How does treating cognition as computation reshape education and work?
This explores what happens when we adopt the premise that thinking is a kind of computation — and how that premise quietly reshapes what learning and knowledge-work are for, once machines can run the computation too.
This reads the question as being less about a technical claim and more about a worldview: if cognition is computation, then it can be offloaded, automated, and measured like any other process — and the corpus shows that move cutting in two directions at once. The most concrete evidence on the cost side comes from a four-month EEG study where brain connectivity scaled *down* with AI reliance: heavy LLM users showed the weakest neural engagement, worst memory retention, and couldn't even recall their own recent work Does AI assistance weaken our brain's ability to think independently?. If thinking is computation you can hand off, the muscle you stop using atrophies — the framing makes outsourcing feel free, and the bill arrives later as 'cognitive debt.'
For work, the deeper shift is structural. One note argues AI doesn't just speed up steps inside intellectual labor — it automates *composition itself*, splitting the outward form of a product from the reasoning and values that used to be inseparable from it Does AI separate intellectual form from the thinking behind it?. Treat cognition as computation and the essay, the brief, the analysis become outputs whose 'use value' floats free from the thinking behind them. That's the same wound the EEG study measures from the inside: the form survives, the formation doesn't.
But the corpus also pushes back on the premise rather than just its consequences. One line insists computation can never bootstrap itself — it presupposes a conscious 'mapmaker' who first chops continuous physics into discrete symbols, and no amount of added algorithmic complexity conjures that agent Can computation arise without a conscious mapmaker?. A related piece, borrowing Habermas, says humans and LLMs look categorically different from the outside yet draw on the same symbolic substrate once both are *inside* a conversation Do humans and LLMs differ fundamentally or just superficially?. Together they suggest education's real job under this framing isn't producing computation — machines do that — but cultivating the experiencing agent who decides what's worth computing.
What's quietly reassuring is that the corpus's own model of machine cognition undercuts the crude 'mind = symbol-cruncher' picture that drives the most anxious takes. Memory-amortized inference frames intelligence as *reusing* prior reasoning paths over a topological memory rather than recomputing from scratch Can cognition work by reusing memory instead of recomputing?, and analysis of millions of pretraining documents finds that reasoning generalizes from broad, transferable *procedural* knowledge — how-to patterns — while only factual recall depends on narrow memorization Does procedural knowledge drive reasoning more than factual retrieval?. The educational implication flips the usual fear: if even machine 'reasoning' is procedural reuse, then the durable thing to teach humans is procedure and transfer, not retrievable facts the computation already owns.
The payoff the reader may not expect: the same cognitive-science lens that lets us reverse-engineer LLMs can be turned back on ourselves. Marr's three levels — what a system computes, how, and on what hardware — are offered as a structured way to explain LLMs layer by layer instead of as black boxes Can cognitive science methods unlock how LLMs actually work?. Reframing cognition as computation, then, isn't only a threat to schooling and jobs; it hands education a vocabulary for teaching *about* thinking itself — provided we don't let the convenience of offloading hollow out the thinker.
Sources 7 notes
A four-month EEG study of 54 participants found that brain connectivity systematically scaled down with AI reliance—LLM users showed weakest neural engagement, poorest memory retention, and impaired ability to recall their own recent work.
Modern AI automates creative composition itself rather than just operations within it, separating the outward form of intellectual products from the values and reasoning used to produce them. This mechanism allows exchange value to float free from use value.
Computational systems depend on a conscious mapmaker who alphabetizes continuous physics into discrete symbols. No increase in algorithmic complexity can generate this agent; it must logically precede the computation it makes possible.
Applied Habermas's observer/participant distinction to AI: from outside, humans and LLMs are utterly different; from within shared discourse, both draw on the same symbolic substrate, making the difference structural rather than absolute.
Memory-Amortized Inference proposes intelligence arises from structured reuse of prior inference paths over topological memory, inverting RL's reward-forward logic into cause-backward reconstruction. This duality explains energy efficiency and suggests memory trajectories form the substrate of adaptive thought.
Analysis of 5 million pretraining documents shows reasoning relies on broad, transferable procedural knowledge from diverse sources, unlike factual recall which depends on narrow, document-specific memorization of target facts.
Cognitive science's 70-year toolkit of behavioral probes, causal interventions, and representational analysis transfers directly to LLM interpretation. Marr's computational, algorithmic, and implementation levels reframe the problem structurally and enable layered rather than monolithic explanation.