given a food reward, but for every incorrect answer they suffer an unwelcome time out in the dark. When ambiguous patterns are presented and the monkeys have a third option of an ‘uncertain’ response with the joystick, which is followed by a new trial or an easier, less rewarding task, they take it. Having learned to avoid time-outs the monkeys tend not to gamble on guessing when presented with ambiguous stimuli. But is this response really indicating their uncertainty: something responsive to internal cues rather than clues about the first-order task? Can’t the third option just be an indication that they judge this is neither a clear food reward slide nor a time-out punishment slide? Or could they just be whizzing through the samples to get to another clear food reward slide? These and other interpretations considered by Proust are all available and indicate a rational use of the ‘uncertainty’ option. But do they all equally rely on the same self-evalu- BOOK SYMPOSIUM 731 ative metacognitive strategy? If so, these responses do not conclusively establish that the monkey is evaluating its own confidence in response to the task. If not, we need different metacognitive models to underpin the different interpretations and some experimental or theoretical way of deciding between them. The evidence of slower timing for uncertain responses favours the first option but nonetheless Proust’s interpretation is a strong one: The comparative evidence amply shows that monkeys and apes are sensitive to some of the constitutive epistemic norms for recurring cognitive tasks (98). This requires them to know when a decision is right or wrong, be able to use confidence judgements appropriately in the tasks, and care about the outcomes. The question remains whether being able to use confidence judgements appropriately—these being implemented by noetic feelings of uncertainty that motivate their response—amounts to self-evaluating confidence judgements. Though I agree with Proust that it certainly doesn’t provide evidence that the monkey uses ‘a theory-based model of being in a mental state’ (108). Non-human animals have more to their minds than many analytic philosophers have supposed, but this is not to suggest that we should see them as languageless humans. IV Having done such useful work to remind us that metacognitive processes of guiding and monitoring first-order states need not be metarepresentational or involve atrributive mentalizing capacities, and that therefore animals without propositional mental states are still capable of flexible and contextsensitive responses that trade on the ability to adjust their perceptual discriminations and monitor their memories, it is surprising that Proust goes in in Chapter Eleven to consider the role of metacognition in constituting persons. Proust tells us that: Constituting a person presupposes the ability to act mentally, which in turn presupposes the ability to monitor and control one’s own mental states on the basis of one’s past experiences and of one’s projects (231). The idea is that keeping track of the reliability of one’s memories and incorporating them into an updated view of oneself is a necessary part of any account of memory’s role in constituting the identity of a person over time. Now the constitutive role played here by metacognition, in continually sewing new memories into the fabric of the unfolding self, is potentially available to other non-human animals, unless this process 732 BARRY C. SMITH requires a language-using or conceptual, narrative view of the self. But why should we be committed to that? On the other hand, this is only a necessary condition for personhood and there is talk of past projects, so perhaps Proust would say that non-human primates lack a sense of a continuing project and so would not quite get there even if they had ‘the ability to monitor and control [their] own mental states on the basis of [their] past experiences.’ Nonetheless, it returns us to the question of whether we are really characterizing the same thing when appealing to person constituting metacognitive processes and task-related metacognitive processes that enable monkeys to act appropriately under conditions of uncertainty. Let me end by pursuing two questions, one related to the above issue: Is metacognition really a single psychological kind; is there an attributive use of procedural metacognition? V When presented with odours to smell, people are notoriously bad at naming them. Though, for familiar odours like vanilla, coffee and tar, they frequently say, ‘I know what that is.’ The question is whether this is a case of the tip of the tongue phenomenon, where one is unable to recall a name but has a strong sense that one will be able to retrieve it from memory; or whether this is a feeling of knowing based on recognizing the odour without being able to identify the associated name for it. Think for a moment of what is involved here. At first sniff there may be a feeling of familiarity: is this an odour I have encountered before or something utterly new? People may be reliable in making these judgements. Then, we may ask them whether they recognize the odour. This requires a subtly different decision. If the odour is familiar it may be because it resembles something one has already had the opportunity to smell, or it could be because one has already encountered this odour but cannot identify it. How do our noetic feelings distinguish here? If it was unfamiliar I might have a feeling of not knowing. But if I recognize the odour—or think I do—this may be because I have a feeling of knowing. But what is such a feeling telling me? Is it also an indication of the tip of the tongue phenomenon (or what Harry Lawless has called the tip of the nose phenomenon). I may be quite sure that I recognize the smell but equally sure that I won’t be able to retrieve its name. Does the same metacognitive system and the same feeling give me both results? Also, in the case of odour naming, the feelings of knowing are not predictive of success, and it has been conjectured that failures of odour naming, despite what is felt, are due to failure to identify the odour source or object by means of the odour (Jonsson et al. 2005). The unreliability of BOOK SYMPOSIUM 733 feelings of knowing in this modality may point to specific features of the case. Odour names are usually names of odour sources, e.g. strawberry, but odours when presented by themselves without the accompanying multimodal clues (i.e. colour and shape of odour source) may significantly reduce the amount of semantic information available for the retrieval process. So while one cannot retrieve the name, this may be because one cannot identity the object to be named. It has yet to be ruled out that the sorts of metacognitive processes and feelings at work in the case of olfaction are unlike those at work in monitoring and controlling cognitive tasks. VI Proust’s procedural account of metacognition is evaluative and self-directed, and therefore it is not used to assess others’ mental states, as it would be in the case of theory of mind. And yet there are ways in which a simulationist can appeal to noetic feelings arising in the context of an encounter with another to diagnose another’s state of mind. I am thinking here of a prescient remark by the developmental psychologist Peter Hobson who works on autism. He points to something that traditionally developing subjects feel when involved in social exchanges with people on the Autistic Spectrum. There is a social or emotional lack in AS individuals and we feel that lack in us when we are dealing with them. That is, we notice some lack of fluency and success in our interactions and we can use that feeling in us about how things are going as diagnostic of the other. In this case, our monitoring of the social situation and communicative success, leads to a feeling that we can use to tell that the other cannot maintain these social interactions. Here, we are using self-evaluative processes for attributive purposes. For another, rather elaborate example, one can listen to the results of music composed to capture the data of EEG recordings of epilepsy patients. This work by Stanford neurologist, Josef Parvizi, and musicologist, Chris Chafe, presents in sonic form three possible states of brain activity: that from the normal brain, that from a brain undergoing seizure, and that from the post-ictal state brain. When played music from these three stages, people without any training in EEG are immediately able to classify the sound recordings as belonging to the categories of normal or abnormal, and they usually report a feeling of discomfort when listening to the abnormal states of the brain. Here, it may be their ability to monitor appropriate timings and rhythms in their own brains that enables them to feel that something is wrong with the rhythms they are hearing from brains beginning to undergo seizure. The reach of self-monitoring metacognitive processes may go further than we think and it may be used in all sorts of unexpected ways. 734 BARRY C. SMITH Many have yet to be explored but Joelle Proust has supplied the philosophical foundations to enable us to do so. References Jonsson, F., Tchekhova, A., Lonner, P., and Olsson, M.J., (2005), ‘A Metamemory Perspective on Odor Naming and Identification’ Chemical Senses 30: 353–36. BOOK SYMPOSIUM 735