Metacognition – Cognitive and Social Dimensions (I)

“Metacognition is a fundamental characteristic of human cognition. Not only do we have cognitive activities but it would seem that they can apply to themselves: we have cognitions about cognition. […] How we answer simple, factual questions involves more complexities than one would expect. We should wonder, for instance, why we keep searching for an answer we do not find immediately, how we know we do not know something […], why we start searching at all […] or what makes us confident in some response we have been able to come up with […]. All the above points are somehow related to metacognition, that is, to the more or less explicit knowledge we have of the way our memory and our mind work. […] The social world requires that we make a staggering number of decisions about the extent and nature of our own knowledge and beliefs as well as decisions about the extent and nature of other people’s knowledge and beliefs. […] Judgments about the trustworthiness of knowledge are just one among many different types of metacognitive judgments. There is virtually no end to the number of cognitive assessments that could be made about the thoughts, feelings, and beliefs of the self and others. What does she think, why does she think it, when did she start thinking it, what would get her to stop thinking it? Do I know the answer to your question, when and from whom did I hear the answer, would I have remembered it if I had heard it, what will help me to remember it, and can I be confident that my memory will be accurate? Human action and interaction depend in crucial respects upon our general success in these rough metacognitive waters. […] Because social behavior and metacognition rely on general cognitive processes, metacognition must also rely on approximations, limited capacity processes, arguable heuristics, intrusion of naive theories, and so forth. The interesting question is whether this explains why our judgments about others and ourselves cannot be better and how they go wrong. The various chapters in this book analyze simple cognitive tasks and a number of problematic social situations. The cognitive analysis seems to converge on a small number of principles that may not be completely understood yet but that seem to agree with what has been observed in social cognition.”

…

This book, from which the above quotes are taken, is not nearly as good as I thought it might be – I was disappointed with the book, which is part of why it got only two stars on goodreads.

I got a Kindle e-reader as a birthday present recently, and I read this book (as well as the habituation book and the Christie novels already mentioned) using that device. It is very hard, bordering on impossible, to highlight stuff when using the Kindle, which means that I’m actually really not a big fan of it. Highlighting and note-taking are very important aspects of reading for me, and when I can’t do those things (or if those things are potentially possible yet prohibitively difficult and time-consuming, as is the case here) reading becomes a very different, and much less enjoyable, experience. Basically I’ve already decided that I’ll only use the device when I have very few relevant alternatives; in general terms I’d say for example that I’d much rather be carrying around a book, two highlighters (yellow and green) and a pen (…stuff I usually carry around with me on occasions where I leave my place and I know that I might have an opportunity to read while I’m away, e.g. in a bus) than this device and nothing else.

I read the book last week, but before writing this post I felt like all I had to show for it were some bookmarks which I added while reading it on the Kindle (bookmarks are much easier to add than are highlights or notes, at least on my device); this was part of why I postponed blogging it and why I didn’t blog it soon after I got back online – I knew I’d have to basically read large parts of the book again in order to blog it, because the key sections to add to a post like this were not separated from the rest of the text the way they usually are when I’m about to blog a book, and given that the book actually isn’t really all that great in the first place it seemed like a lot of work. But here we are.

The book has 12 chapters of varying quality – here’s what the chapters deal with: 1) From Social Cognition to Metacognition, 2) Illusions of Knowing: The Link between Knowledge and Metaknowledge, 3) Rapid Feeling-of-Knowing: A Strategy Selection Mechanism, 4) The Feeling-of-Knowing as a Judgment, 5) Knowing Thyself and Others: Progress in Metacognitive Social Psychology, 6) Social Influence on Memory, 7) Beliefs, Confidence and the Widows Ademoski: On Knowing What We Know about Others, 8) Social Judgeability Concerns in Impression Formation, 9) The Consciousness of Social Beliefs: A Program of Research on Stereotyping and Prejudice, 10) Protecting Our Minds: The Role of Lay Beliefs, 11) The Metacognition of Bias Correction: Naive Theories of Bias and the Flexible Correction Model, and 12) Correction and Metacognition: Are People Naive Dogmatists or Naive Empiricists during Social Judgments?

The book is an interdisciplinary work in the sense that it includes research from both social psychology and cognitive psychology. It’s to some extent already a slightly ‘old book’; it was published in 1998. A lot of stuff has happened since then in related fields, and presumably in this field as well, but it looked like an interesting book so I decided to give it a shot anyway.

