From Science to Philosophy

0. A Counterweight

In From Philosophy to Science, I argued against speculation that floats free of evidence. Concepts like consciousness and free will, I wrote, had been debated for centuries without anchoring themselves to anything observable, and that failure to anchor was not a mark of depth but a mark of drift. The remedy was formalization: state a concept precisely enough that it generates testable questions, and let observation correct it.

That essay was half of something. This one is the other half.

This is not a retraction, and it is not a mechanical rebalancing — as if philosophy and science were two weights on a scale that simply need to be evened out. The relationship is sequential, not symmetrical. Description precedes prescription; empirical investigation precedes normative commitment. You investigate the world and arrive at knowledge — and then, only then, does the question of what to do with that knowledge, how to hold it sincerely, how to act on it, begin. A body of knowledge that stops at investigation and never moves to the next step is not humility. It is knowledge left unused.

What follows is an attempt to take that next step honestly — grounded in observation, but ending somewhere observation cannot reach. I will not offer a solution. I will show the movement itself: from watching the world, to noticing a question observation cannot settle, to speculating about it without pretending the speculation is a discovery.

1. What the News Keeps Showing Me

Start with what is actually happening, not with abstraction.

In the United States, the current administration has pursued an aggressive program of tariffs alongside sharply restrictive immigration policy, justified as a way to rebuild domestic manufacturing and protect American workers. By 2026, new tariffs applied to roughly 54 percent of US goods imports, pushing the applied tariff rate to a level not seen in decades. The promised revival of manufacturing did not materialize at the scale advertised: manufacturing output growth slowed to a near standstill by the final quarter of the year, and the sector shed tens of thousands of jobs even as the policy was in full effect. Immigration restriction, layered on top, threatened to slow job creation further rather than free up positions for native workers, since much of the labor growth in tight sectors had been coming from recent immigrants in the first place.

At the opposite pole, New York City elected a democratic socialist mayor whose platform includes rent freezes, government-run grocery stores, and aggressive taxation of the wealthy to fund expansive public services. The rent freeze, intended to protect tenants, is a price ceiling — and price ceilings on housing have a well-documented tendency to suppress the incentive to build and maintain rental units, deepening the very shortage they are meant to solve. The redistributive model funding it all rests on the assumption that an expanding tax base can support ever-larger public commitments, an assumption that historically has not survived the erosion of that base once high earners and capital begin to relocate.

Neither of these is a hypothetical. Both are underway, right now, defended by people who are not stupid and who believe, sincerely, that they are solving the problem in front of them.

And it is not only the United States. Across Europe, populist parties — far-right and far-left alike — won 23.2 percent of the vote across seven countries holding elections in 2026. But the picture is not one of steady advance. Hungary and Poland, once far-right strongholds, elected more moderate governments in their most recent elections. Sweden’s far-right party lost support, dropping to third place. In France, the far-right Rassemblement National lost the mayoral races that mattered most — Marseille, Toulon, Nîmes — even as it consolidated its grip on smaller towns. The pattern is not a line moving in one direction. It oscillates.

So the question that forms, watching all of this, is not “which side is right.” Both are being tested against the only standard I find myself actually caring about: does it solve the problem it claims to solve? And by that standard, both are failing, in mirror-image ways. One promises order and delivers a slower, harder economy. The other promises fairness and delivers scarcity. And across an entire continent, no single answer holds still long enough to call it a winner.

Which raises the real question. Is there a coherent balance point that these systems are failing to find — a place where governance actually works — or is “balance” nothing but the name we give, after the fact, to whatever survives the oscillation between extremes that fail? Is equilibrium a location, or only a label retroactively applied to whatever tends to remain standing?

2. Why This Is Not a Scientific Question

It would be convenient if this were a scientific question — if enough data, run through the right model, could simply output the answer.

I have tried that move before. In From Philosophy to Science, I formalized freedom as DoF(s, a) > 1 — the size of an available action set — precisely so that questions about it could be handed to cognitive science, to economics, to measurement. The same instinct shows up here: define an objective function, maximize the total degree of freedom across a population subject to the constraint that the system must persist, and let a sufficiently capable analytical engine compute the answer. This was, more or less, the design I proposed in A Perpetual Peace for the 22nd Century — an AI that simulates outcomes and reports their expected value, while leaving the final choice to human deliberation.

