The Philosophy of Metacognition: Mental Agency and Self-Awareness

by Joëlle Proust

Does metacognition, i.e. the capacity to form epistemic self-evaluations about one's current cognitive performance, derive from a mindreading capacity, or does it rely, at least in part, on sui generis informational processes? In The Philosophy of Metacognition Joëlle Proust provides a powerful defense of the second position. Drawing on discussions of empirical evidence from comparative, developmental, and experimental psychology, as well as from neuroscience, and on conceptual analyses, she purports to show that, in contrast with analytic metacognition, procedural metacognition does not need to involve metarepresentations. Procedural metacognition seems to be available to some non-humans (some primates and rodents). Proust further claims that metacognition is essentially related to mental agency, i.e. cognitive control and monitoring. "Self-probing" is equivalent to a self-addressed question about the feasibility of a mental action ('Am I able to remember this word?').
"Post-evaluating" is a way of asking oneself whether a given mental action has been successfully completed "'Is this word the one I was looking for?"). Neither question need be articulated conceptually for a feeling of knowing or of being right to be generated, or to drive epistemic control. Various issues raised by the contrast of a procedural, experience-based metacognition, with an analytic, concept-based metacognition are explored, such as whether each is expressed in a different representational format, their sensitivity to different epistemic norms, and the existence of a variety of types of epistemic acceptance.

Genres: Philosophy

366 pages, Hardcover

First published November 20, 2013

Reviews

[Alina · Rating 5 out of 5]

Proust examines the phenomenon of metacognition and argues that there are two distinct forms of it, procedural and analytic, and that the latter depends on the former. Metacognition refers to any case of evaluative or informative awareness of other mental states. For example, we can have metacognition about memory; one form of such metacognition is that we can have a sense before we attempt to remember something of how likely it is that we will successfully remember it. As another example, we can have metacognition about perceptual or imaginative states that function as claims; we can have a sense of whether they are accurate, exhaustive, or agreeable to a social consensus, for example. While Proust's book includes 14 chapters, the last 4 chapters are purely applications of her view and so can be skipped; and among the first 14 chapters, I found only chapters 1-4 and 7 to be essential to articulating her view. As a whole, each chapter is quite stand alone, which makes it easy for a reader to learn a lot from just reading individual chapters of especial interest; but also this means the chapters aren't always coherent with one another, and there is much repetition of ideas between them.

Here are the essentials of Proust's view. There are two forms of metacognition. One is procedural; its representational format is non-conceptual and it is essentially contextual. So it doesn't involve generalizations and can't yield inferential knowledge. Its format is rather like a brute sense or feeling. Moreover, we do not have to access it consciously in order for it to guide and inform our mental actions. An example of procedural metacognition would be having an intuitive sense of how likely it is to remember something; a "tip-of-the-tongue" experience is a case of our having an inaccurate metacognition; we feel that we remember something, and yet we fail to produce it. Some non-human animals are capable of metacognition; apes, for example, may be given the choice of performing task A or B. A involves a lower reward than B, but B is much more difficult than A. Apes are able to prefer task A; this reflects they have procedural metacognition that they know how to perform A with more certainty than B.

The causal basis of procedural metacognition is that there are phenomenological dimensions of the mental action at hand (e.g., fluency: how forceful an affordance is, or how familiar we are with performing a certain action under a certain circumstance, so that it seems very easy and appropriate), and feedback systems for the control of action allow for such feelings to deliver information about the cognitive states that constitute that action. Procedural metacognition is thus possible only when we are already actively engaged in the mental action which it is evaluating. Only when engaged is it possible for us to receive such feelings of fluency or lack of. This implies that there is very special first-person access to metacognitive knowledge; we cannot have procedural metacognitive knowledge about other people's mental states, only our own.

Analytic metacognition refers to our explicit or deliberate thoughts or judgments about other mental states. The format of these is conceptual; it is potentially linguistic and can yield inferential knowledge. For example, we may explicitly ask ourselves whether it is likely for us to remember the name of a place; to estimate this, we might think about related cases of our failing or succeeding in remembering a name, and draw inferences from these probabilities for determining that of the task at hand. Non-human animals are likely incapable of analytic metacognition, because they lack linguistic capacities, which are requisite for the fine-grained conceptual format of this kind of metacognition.