I think that I in general liked the second half better than the first half. Below I have covered stuff from the first five chapters of the book.

…

“Research has shown that human subjects can express a number of more or less valid judgments about various aspects of experimental cognitive tasks […] Research also suggests that these judgments are more or less directly related to the optimization of cognitive processes, determining how long one will search for an answer, keep rehearsing when learning, etc. […] Most cognitive processes are normally accompanied by metacognitive operations that supervise and control various aspects of these processes. Thus, when we make an appointment, we often have to take precautions not to miss it, and these precautions depend on our assessment of their effectiveness as well as on our assessment of the chances of missing the appointment if these precautions are not taken. […] In attempting to retrieve a piece of information from memory, we can often tell whether it is indeed in store and worth searching for, and when we finally do succeed in retrieving the solicited information, we can generally assess the likelihood that it is the correct information. What is important about the subjective feelings that ensue from monitoring operations is that they generally have measurable effects on our behavior […] For example, the stronger my feeling of knowing [FOK] about an elusive name, the more time I am likely to spend searching for it before giving up […] One can conclude from this line of research that the accuracy of feeling-of-knowing judgments is well above chance yet “far from perfect” […] In addition to prediction of accuracy, feeling-of-knowing has been shown to be correlated with search duration […] when subjects [experience] stronger feelings of knowing, they [search] longer. […] Several researchers exploring the feeling-of-knowing phenomenon have speculated about the underlying mechanisms that are involved in this process. One viewpoint that has received some attention is the trace access hypothesis. This presumes that subjects have partial access to, and are able to monitor some aspects of, the target item during feeling-of-knowing judgments […] An alternative viewpoint to the trace access hypothesis is the cue familiarity hypothesis. This position argues that feeling-of-knowing judgments rely on the familiarity of the cues in the questions themselves”

So which one is right? Here are some relevant comments from chapter two:

“Although subjects are generally successful in monitoring the availability of inaccessible information, this is not because they have privileged access to the underlying memory trace. Rather, the FOK monitors the accessibility of partial information regardless of its correctness, and its accuracy derives from the fact that most of the information that comes to mind is correct. In this sense the accuracy of metamemory can be said to constitute a by-product of the accuracy of memory itself. […] results suggest that the ease with which information comes to mind can serve as a valid cue for the accuracy of that information, and that FOK judgments do in fact monitor ease of access. The reliance on ease of access, then, can also contribute to FOK validity in predicting memory performance. In this manner FOK judgments can function in two capacities, as predictors of the future recognition of the full target (prospective monitoring), and as postdictors of the accuracy of the partial information that has already been accessed (retrospective monitoring). […] The accessibility account of the FOK is […] consistent with findings in social psychology indicating that subjective experiences and social judgments are affected by the fluency with which stimuli are processed, and by the ease with which information comes to mind. These findings too suggest that under some conditions judgments are based on a nonanalytic, inferential process rather than on direct access to the judged attribute, and that people are not always capable of monitoring the validity or relevance of the associations that come to mind”

The authors of chapter 3:

“There is empirical evidence supporting both views [the cue-familiarity hypothesis and the trace access hypothesis]. Thus, one can say that both hypotheses are correct (or wrong). However, we believe that this must be qualified by stating that the accuracy of the hypothesis depends on the methodology one uses to investigate feeling-of-knowing. If the researcher is investigating a feeling-of-knowing process that occurs after a memory retrieval failure, then it is likely that the feeling-of-knowing that is assessed is a by-product of the retrieval attempt. When the methodology is concerned with strategy selection, feeling-of-knowing turns out to be based solely on the cues of the question. […] It seems to us that there is a more central function to feeling-of-knowing than just rating questions after the fact, specifically, to regulate strategy selection. It is important to emphasize that rapid feeling-of-knowing, assessed before retrieval failure, serves this function […] when presented with a general knowledge question, a person can decide, based on this initial, rapid feeling-of-knowing, whether to try to retrieve the answer from memory or to use another strategy, such as reasoning or looking up the answer in a textbook. Similarly, after presentation of a novel math problem (e.g. 26 x 43), it would be this initial rapid feeling-of-knowing which would help a person decide either to retrieve or to calculate the answer. If the initial feeling-of-knowing for the problem is high, the person tries to retrieve the answer; if the feeling-of-knowing is low, the person computes the answer. Note that it is possible for the person to choose to compute even if the answer is stored in memory, and conversely, it is possible to choose to retrieve when the answer is not known. […] the rapid, preliminary feeling-of-knowing stage necessarily occurs before the retrieval attempt, because it is a product of parsing the question. Consequently, feeling-of-knowing is a natural precursor to the retrieval process. […] trace access theories can explain a positive feeling-of-knowing when access is blocked or delayed, but they do not easily explain illusions of knowing: cases in which a high feeling-of-knowing rating is – wrongly – assigned to a question while the correct response is totally unknown. If the feeling-of-knowing is to be generated by a partially accessed trace, it is difficult to imagine how a feeling-of-knowing can be generated when the subject does not know the answer.”

