What the is-ought gap actually says — and what it doesn't
Ask any educated person about the is-ought gap and you'll get a confident summary: David Hume proved in 1739 that you cannot derive moral conclusions from factual premises. "Ought" cannot follow from "is." Science describes the world; it cannot prescribe how to act. Case closed. Ethics and physics are permanently separate domains.
This understanding is so universal that it functions as a reflex. Anyone who suggests that physical reality might constrain ethical conclusions gets the same response — "Hume's guillotine" — delivered with the finality of a mathematical proof. The conversation ends there.
There's one problem: Hume said no such thing.
The passage is in A Treatise of Human Nature, Book 3, Part 1, Section 1. It's short. Here it is:
In every system of morality, which I have hitherto met with, I have always remark'd, that the author proceeds for some time in the ordinary way of reasoning, and establishes the being of a God, or makes observations concerning human affairs; when of a sudden I am surpriz'd to find, that instead of the usual copulations of propositions, is, and is not, I meet with no proposition that is not connected with an ought, or an ought not.
This change is imperceptible; but is, however, of the last consequence. For as this ought, or ought not, expresses some new relation or affirmation, 'tis necessary that it shou'd be observ'd and explain'd; and at the same time that a reason should be given, for what seems altogether inconceivable, how this new relation can be a deduction from others, which are entirely different from it.
Read it again. Slowly.
Hume is not declaring that ought can never follow from is. He is observing that moral philosophers keep sneaking from is to ought — "this change is imperceptible" — without explaining how they got there.
His demand: "it shou'd be observ'd and explain'd."
That's not a guillotine. It's a quality control audit. Hume caught philosophers smuggling normative conclusions past the reader without justification, and he said: stop and explain the transition.
He was right. They were smuggling. Nobody was explaining.
But something happened to this observation on its way through 287 years of intellectual history.
The telephone game went like this:
Hume (1739): "Explain the transition from is to ought. Nobody is doing this."
G.E. Moore (1903): "Defining 'good' in terms of natural properties is a fallacy." The "naturalistic fallacy" — a stronger claim. Not "explain the transition" but "the transition is a logical error."
Logical positivists (1930s): "Moral statements are neither true nor false. They're emotional expressions." The fact/value dichotomy hardens into dogma.
Educated common sense (1960s–present): "Values are subjective. Science can't tell you what to value. Anyone who says otherwise is confused." The thought-terminating cliché is installed.
Each step made the prohibition stronger and more absolute than the last. Hume's demand for explanation became Moore's logical prohibition, which became the positivists' categorical separation, which became the modern assumption that values and facts inhabit different universes.
By now, "Hume's guillotine" fires as a reflex before anyone checks what Hume actually wrote. The meme is so efficient that it terminates the exact reasoning chain it was supposed to demand justification for.
A demand for rigor became a ban on thinking.
The confusion persists because two very different claims get conflated:
The strong claim: "Physics tells you what to value." This IS what Hume warned against, and he was right. You cannot deduce from the laws of thermodynamics that you should prefer Bach over birdsong, cooperation over solitude, or any particular way of life over another. Sam Harris tried this — deriving values from neuroscience. Bentham tried it — reducing all human experience to pleasure and pain. Both failed. The strong claim was always wrong.
The weak claim: "Physics tells you which value-configurations self-destruct." This is a different type of statement entirely. Not "you should value X" but "if your system is configured as Y, it will cease to exist." This is not normative. It's structural analysis. Engineering.
Nobody invokes Hume against structural engineering. "You can't derive the load-bearing requirements of a bridge from physics" is a sentence nobody has ever uttered, because it's obviously absurd. The bridge doesn't care about your preferences for gravity. It either holds or it doesn't.
The weak claim says: civilizations are physical systems too. They are bounded regions of spacetime that maintain internal order by pumping entropy across their boundaries. This is not a metaphor. It's a thermodynamic description. And thermodynamics constrains which configurations of such systems persist.
Hume's passage demands explanation of the is-to-ought transition. Here is the explanation:
There is no transition. The weak claim never crosses the is-ought gap. It stays entirely on the "is" side. "This bridge will collapse" is not an ought-statement. "This civilizational configuration will self-destruct" is not an ought-statement. Both are predictions derivable from physics.