But formalization only gets you to the base of the wall. It does not get you over it. Because before any function can be maximized, someone has to decide what counts as the thing worth maximizing — the total sum of freedom across everyone, or the floor available to the worst-off? Stability purchased at the cost of individual liberty, or liberty purchased at the cost of stability? No measurement resolves that choice, because the choice is not a fact about the world. It is a commitment about what matters, made prior to any measurement.

This is not a new discovery. Hume noticed it more than two centuries ago: a description of what is does not, by itself, yield a conclusion about what ought to be. What is new, if anything, is realizing that this is not a defect in the question about political balance — it is the reason the question is a philosophical one to begin with. Value has no answer key. That absence is not a temporary gap waiting for better instruments. It is the permanent shape of the territory.

So I stop reaching for the calculator, and I try to think instead.

3. Pushing Both Extremes to Their Limit

Take total decentralization first. Push it all the way.

Cooperation, once you strip away the sentiment, is an invented device — built because individuals too weak to survive alone needed a way to coordinate at a scale personal acquaintance could no longer manage. Law, money, shared narrative: none of it is physically necessary, all of it is functionally necessary, the moment a group outgrows what everyone can verify about everyone else in person. Erase that device, push decentralization to its limit, and what remains is not a lightly-governed society. It is no society at all — a return to whatever scale of trust direct, personal contact can still support. Not a weaker version of the collective. Its dissolution.

What strikes me, watching this play out in real political movements, is the psychology underneath it. Those who push hardest toward this limit are rarely trying to destroy the coordination they depend on. Quite the opposite — they act as though it is indestructible. They were born inside a system that already worked, and a system that works stops being visible as a system at all; it becomes background, like air. So they test its limits freely, confident that the background will simply remain in place no matter how hard they push against it. This confidence is not evidence that the system is sturdy. It is closer to a symptom of how well it has worked so far — because a coordination device only fades from view once it has stopped being questioned, and a device that has stopped being questioned is precisely the one people forget they are standing on.

Now take the opposite limit. Total centralization.

Here something asymmetrical happens, and I think this asymmetry is the part worth dwelling on. Decentralization’s failure is loud — it collapses, visibly, into fragments. Centralization’s failure is quiet, and in a sense far worse, because it does not fail by collapsing. It fails by persisting. A system that pours all of its effort into cohesion does not fall apart; it holds, sometimes for a very long time, precisely because it has eliminated the friction that dissent, error, and disagreement usually cause. What it sacrifices instead is the individuals inside it — not the system’s death, but a slow attrition of the people the system was supposed to serve, right down to what A Theory of Value calls the floor: the one thing no arrangement is entitled to take, the freedom to choose even one’s own end. A system that removes even that has not merely restricted freedom. It has done something worse than killing the people in it, because it has taken away their right to die on their own terms too.

And this is where the deeper tragedy sits. Such a system does not need a philosopher-king to be dangerous — it only needs to succeed too completely at the one thing it was built for. Perfect cohesion requires erasing the local units that might otherwise notice, first, that the world has changed. So when the environment does shift — and it always eventually does — nothing in the system is left in a position to notice early enough to matter. The empire that never bends is not stable. It is simply storing up all of its bending for one moment, later, when it breaks entirely.

Two extremes. Two failures. A system that dies by fragmenting, and a system that dies slowly while insisting it is thriving. Different in almost every visible way — and yet, looking at them side by side, I keep suspecting they are not two different diseases. They look like the same underlying tension, expressed through two different symptoms: the tension between holding together and staying responsive, between what a system needs to remain one thing and what it needs to remain capable of noticing that it is wrong.

4. The Same Tension, Under Many Names

I do not think I am the first to notice this tension, and I do not want to pretend otherwise. What I want to note is only that it keeps reappearing, in field after field, dressed in different vocabulary.

Tocqueville worried about exactly this in early American democracy: centralization is efficient but deadens the political vitality of individuals, while unchecked individualism dissolves the bonds that make collective action possible at all. His answer was neither pole, but layered intermediate institutions — local government, voluntary associations — that let cohesion and responsiveness coexist at different scales simultaneously.