“The feeling-of-knowing may be affected by numerous variables. Cue familiarity, domain familiarity, contextual or normative information, recency of use have all been found to influence the feeling-of-knowing rating […]. It can also be affected by episodic elements like remembering the context in which the response might have been acquired, and even by more social or normative elements like estimates of the probability that the response will be known by people the subject usually interacts with […]. Most of these elements are correlated, so it is not easy to determine which has the most important weight.”

“The problems involved in assigning a rating to a dimension like accessibility also appear theoretically similar to the problems involved in Clore’s (1992) description of how mood can be used as information. […] In both cases a single continuum is involved and a single intensity value is returned. Just as various events have an influence on our mood, various inferential cues condition accessibility. In both cases this “confounding” of influence sources is usually acceptable from the subjects’ point of view because there is a general correlation between the sources, but both cases involve a potential attribution problem that is directly related to this “confounding.” In the case of the feeling-of-knowing, high accessibility may be due to a number of factors, some of them related to the existence of an appropriate response, some related to extraneous considerations (inferential aspects), some related, even, to elements of information that do not constitute a proper answer.”

“empirical social psychology provides ample evidence that people make use of metacognitive assessments, whether they are aware of those assessments or not. For instance, it has been found that people search the facial expressions and non-verbal behaviors of others in order to figure out how others perceive a given situation, and this helps them to determine how they themselves feel about that situation […] In research on social influence, it is found that people almost always revise their own judgments to make them more similar to the judgments of others […], especially when people have (even the slightest) reason to question the credibility of their own knowledge […]. Some social psychologists have argued that we do not know what we think until we know what other people think […]. Even when our own thoughts and feelings are relatively clear to us, we still estimate and adjust for the thoughts and feelings of others in order to best follow our own interaction goals […]. All of these cases involve metacognitive judgments, that is, cognitive assessments that are about the cognitions of the self.”

“That situationally given “data” are interpreted in terms of participants’ extant theories or conceptions is especially clear in recent research by Strack and his colleagues on the subject of reconstructive memory […] people hold intuitive theories about what sorts of things they are likely to remember and what sorts of things they are likely to forget, and these notions play an important role in confidence judgments about whether or not a stimulus has appeared previously. For instance, students may be more certain about the fact that their professor was not wearing a sombrero during the last class period than about the fact that their professor was not wearing a pair of glasses […]. This is because people make important metacognitive judgments about whether they would have remembered a particular occurrence, even when they do not remember the actual occurrence […]. Strack and Bless (1994) have demonstrated further that people possess very subtle assumptions that numerically distinctive categories will be remembered better than numerically non-distinctive categories. As a result, people are more confident in their decision that a given event did not occur when that type of event would have been infrequent than when it would have been a frequent occurrence. Other work indicates that people are also more confident in their memories when the previously exposed set of items was small rather than large […] in attempting to remember one’s past attitudes, feelings, or abilities, one often begins with present attitudes or mental states and makes adjustments to those on the basis of implicit theories about personal stability and change. Because people generally believe that factors such as intelligence and political party affiliation are stable entities, they tend to overestimate the extent to which their present capacities and affiliations are the same as those they possessed in years gone by (Ross, 1989).”

