Where selection stops explaining
Why Your Mind Can Understand the Universe
Selection built a mind good enough for the middle-sized world. It keeps turning out to be right about everything else.
Why can a human being understand quantum mechanics, curved spacetime, and the infinity of the continuum? None of it helped anyone survive. Proving a theorem about uncountable sets caught no food and avoided no predator. So why is the ability there?
The obvious answer is more than half right. Selection built a general problem-solver, and generality is the point of the thing rather than an accident of it. Ask it about seasons and it plans a harvest; ask it about particles and it does physics. That disposes of most of the puzzle. The residue is not about power. It is about range.
What makes a domain hard is not unfamiliarity — an unfamiliar problem is what a general solver is for. The hard case is a domain that is contrary: one that denies what the solver has built into it. Quantum mechanics denies locality, denies that things have definite properties before measurement, denies that a measurement has one outcome. Each is an assumption an environment-modeller carries structurally, because for middle-sized things they hold, and a creature that doubted them would have been eaten. So this is where a recruited faculty ought to fail — not to struggle, but to fail, the way a hand fails at a task that needs a wing.
Instead it goes to twelve decimal places, and the decimal places are right.
Matt Dorman put the sharpest objection to me: our cognitive profile is contingent. Had our ancestors stayed in the trees, we would have a different one. I agree — and that is the difficulty rather than the answer. If the profile is contingent, its fitting the parts of reality nothing prepared us for is a coincidence, and not a single one. Non-Euclidean geometry was a curiosity decades before gravity needed it; group theory was developed for its own sake and turned out to describe particles nobody had looked for. The tools keep being there before the domain is.
That is not a missing link but a missing constraint. Selection explains why we build models, not why they are accurate where selection never reached. A faculty good enough for the middle-sized world is satisfied by models roughly right there and wrong everywhere else. That is what we should have got.
Two closings are usually offered, and both fail. Inheritance: the structure was there first, in a mind above the world, and ours copies it — which explains the fit completely, at the cost of an author. Emergence: complexity accumulated until a general modeller appeared — but emergence is contingent by construction, so it answers a question I am not asking. Kant puts the structure in us rather than the world, which makes the accuracy worse: our categories were shaped by the middle-sized world, and quantum mechanics violates them.
What I suggest is none of the three. The correspondence is not handed down, not accumulated, and not projected, but shared: the thing doing the modelling and the thing being modelled were made the same way, so the match sits in the construction rather than in the history.
Which part would I defend hardest? That a representing system must exceed what it represents is a deduction: it follows once you grant the starting points. What the excess amounts to, and how to count it, is where I am exposed — and I would rather show the load-bearing step than blur it into the confident parts.
But the question underneath is not mine, and it does not go away. The mind’s reach exceeds what survival asked of it, and it does so in a particular direction: towards the parts of the world we never met.
Where to go next
Both are free, and both are running now. The essays come out twice a week; the video course builds the argument from the beginning, a new one every few days.
A longer version of this essay is on Substack. Where this account has changed, or where I have been wrong, it is on the corrections page.