Metacognitive knowledge

A student closes a book and says, with complete sincerity, "I learn best by reading it again."

The statement sounds like self-knowledge. It might be. It might also be a theory about learning that has never been tested against the student's actual memory.

This distinction is the beginning of metacognitive knowledge.

Metacognitive knowledge refers to stored knowledge and beliefs about cognition: what a person believes about themselves and other people as thinkers, about the demands of a task, about available strategies, and about how these elements interact. The construct has roots in developmental and educational psychology, especially John Flavell's work on metacognition. It is not the same thing as the moment-to-moment monitoring of a particular answer, and it is not the same thing as changing behaviour in response to that monitoring.

A person can know a great deal about learning and still make a poor learning decision. They can also hold confident metacognitive beliefs that are wrong.

Knowledge about the learner is only one part

Flavell's classic account distinguished knowledge about persons, tasks and strategies. The labels are simple, but the interactions are not.

Person knowledge includes beliefs about one's own cognitive tendencies and those of others. Someone may know that they lose track of arguments when reading late at night, remember spatial layouts more easily than names, or become overconfident in familiar domains. Some of this knowledge can be well grounded in repeated experience. Some can become identity language that exceeds the evidence, as when a preference is treated as a fixed learning style.

Task knowledge concerns what different cognitive demands require. Memorising vocabulary, evaluating an argument, learning a motor sequence and debugging code are not the same learning problem. A strategy that helps one task may be poor for another.

Strategy knowledge concerns possible cognitive actions and their conditions of use: retrieval, spacing, elaboration, checking, summarising, seeking feedback, comparing examples, or deliberately slowing a decision.

Metacognitive knowledge becomes useful when these three kinds of knowledge meet. "This is a difficult inference task; I tend to accept fluent explanations too quickly; I need an external check" is richer than "I know myself."

Knowing that, knowing how, and knowing when should not be collapsed

Educational writing sometimes speaks of metacognitive knowledge as if possessing a named strategy were enough. It is not.

A learner may know that retrieval practice can support retention, yet not know how to construct a useful retrieval attempt. They may know how to create practice questions but not know when retrieval is preferable to rereading, worked examples, feedback, or initial instruction. They may know when a strategy is useful but fail to apply it because the environment rewards speed or because the easier method feels more fluent.

This matters because metacognitive knowledge has both declarative and conditional dimensions. Knowledge can be technically accurate yet too abstract to regulate behaviour.

The distinction also protects the construct from becoming a list of study tips. Metacognitive knowledge is not the possession of approved advice. It is a structured model, often incomplete, of cognition in context.

Beliefs about cognition can be wrong

The word knowledge can mislead if it is heard as "true information." In this literature it often includes beliefs that a person holds about cognition whether or not those beliefs are accurate.

A learner may believe that familiarity equals mastery. An expert may believe that years of experience make bias irrelevant. A manager may believe that multitasking improves their productivity because task switching feels rapid. A person may infer from ease of processing that an explanation is trustworthy.

These beliefs can regulate behaviour very effectively while being false.

That is why strong research on metacognition cannot rely only on asking people what they think they know about their own minds. Self-report reveals beliefs and reported strategies. It does not automatically validate those beliefs against behaviour or performance.

The practical question is therefore twofold: what does the person believe about cognition, and how well does that belief predict or improve what happens in the task?

Metacognitive knowledge is not metacognitive monitoring

The neighboring entry Metacognitive monitoring concerns online judgments such as confidence, judgments of learning, feelings of knowing and estimates of current performance.

Metacognitive knowledge can inform those judgments, but the two are not interchangeable.

A reader might know in general that dense legal language creates comprehension errors for them. That is metacognitive knowledge. Halfway through a particular contract, noticing that they can no longer explain the previous paragraph is monitoring. Returning to the paragraph, annotating it or asking for clarification is control.

The sequence can also run in reverse. Repeated monitoring experiences can update knowledge. A learner who repeatedly discovers that rereading produces high familiarity but weak delayed recall may revise their theory of how they learn.

The components form a system, but retaining the distinctions makes failure diagnosable. If a learner knows the right strategy but does not deploy it, the problem is different from not knowing the strategy exists.

Knowledge about strategy effectiveness needs evidence, not fashion

Metacognitive knowledge is often culturally supplied. Teachers, parents, workplaces, social media and educational software all teach implicit theories of learning.

Some are well supported. Some are plausible but weakly tested. Some survive because they are easy to communicate.