The causal basis of analytic metacognition is similar to the causal basis of ordinary reasoning. We observe something and form judgments about it. So we can have analytic metacognition about a mental state without actually having that state (just as we can form judgments about a future bodily action that we have not yet initiated). We form analytic metacognition towards ourselves in the same way as we form it towards others; a friend might tell us what she is undergoing, and we might form a judgment about her mental state, and likewise, we might notice our own mental state, apprehend it in a conceptual manner, and form a subsequent judgment about it.

According to Proust, analytic metacognition often depends upon procedural metacognition. In order to form a conceptual judgment about a mental state, we must have some feeling about that state (procedural metacognition) that supplies basic information, which we work with and develop in forming the judgment.

All mental actions (e.g., imagining, remembering, reasoning, planning) involve a dimension of procedural metacognition, which enters at two stages: there is a moment that precedes the mental action, and then a moment that follows it. The procedural metacognition that precedes the mental action involves a prediction of how successful the mental action will be, or of what the mental action might be like according to other normative dimensions than success (depending on what sort of mental action it is -- more on this below). The procedural metacognition that follows the mental action involves an evaluation of that action according to relevant normative criteria.

The first moment of procedural metacognition will guide our mental actions. If procedural metacognition informs us that we are unlikely to remember something if we try, for example, we might decide to not attempt to remember it at all. Or, as another example, procedural metacognition might inform us that a certain belief would not cohere with the social consensus, and this might lead us to refrain from utilizing this belief in further reasoning.

In contrast, physical actions are rarely accompanied by procedural metacognition. This marks mental action out to be a unique kind. This goes against a traditional view on which mental and physical actions alike ought to be regarded as falling under the same kind, whose essential feature is that the action involves a sense of "trying." Proust shows that mental action involves something much more nuance than just trying; it involves metacognition.

Whenever we have procedural or analytic metacognition, this second-order state that involves an evaluation or prediction regarding another state heeds to certain norms (e.g., truth or accuracy; relevance; coherence with other knowledge; agreeableness to the social consensus of a relevant community). Sensitivity to these norms determines the content of an evaluation. According to Proust, these norms constitute the mental actions of which the procedural metacognition, which heeds these norms, necessarily precedes and follows. For example, metacognition regarding an imaginative state will not heed to the norm of truth or accuracy; but metacognition regarding a perceptual or judgment-based state will. In other words, imagination does not necessarily aim at truth, but belief does. Proust argues that different mental attitudes (e.g., belief, memory, imagination) are natural kinds, which are delimited on the basis of which norms towards which their accompanying metacognitions are sensitive.

As a whole, I was impressed by how precise and thorough Proust was in articulating her views, raising objections, and responding to them. It is so thorough that it can be difficult to read; I found myself losing the main thread and getting caught up in the ideas contained by objections. I was also impressed by how Proust deftly draws on empirical evidence across cognitive sciences to make her points. Her work might be taken as a model for doing interdisciplinary or naturalistic philosophy of mind. I'd highly recommend this book to anyone interested in how we can distinguish between different types of mental states (e.g., belief v. imagination), and in whether or how to characterize mental activities as actions.

Here are some of my side thoughts. I was initially drawn to this book after reading some of Proust's papers; it seemed to me I could use this concept of metacognition to make sense of a phenomenon I care about: how can it be that the contents of certain imaginings show up to us as part of reality, while the contents of others fail to? Perhaps procedural metacognition that accompanies the former obey norms typical of belief (e.g., truth), while that which accompanies the latter do not. I am not sure, however, how much progress this idea can lead me towards. It just opens the question: what accounts for the fact that an instance of procedural metacognition will draw upon one set of norms rather than the other? This is not a failing on Proust's part at all, but might reflect that my interests do not directly align with hers.

I really appreciate her giving me the vocabulary of how to make sense of an informational state that is neither explicit or deliberate, and may even be unconscious. It is not directly a matter of whether it is "personal" or "subpersonal"; rather it is a matter of representational format. Proust's discussion of conceptual v. non-conceptual formats of mental states may be dissociated from the topic of metacognition but is interesting and rich in itself. This helps me make sense of not just how we can build in as much information about a mental state as whether it reveals reality v. is detached from reality (and perhaps reveals make-believe) into that state, without our needing to think through these facts whatsoever -- this also helps me make sense of how we might understand affordances (e.g., information of what actions or behaviors a certain object we encounter affords, which amounts to the identity of that object for us).

[Grace · Rating 3 out of 5]

Warning: This book is incredibly thick and difficult to read. It is not really for the layman. It's also boring as hell, but very, very insightful.