Metacognition

Metacognition refers to awareness of one's own cognitive processes—that is, thinking about how one thinks.

The term metacognition was coined by American developmental psychologist John Flavell in 1976, referring to cognition about cognition—that is, an individual's awareness, monitoring, and regulation of their own thought processes.

Having one AI generate prompts to pass to another AI for execution is described by the author as the first aha moment in AI usage for many people.

The concept of metacognition was proposed by psychologist John Flavell in the 1970s.

The concept of metacognition originates from cognitive science, where it originally refers to humans' ability to be aware of their own thought processes.

Treating LLMs as genuine intelligence leads to overestimating their reliability, mistakenly trusting their reasoning capabilities, and lacking vigilance regarding the hallucination problem.

The Dunning-Kruger effect describes how people with insufficient competence tend to overestimate their own ability, while those with greater competence tend to underestimate themselves.

A model with metacognition will proactively express uncertainty and suggest calling external tools when it exceeds the boundaries of its knowledge, rather than generating hallucinated answers.

Research shows that individuals with strong metacognitive ability are better at calibrating their confidence.

Metacognition is a core concept in cognitive science, referring to an individual's ability to monitor, evaluate and regulate their own thought processes.

Medical fine-tuning changes the model's style of expressing itself, not its metacognition; after fine-tuning the model is less overconfident, but the internal gauge for distinguishing what it knows from what it does not has not become more accurate.

What Krouse emphasizes—"how to ask good questions" and "how to adjust strategies after failure"—essentially falls within the scope of metacognition.

Metacognition refers to an individual's ability to be aware of and regulate their own cognitive processes, including planning learning paths, monitoring comprehension, and evaluating the effectiveness of strategies.

In educational psychology, metacognitive ability is widely regarded as the key variable distinguishing efficient learners from inefficient ones.

Metacognition usually comes from the accumulation of feedback in long-term practice and is very difficult to acquire through reading or short-term study.

Structured guidance helps LLMs maintain clearer stage divisions during the inquiry process (what to do first, what to do next, and when to go back and verify), compensating for the model's shortcomings at the metacognitive level.

The experiments found that on most test instances the optimized schedule brought substantial performance gains, but in certain problem environments, an overly aggressive early elimination strategy instead incurred higher cost.

Subsequent research shows that the correlation between metacognitive level and academic achievement is often higher than that of IQ.

元认知（metacognition）是指"对自身认知过程的认知"，由发展心理学家约翰·弗拉维尔于1970年代提出，如今在AI协作场景中变得尤为关键。

The value of recording mistakes into an 'error notebook' lies in metacognition; documenting the structure of 'I thought X, but the reality was Y, because Z' helps abstract and attribute errors.