What am I doing?
Remembering, reasoning, perceiving, deciding?
Monitoring depends on identifying the cognitive task actually underway.
Side 85
Metacognition
A study of cognition monitoring itself. Metacognition estimates what is known, detects uncertainty and error, chooses strategies, allocates effort and updates confidence after feedback.
monitor→estimate→evaluate→regulate→update
06metacognitive lenses
05confidence problems
05regulation moves
85Side
Metacognition begins with an estimate of one’s own cognitive state.
The system asks not only “what do I think?” but also “how certain am I, and should I trust this process?”
What internal evidence is available?
Fluency, conflict, familiarity, effort?
Metacognitive judgments often rely on indirect cues rather than direct access to correctness.
How likely am I to be right?
Confidence judgment.
Confidence compresses many internal signals into an actionable estimate.
Does confidence match performance?
Calibration.
Reliable self-monitoring requires confidence to track actual accuracy over repeated cases.
Should I continue, check or change strategy?
Control follows monitoring.
Metacognition matters because estimates alter behavior.
Confidence is useful precisely because certainty is unavailable.
It is a second-order judgment about the likely quality of a first-order judgment.
Strong internal evidence.
Can justify acting quickly when calibration is good.
Weak or conflicting evidence.
Can trigger information seeking or slower reasoning.
Ease can feel like truth.
Fast or familiar processing often inflates confidence even when accuracy does not improve.
Recognition can masquerade as knowledge.
Something feeling known does not imply that its details can be retrieved or defended.
Competing responses reduce certainty.
Internal disagreement can act as an error-warning signal.
Fast answers can feel stronger.
Response fluency influences confidence independently of objective correctness.
Good metacognition is not high confidence; it is calibrated confidence.
A calibrated observer is right about as often as their stated confidence predicts.
| Pattern | Confidence | Accuracy | Problem |
|---|---|---|---|
| Calibrated | Tracks performance | Matches expected rate | Minimal mismatch |
| Overconfidence | Too high | Lower than expected | Insufficient checking |
| Underconfidence | Too low | Higher than expected | Unnecessary hesitation |
| Poor discrimination | Similar across answers | Varies substantially | Cannot identify which judgments are weak |
| Good discrimination | Higher when correct | Lower when wrong | Useful self-monitoring |
Monitoring without control changes nothing.
Metacognitive regulation uses confidence and error signals to allocate time, attention and strategy.
Continue
Stay with the current strategy.
Appropriate when progress and confidence are both strong.
Slow down
Increase processing depth.
Useful when conflict or uncertainty rises.
Check
Seek external evidence.
Verification is especially valuable when consequences are high.
Switch
Change strategy.
Persistent difficulty can indicate that effort is being applied to the wrong process.
Stop
Recognize diminishing returns.
Good regulation includes deciding when more thought is unlikely to improve the answer.
Metacognitive errors often arise because the mind monitors proxies.
Ease, familiarity and coherence are useful signals on average, but each can become misleading.
Exposure feels like mastery.
Rereading can increase familiarity without increasing retrievable knowledge.
Past uncertainty disappears.
Once an outcome is known, the original alternatives become harder to reconstruct.
Intervals are too narrow.
People can underestimate how much uncertainty surrounds their estimates.
Search stops too early.
A coherent first explanation can suppress consideration of alternatives.
Bias is easier to see in others.
Knowing about a bias does not guarantee accurate detection in one’s own reasoning.
Doubt can also be excessive.
Metacognition fails when correct judgments are repeatedly overridden by unjustified uncertainty.
Metacognition can be externalized.
Records of confidence, predictions and errors make self-assessment testable rather than purely introspective.
Predict
State expected performance before seeing the result.
Commit confidence
Record how certain the judgment feels.
Resolve
Compare the judgment with external evidence or outcome.
Diagnose
Ask whether error came from knowledge, attention, strategy or confidence estimation.
Update
Change future confidence or strategy based on recurring error patterns.
Metacognitionmonitoring and control
Judgments of Learningpredicted versus actual recall
Calibration Researchconfidence and accuracy
Error Monitoringconflict, feedback and adjustment