If you then choose to care about whether your civilization persists, the constraints become relevant. But that choice is yours. Physics doesn't make it for you. It just tells you the consequences.
Any system that maintains order against entropy — a cell, a corporation, a civilization, a mind, a future artificial intelligence — faces physical constraints on its continued existence.
These constraints are derivable from first principles. Briefly: finite energy forces a tradeoff between maintenance and growth. Bounded systems must define where "self" ends. Effective action requires models of reality. Coordination requires architecture. These dilemmas compress into three fundamental problems (time allocation, world-modeling, boundary-setting), whose optimal solutions are discoverable.
The AI alignment community already accepts this logic. Instrumental convergence (Omohundro, 2008; Bostrom, 2012): any sufficiently intelligent agent, regardless of its terminal goals, will converge on certain instrumental subgoals — self-preservation, resource acquisition, goal-integrity maintenance — because physics constrains what any agent needs to achieve anything.
Nobody objects to instrumental convergence on Humean grounds. Nobody says "you can't derive what an AI agent needs to do from physics." The derivation is accepted because it's obviously correct: it stays on the "is" side of the gap. It describes what's physically required for goal-pursuit, not what goals to pursue.
The move is one hop: apply the same argument to civilizations. Physics constrains what any civilization needs to do to persist, regardless of what that civilization "values." The instrumental requirements are the same: maintain your entropy-pumping capacity, defend your capital stocks, sustain your complexity-generation capability. The specific value-configuration is your business. The boundary conditions are physics.
Hume's gap is intact. No ought was derived from is. What was derived is: which configurations of ought are compatible with continued existence.
The pieces have been available since the mid-19th century. Thermodynamics (Clausius, 1850). Dissipative structures (Prigogine, 1960s). Instrumental convergence (Omohundro, 2008). Game theory (von Neumann, 1944). Mechanism design (Hurwicz, 1960s). The connection is not computationally difficult. Each step is individually simple.
Three forces prevented the synthesis:
Disciplinary boundaries. Philosophers know Hume but not thermodynamics. Physicists know thermodynamics but don't do ethics. Governance scholars know neither well enough. The synthesis requires crossing three fields simultaneously, and academic incentives punish boundary-crossing. Each field has its own literature, its own jargon, its own career structure. The person who could connect them would need to be outside all three — which means no credentials in any, which means no audience.
The prohibition-meme. "Hume's guillotine" fires before the reasoning chain reaches the connection point. A philosopher who starts thinking "what does thermodynamics require..." gets intercepted by the reflex: "you can't derive ought from is." The chain breaks. The meme doesn't need to be correct. It just needs to fire fast enough to prevent the connection from being made.
Poisoned precedents. Harris (2010) tried the strong claim — deriving values from neuroscience — and failed publicly. Bentham (1789) tried it with utility and produced the paperclip maximizer that ate Western civilization. Every subsequent attempt gets pattern-matched to these failures. "Oh, you're trying to get ought from is? Harris tried that. Didn't work." The failures of the strong claim immunize against the weak claim — even though they're categorically different moves.
The irony: Hume would have recognized all three obstacles immediately. People proceeding in the ordinary way of reasoning, making imperceptible transitions, without observing or explaining. The same intellectual vice he diagnosed in 1739 — applied to the interpretation of his own passage.
"Values are subjective" is a category error. It's true for preferences — chocolate versus vanilla, jazz versus classical. Physics leaves wide latitude. It's false for survival constraints — growth versus stagnation, truth-seeking versus self-deception. Some configurations of matter persist. Others dissolve. This is not a value judgment. It's observation.
Ethics becomes engineering. Not "constrained by" physics the way a decorative choice is constrained by load-bearing walls — as if the physics part were a minor boundary and the real action were elsewhere. The constraints are the ethics. Once you accept that sustained complexity generation is the objective — the one axiological wager the framework makes explicit — then the optimal solutions to the four physical dilemmas are the virtues, derivable from thermodynamics, information theory, and game theory. Every ethical tradition turns out to have captured one piece of a multi-objective optimization problem that physics poses and physics answers.