The learning-styles idea is a useful caution. People plainly have preferences, histories and areas of strength. The stronger claim that instruction should be matched to a person's supposed sensory learning style to improve learning is a different proposition and requires direct evidence. Metacognitive knowledge becomes less reliable when preference, identity and causal effectiveness are folded together.

The same problem appears in productivity culture. A strategy can feel personalised, efficient or motivating without being the best explanation of learning gains.

For an Institute knowledge library, the standard should therefore be stricter than "people find this useful." The relevant question is what kind of evidence supports the belief, for which task, over what retention interval, and with what comparison condition.

Development changes what people can know about their own cognition

Flavell's early work was developmental. Children gradually acquire richer knowledge about memory, comprehension, communication and strategy use. They become better able to represent differences between task demands and to understand that internal states can be monitored and managed.

Development, however, does not terminate in perfect adult self-knowledge.

Adults remain vulnerable to miscalibration, fluency effects and incomplete models of their own cognition. Expertise can improve metacognitive knowledge by providing a richer model of task structure, but expertise can also create blind spots when familiar cues are mistaken for certainty.

This is one reason metacognitive education cannot be treated as a childhood remedial topic. The relevant knowledge changes with domain, technology and role. A surgeon, teacher, software developer and policy analyst confront different uncertainty structures.

What develops is not a single inner dashboard. It is a repertoire of models about cognition that can become more differentiated and better tested.

Measurement is harder than a questionnaire makes it look

Metacognitive knowledge is often assessed through interviews, questionnaires, think-aloud protocols, strategy inventories and task-specific prompts. Each method captures a different slice.

A questionnaire can efficiently sample declared beliefs, but it is vulnerable to social desirability, memory error and differences in how respondents interpret general statements. Think-aloud methods can expose strategy knowledge in context but may alter task performance and are demanding to code. Performance tasks show what people do, but behaviour alone does not reveal the belief that produced it.

Measurement therefore needs construct alignment. If the research question is whether students understand when retrieval is useful, a broad "I use good learning strategies" item is too far away from the claim.

Contemporary metacognition research adds another caution. Measures of metacognitive ability can have strong dependencies on task performance and uneven test-retest reliability. Those findings concern particular performance-based measures, not metacognitive knowledge as a whole, but they reinforce the broader principle: measurement properties must be demonstrated rather than assumed.

Metacognitive knowledge can remain inert

One of the most important boundaries in this field is the knowledge-action gap.

Knowing that interruptions impair concentration does not turn off notifications. Knowing that a decision is emotionally loaded does not guarantee a pause. Knowing that retrieval is useful does not make retrieval pleasant. Knowing that an AI answer should be checked does not create time, expertise or institutional permission to check it.

Metacognitive control depends on goals, incentives, effort, emotion and available actions. Self-regulated learning is broader still, incorporating motivation, behaviour and context.

This matters for interventions. If poor performance is caused by time pressure, lack of feedback or inaccessible resources, teaching more metacognitive facts may have little effect. The person may already know what good regulation would require.

The Institute should resist turning every failure of practice into a deficit inside the individual.

The AI environment changes the object of metacognitive knowledge

Generative systems add new things that people need to know about cognition and about tools that interact with cognition.

A user may need a realistic model of what language-model fluency does and does not indicate, how automation can change effort, when external assistance reduces useful practice, and when checking the tool requires expertise the user does not possess.

There is established evidence for many underlying metacognitive phenomena, including confidence, monitoring and learning strategy use. The broader proposition that AI literacy should include metacognitive knowledge about human-tool interaction is an Institute synthesis.

It should be treated as a design and research question, not as a settled scientific fact.

The strongest version of the claim is modest: when technology changes the cues available to a learner and the actions they can outsource, the person's model of how learning and judgment work may need updating.

From individual belief to institutional knowledge about knowing

Institutions also carry assumptions about cognition. A school timetable encodes beliefs about attention. An assessment system encodes beliefs about what counts as learning. A workplace policy encodes beliefs about interruption and productivity. An AI procurement decision can encode beliefs about which cognitive tasks are safe to externalise.

Calling these institutional assumptions metacognitive knowledge is an analogy, not a direct application of a human psychological construct. Organisations do not possess a unitary mind.

Still, the analogy is useful when kept labelled. Institutions can make their assumptions about human cognition explicit, compare them with evidence, and revise practices that rest on outdated models.

At civilizational scale, this becomes a question of whether cultures update their theories of learning as fast as they update the technologies that change how learning occurs.