“In sum, it appears that the assessment of self-knowledge is derived from whatever specific information is perceived as relevant to the task at hand or to the general processes by which knowledge is achieved. Such information may consist of subjective feelings and experiences as well as cultural theories and category-based information about people in certain social groups. Subjective feelings are interpreted in terms of prior cognitive notions that may be momentarily activated in the individual’s memory. Such notions may include long-standing beliefs as to how knowledge states are represented in momentary sensations (e.g. that a sense of fluency or familiarity indicates the validity of a hypothesis one was entertaining) or naive theories about the workings of memory (e.g. that one may have better memory for unusual or infrequent events). Metacognitive theories engender expectations about cognitive performance, and these expectations are capable of impacting actual judgments, memories, and intellectual performances through social psychological processes of “expectancy confirmation” or the “self-fulfilling prophecy” (Jones. 1990). […] Social metacognitive phenomena of the type we have been reviewing have an obvious bearing on the accuracy of people’s self-perceptions and judgments. To the extent that people’s attributions about their momentary states, such as causal explanations pertaining to their mood, are false, any inferences based upon those attributions, such as judgments of global life-satisfaction, are likely to be inaccurate. If, however, correct attributions are made, then the inferences derived from those attributions have a greater likelihood of being accurate. […] people often accept their own perceptual experiences as unquestionably true. […] “feelings,” “sensations,” and “experiences” are [however] heavily influenced by interpretive and inferential processes based upon potentially malleable lay theories, and so the metacognitive inferences drawn on the basis of those feelings may often lack veridicality, despite the directness of the subjective experience […] subjective experiences or unconscious memories may often give rise to faulty judgments”

“The metacognitive processes whereby momentary feelings or sensations are identified and inferences are drawn from these identifications are highly dynamic. Social psychological research indicates that the extent, directionality, and duration of metacognitive inferences should be affected by the individual’s motivation and cognitive capacity. When cognitive capacity or processing motivation are low, metacognitive processes are likely to be limited and early interpretations are likely to be anchored or “fixated” upon. However, when information processing capacity and motivation are high, a number of metacognitive interpretations may be entertained. This multiplicity of conceptions may occasionally produce confusion and undermine the individual’s sense of secure knowledge. Thus, recent research suggests that conditions such as time pressure or environmental noise tend to limit a person’s cognitive capacity and motivation and consequently reduce the number of interpretative hypotheses that he or she considers, increasing at the same time his or her subjective sense of judgmental confidence. By contrast, conditions that increase a person’s capacity or processing motivation not only increase the number of hypotheses generated, but they concomitantly reduce his or her confidence”

“Divining other people’s knowledge is essential to functional social interaction. […] But how does one go about assessing another persons’ knowledge? Research has identified several informational sources that people may use for that purpose. First, people may use their own knowledge or opinions as a basis for making projections about the knowledge of others. Second, people may use “actuarial” knowledge or statistical information about what people in general think or what a specific group of people think about a given topic. Third, people may attend to a given person’s individualized reactions to a stimulus. All three of these informational sources have figured in social psychological research on metacognitive assessments of other people’s minds. […] people are more likely to impute a piece of knowledge to others if they possess this knowledge themselves than if they do not, and people are likely to overestimate the commonality of their own knowledge. […] In a wide variety of domains, it appears that people exaggerate the commonality of their own attitudes, feelings, and behaviors […] Subjective “feelings,” “sensations,” and so-called “inner experiences” are relevant and accessible only with regard to one’s own assessment of knowledge. Because subjective events of this type can be subject to misidentification (or misattribution), however, they do not necessarily confer an accuracy advantage to the assessment of self-knowledge as compared with the assessment of others’ knowledge.”

“Rather than using self-knowledge as a standard from which other people’s beliefs are inferred, perceivers sometimes make use of statistical or “base rate” information […] Several kinds of research suggest that people generally underutilize category-based, actuarial information in the face of even the slightest amount of individuating information […]. Nevertheless, it has been found that base rate information is taken into account by lay perceivers when it is perceived as specifically relevant to the judgment being rendered […], when the diagnosticity of the individuating information is diminished in some way […], or when the categorical information is congruent with the perceiver’s information-processing goal or task orientation […]. Furthermore, research on social stereotypes demonstrates that category-based information is utilized when it is readily accessible in memory […], when the perceiver is highly motivated to achieve cognitive closure […], or when the perceiver possesses an especially low degree of accuracy motivation […]. Under these diverse sets of circumstances, it should be expected that people will rely on actuarial information to draw inferences about the mental states of others.”