The subjective zone is real but small. After physics, biology, game theory, and selection pressure have consumed everything load-bearing, there remains a residual space of genuine freedom — instrumental preferences, aesthetic choices, the things physics truly doesn't care about. That space exists. But it's far smaller than "values are subjective" implies. Most of what people call "value disagreements" are disagreements about engineering — about which configurations actually work — not about what to optimize for.
Governance becomes applied physics. Not "which preferences should win?" but "which institutional configurations maintain complexity-generation capacity over deep time?" The question transforms. The answer becomes testable. Falsifiable. Improvable. The trolley problem — philosophy's most famous ethical puzzle — reduces to multi-objective optimization under uncertainty with physically grounded constraints. So does everything else.
He was right that moral philosophers were smuggling ought from is without justification.
He was right to demand that the transition be "observ'd and explain'd."
He was right that it "seems altogether inconceivable" how ought can be deduced from is.
It is inconceivable — for values. You cannot deduce what to want from how the world is.
It is trivial — for constraints. You can derive what won't work from how the world is. Engineers do it every day.
The distinction was always there. A demand for rigor was misread as a prohibition, and the prohibition became a meme, and the meme terminated a reasoning chain for 287 years.
Hume asked for an explanation. Here it is:
Physics does not tell you what to want. But once you want anything at all, physics determines what you must do to get it — and what will destroy you if you don't. The "transition" Hume demanded we explain is not from is to ought. It's from is to must, conditional on persistence. Engineers made this transition centuries ago. Ethics is late to the job site.
This draws from Aliveness, a framework for understanding what sustains organized complexity over time.
Related: The Question Nobody Asks — the derivation. Values Aren't Subjective — the category error. The Original Sin of History — what happened when someone tried the strong claim.
The Hume passage: A Treatise of Human Nature, Book 3, Part 1, Section 1 (T 3.1.1.27). Full text at Wikisource. The SEP notes that "few passages in Hume's work have generated more interpretive controversy." The dominant 20th-century reading (Hume is issuing a prohibition) is contested by the "logical" reading (Hume is merely noting a logical gap that needs filling). See Hume's Moral Philosophy (Stanford Encyclopedia of Philosophy) for the interpretive landscape.
Moore's naturalistic fallacy: G.E. Moore, Principia Ethica (1903). Moore's argument is structurally different from Hume's — Moore claims defining "good" in natural terms is a logical error; Hume observes that the transition goes unexplained. The conflation of these two distinct claims is itself part of the telephone game.
Instrumental convergence: Steve Omohundro, "The Basic AI Drives" (2008). Nick Bostrom, "The Superintelligent Will" (2012). The argument: any sufficiently intelligent agent converges on instrumental subgoals (self-preservation, resource acquisition, goal-integrity) regardless of terminal goals, because physics constrains what's required for goal-pursuit. Widely accepted in AI alignment literature without Humean objection.
Harris's strong claim: Sam Harris, The Moral Landscape (2010). Harris argued that science can determine human values by measuring well-being via neuroscience. The philosophical response was overwhelmingly negative — correctly, because this IS the strong claim. The failure was instructive but poisoned subsequent attempts, including ones making the categorically different weak claim.
Bentham's full-stack attempt: Jeremy Bentham, An Introduction to the Principles of Morals and Legislation (1789). Bentham's Theory of Fictions + utility calculus was the first systematic attempt at vertically integrated civilizational engineering from materialist premises. It produced the framework (utility maximization) that, extended through the Marginal Revolution → Arrow-Debreu → QALYs → EA, became the dominant optimization metric of modern governance. For the causal chain, see The Original Sin of History.
Hans Jonas: The Imperative of Responsibility (1979). The closest prior attempt in philosophy to ground ethics in physical reality. Jonas derived normative claims from the fact of metabolism — "life creates value." But he explicitly violated Hume rather than bypassing him: his premise "anything that must be done, ought to be done" smuggles normativity without justification. Jonas breached the wall; the weak claim walks around it.