Metacognitive Laziness
Metacognitive laziness is the tendency to stop monitoring and evaluating your own thinking when an AI is available - accepting its output without checking it against your own judgment. You stop asking "do I understand this?" because the answer no longer seems to matter.
Metacognition is the supervisor process: the part of you that watches your own thinking and asks whether it’s working. Am I understanding this? Is this approach failing? Should I re-read that? It’s the skill that separates experts from people who merely have experience.
The research term “metacognitive laziness” was coined for what happens to that supervisor when ChatGPT enters the room: it goes home early. In Fan et al.’s 2024 study, learners using ChatGPT got better task scores while doing measurably less planning, monitoring, and self-evaluation - better output, worse learning. The grade went up; the student didn’t.
What it looks like in development
The supervisor question for a developer is “do I understand this change?” Metacognitive laziness is the state where the question stops being asked, because the practical answer - the tests pass and the agent seemed confident - arrives faster than the honest one.
It compounds with automation complacency: first you stop checking the AI’s reasoning, then you stop checking your own. The Microsoft + CMU survey of knowledge workers caught the pattern at scale - confidence in the AI predicted less critical thinking, while confidence in one’s own skills predicted more (sources in the research review).
The countermeasure is annoyingly simple: re-ask the supervisor questions on purpose. Before accepting a diff - could I have written this? could I explain it? Cheap to ask, and the asking itself is the exercise.
FAQ
Where does the term "metacognitive laziness" come from?
From learning-science research on ChatGPT in education - notably a 2024 study by Fan and colleagues in the British Journal of Educational Technology titled "Beware of metacognitive laziness." Learners with ChatGPT improved their essay scores but engaged less in the self-monitoring behaviors (planning, checking, evaluating) that produce durable learning.
How is metacognitive laziness different from regular laziness?
Regular laziness skips the work. Metacognitive laziness does the work but skips the supervision of it - you produce the output without ever evaluating whether you understand it or whether it is right. It is more dangerous precisely because the work still ships, so nothing visible flags the problem.
What does metacognitive laziness look like when coding with AI?
Accepting a diff after a skim that checks style but not logic. Re-prompting when something fails instead of forming your own hypothesis. Realizing in code review that you cannot explain why your own change works. The tell is the absence of the inner question "wait, do I actually buy this?"