Metacognitive knowledge can be taught, but not as slogans

Instruction can make metacognitive knowledge more explicit. Teachers can explain why a strategy works, model how task demands differ, compare predictions with outcomes, and ask learners to articulate not only what they did but why they chose it.

The important word is conditional.

A learner does not need a larger bag of named techniques so much as a better model of the conditions under which techniques are useful. Retrieval can strengthen memory, but it is not the first move when nothing has been encoded. Detailed worked examples can support novices, while experts may need less scaffolding. A comprehension strategy that helps with argumentative prose may not transfer unchanged to mathematical proof.

This conditionality also limits the appeal of universal productivity systems. Metacognitive knowledge becomes mature when it can represent exceptions, tradeoffs and uncertainty rather than merely repeat rules.

For institutions, the educational task is therefore not to publish a list called "how the brain learns." It is to create recurring opportunities for people to test beliefs about learning against outcomes and revise them.

Expertise can deepen metacognitive knowledge and distort it

Experts know structures novices cannot yet see. They can recognize diagnostic cues, anticipate common failure modes and select strategies with less explicit deliberation. This richer task model can support sophisticated metacognitive knowledge.

Yet expertise changes experience. Procedures that once felt difficult become fluent. Intermediate steps become compressed. Experts can forget which assumptions a novice does not possess, producing the familiar problem sometimes called the curse of knowledge.

That creates a paradox for teaching and leadership. The person with the best domain model may be poorly positioned to remember what the task feels like before that model exists.

Metacognitive knowledge about other minds therefore matters alongside self-knowledge. Good instruction requires an estimate of what another person knows, what they can infer, and which cues will mislead them.

The evidence base does not justify a simple claim that experts are either better or worse at metacognition. It suggests a more precise question: better at monitoring and regulating what, under which task conditions, using which cues?

What the evidence does not show

Metacognitive knowledge is not guaranteed to be accurate simply because it is introspective.

Knowing the name of an effective strategy does not show that a person can execute it, choose it at the right time or persist with it.

A self-report of learning preferences does not establish that matching instruction to those preferences improves learning.

Metacognitive knowledge is not equivalent to confidence, online monitoring, executive control or self-regulated learning.

Evidence that a strategy supports learning in one task does not establish universal superiority across tasks, learners and goals.

And the established psychological construct does not by itself validate an Institute programme, an AI-readiness curriculum or an institutional score for "knowing how to know."

A more useful question than "what kind of learner am I?"

The familiar question asks for an identity.

The stronger metacognitive question asks for a model that can be tested.

What is this task asking of cognition? Which cues am I using to judge whether I understand it? Which strategies have evidence under conditions like these? What tends to mislead me? What would count as evidence that my theory about my own learning is wrong?

At the individual level, metacognitive knowledge makes cognition available for deliberate choice.

At the educational level, it can help learners distinguish ease from learning and preference from effectiveness.

At the institutional level, it invites explicit scrutiny of the theories of mind built into systems.

Its value is not that it makes people certain about themselves.

It gives them a model of cognition that can be revised when reality disagrees.

Research references

1. Flavell JH. Metacognition and cognitive monitoring: A new area of cognitive-developmental inquiry. American Psychologist. 1979;34(10):906-911. https://doi.org/10.1037/0003-066X.34.10.906
2. Veenman MVJ, van Hout-Wolters BHAM, Afflerbach P. Metacognition and learning: conceptual and methodological considerations. Metacognition and Learning. 2006;1:3-14. https://doi.org/10.1007/s11409-006-6893-0
3. Fleming SM. Metacognition and Confidence: A Review and Synthesis. Annual Review of Psychology. 2024;75:241-268. https://doi.org/10.1146/annurev-psych-022423-032425
4. Rahnev D. A comprehensive assessment of current methods for measuring metacognition. Nature Communications. 2025;16:701. https://doi.org/10.1038/s41467-025-56117-0
5. Koriat A. Monitoring one's own knowledge during study: A cue-utilization approach to judgments of learning. Journal of Experimental Psychology: General. 1997;126(4):349-370. https://doi.org/10.1037/0096-3445.126.4.349
6. Nelson TO, Narens L. Metamemory: A theoretical framework and new findings. Psychology of Learning and Motivation. 1990;26:125-173. https://doi.org/10.1016/S0079-7421(08)60053-5
7. Metcalfe J, Kornell N. A Region of Proximal Learning model of study time allocation. Journal of Memory and Language. 2005;52(4):463-477. https://doi.org/10.1016/j.jml.2004.12.001