In cybernetics, Ashby’s law of requisite variety states the same tension formally: a system can only regulate a disturbance if its own internal variety is at least as great as the variety of the disturbance it faces. Strip out internal variety in the name of unity, and you strip out, by the same motion, the system’s capacity to absorb whatever the world throws at it next.

Isaiah Berlin’s distinction between negative and positive liberty is aimed at a related danger — the way “positive freedom,” the idea that some authority knows what you truly want better than you do, has historically been used to justify exactly the kind of centralization that erases the individual’s capacity to notice and object.

And in a completely different register, Kahneman’s System 1 and System 2 — fast, parallel, locally responsive intuition against slow, unified, deliberate integration — describe, inside a single mind, the same architecture problem that the others describe across an entire society.

None of these thinkers are the reason I hold this view, and I want to be precise about what that means. I did not arrive here by consulting their arguments and picking a winner. I arrived here by pushing two extremes as far as they would go and watching what broke. That several people, working in different centuries and different disciplines, kept running into the same wall is not a debt I owe them. It is what happens when independent inquiries starting from different observed facts converge on the same shape. I call that shape rational liberalism — reason supplying the coherence a system needs to hold together, liberalism supplying the standing local variety it needs to stay capable of noticing when it is wrong. Not a compromise between the two. A recognition that a system missing either one is not a weaker version of a system. It is a different kind of failure altogether.

5. A Personal Answer

I do not think this tension resolves differently depending on whether the system in question is embodied or not. A brain built by evolution over hundreds of millions of years and a society assembled deliberately over a few thousand are different in almost every material respect — one is written into flesh and cannot be rewritten without ending the organism it belongs to; the other is closer to software, revisable, sometimes violently, by the people running underneath it. But the structural demand each one faces — hold together, and still notice when you’re wrong — does not care which substrate it’s implemented in. That is a claim about mechanism, not about material, and I think the difference between the two matters enormously for what can be redesigned and what cannot, without erasing the fact that the underlying problem is the same one.

I have tried to answer this question directly before, more than once, and I will not repeat those arguments here — only point toward them, in case the reader wants to see where this speculation actually led.

A Theory of Value built the floor this essay leans on: freedom as something no arrangement is entitled to fully extinguish, and value as an invented device whose entire justification is enabling cooperation at a scale personal trust cannot reach on its own.

A Perpetual Peace for the 22nd Century was my attempt at an actual architecture for this tension — not balancing cohesion and responsiveness within a single layer, but separating them: centralizing the analysis, so that error can actually be detected at scale, while leaving the decision itself distributed, so that no single point of failure can extinguish the system’s capacity to notice it has gone wrong.

And Intelligence Is a Theory of the World names the deeper structure underneath both of those essays more precisely than I have managed here: intelligence, at any scale — a cell, a brain, an organization, a civilization — is the capacity to build a model, act on it, and correct it when reality disagrees. Read through that lens, centralization’s failure is not “too much cohesion.” It is the collapse of the loop’s ability to detect error. And decentralization’s failure is not “too much freedom.” It is the collapse of the loop’s ability to preserve and pass on a correction once it’s found. Not two competing resources fighting over a fixed budget — two segments of one loop, either of which can be severed independently, and each severing looks, from the outside, nothing like the other.

6. An Invitation, Not a Conclusion

I have not resolved anything here, and I want to be honest that this was the intention from the start.

What I wanted to show, rather than argue, is what philosophy does once science has said everything it is able to say. It does not get to invent facts, and it does not get to claim its speculation as discovery. But it does get to ask what ought to be done with the facts once they run out — and it can do that without floating away from the world that produced the question, if it is willing to keep testing itself against that world at every step, rather than only at the beginning.

Investigation of things, extension of knowledge — that much, the tariffs and the rent freezes and the electoral maps supplied on their own. What comes after, the sincerity of intention, the rectification of the mind, is not a departure from that investigation. It is refusing to let it go to waste.

I don’t know if there is a location called balance. I suspect there might only be a loop that stays unbroken, or doesn’t. I would be glad to be shown wrong by someone thinking alongside me rather than by silence.