137 | Justin Clarke-Doane on Mathematics, Morality, Objectivity, and Reality
What this covers
Justin Clarke-Doane and Sean Carroll explore a surprising structural parallel between mathematics and morality. Clarke-Doane argues that mathematical realism and moral realism rest on equally solid ground, yet most philosophers who accept one reject the other. The conversation centres on why this asymmetry persists among naturalists, who typically embrace the independence of mathematical facts (since science cannot function without them) while dismissing moral facts as explicable through evolution and social convention. Clarke-Doane dismantles the standard dis-analogies — that math settles questions definitively while ethics endures endless dispute, that math applies to physics while morality does not, that evolutionary accounts explain moral belief without appealing to moral truth — showing none survives close inspection.
The discussion then turns on a crucial distinction: realism (whether independent facts exist) and objectivity (whether a unique answer exists) are separable properties. Mathematics itself provides the model. Some areas like geometry and algebra lack objectivity — there is no fact of the matter whether the parallel postulate or commutativity is "really true" — yet realists can still treat these as dealing with real, mind-independent facts. By this logic, moral facts could be both real and yet leave the practical question of action undecided. Clarke-Doane presses further: no moral fact can settle what to do, because choosing which moral property to act on cannot itself be a further moral fact without triggering infinite regress. The deliberative question of action, he argues, occupies a separate domain from fact altogether — it is what remains once all facts, moral and otherwise, are exhausted. This move insulates moral realism from the charge that it fails to guide conduct, reframing that failure as conceptually inevitable rather than metaphysically embarrassing.
Clarke-Doane argues that mathematical realism and moral realism stand on equally good footing, but that 'being real' and 'being objective' are distinct, so morality can be just as real as mathematics while still failing to settle questions of action, which are deliberative rather than factual.
- Naturalists who must accept mathematical realism (because science is saturated with math) have no principled basis to reject moral realism, since the standard dis-analogies don't survive scrutiny
- Reality (independent facts) and objectivity (uniqueness of answer) are separable: math is real but largely non-objective like geometry, while moral facts can be real yet still leave the practical 'what to do' question undecided
- No moral fact can settle action because asking whether to consult one 'ought' rather than another cannot itself be a further factual 'ought' on pain of regress
Mathematical realism posits that mathematical statements have mind-independent truth values not dependent on human convention or observation.
- It is not a coincidence that mathematical realism is called Platonism, because it is very hard to see how anything satisfying realism's constraints could be other than Platonic abstract objects — numbers are not the kinds of things observable through a telescope or discoverable by building a bigger particle accelerator, so if there are infinitely many twin prime numbers they would be something strange from an ordinary naturalist standpoint.
“It doesn't seem like numbers, for instance, are the kinds of things that we're ever gonna observe through a telescope, that building a bigger particle accelerator is gonna discover”
- Mathematical realism is the view that mathematical statements of the sort encountered in ordinary math classes (number theory, geometry, set theory) have truth values — they can be true or false — and whether they are true is not up to us; we cannot change our conventions to make the twin primes conjecture false, even though we can mean different things by the same symbols.
“mathematical statements of the sort that you encounter in an ordinary mathematics class, whether it be a number theory or geometry or set theory, these have truth values. They are the kinds of things that can be true or false. And whether they're true is not up to us”
- Quine wanted to be a nominalist about abstract objects and, with Goodman, tried in 'Steps Towards a Constructive Nominalism' to reformulate physical theories without reference to numbers, but concluded it could not work and became a reluctant Platonist, reasoning that accepting science (e.g. the standard model of particle physics) while denying math is unintelligible — an awkward component of his radical empiricism.
“Quine came to think, this is not gonna happen... And he became a 'reluctant' Platonist... he felt like I got to at the end of the day accept math and be a Platonist about it, because I want to accept science and what in the hell could it mean to say I accept the standard model of particle physics and I don't believe in math?”
- The twin primes conjecture states that there are infinitely many primes P such that P+2 is also prime; number theorists generally regard it as very likely true, and they have proved that there are infinitely many primes P such that P plus some bounded value is prime where that value keeps getting lower, but the P+2 case itself remains unproven.
“It's just the claim that there are infinitely many prime numbers, P, such that P plus two is also prime.”
Standard logical strategies to avoid commitment to abstract mathematical objects—treating math as disguised logic or conditional claims—fail epistemically and ontologically.
- The sentence 'one plus one is two' arguably goes proxy in ordinary language for a claim of pure first-order logic — e.g. that if there is a book on the table and a book on the chair and no book is on both, then there are exactly two books — where 'exactly two' can be defined without any reference to numbers, an accomplishment Frege demonstrated, making such rudimentary cases distracting examples for the realism debate.
“this was a famous accomplishment of Frege's to show that sentences like that can actually just be understood as claims of pure logic. Where by pure logic... I mean, pure first order logic”
- The attractive view that any mathematical claim P just means 'if the axioms are true, then P' (avoiding commitment to numbers) fails: if read as a material conditional, an unsatisfied set of axioms makes both P and not-P come out true; and if read as a logical-consequence relation, Gödel's second incompleteness theorem shows it is consistent to say false things about consistency, since a consistent arithmetic conjoined with the claim that it proves an inconsistency is itself consistent.
“if we take the claim if the axioms, then the twin primes conjecture to be what's called the material conditional... if the axioms aren't satisfied, that whole thing comes out true no matter what.”
- Second-order logic does not rescue the idea that math is just logic in disguise, because it is incomplete: you can build proof systems where provability implies second-order validity but none where second-order validity implies provability, and since logic should correspond to provability, second-order logic is committed ontologically to roughly the same sets and objects as mathematics while offering no epistemological advantage in knowing second-order consequence.
“second order logic is incomplete. And what I mean by that is you can come up with proof systems such that if the things that... If you can prove the thing, then it's valid, second order valid, but you can't come up with something such that if it's second order valid, you can prove it”
- Even if one could eliminate all reference to mathematical entities (as Hartry Field attempts), nothing is gained epistemologically, because everyone still needs to talk about consistency — a logical-possibility claim of the form 'diamond P' — which carries exactly the same epistemological problem as arithmetic claims: there is no experiment or observation that tells you it is true, so you have merely traded beefed-up ontology for beefed-up ideology with identical knowledge difficulties.
“you just trade all that ontology, that is all that reference to mathematical entities, for beefed-up what Quine called ideology... where exactly the same epistemological problems arise. How am I supposed to know that? There's no experiment, there's no observation.”
Moral realism and mathematical realism face parallel justificatory structures; both posit independent facts that naturalists should accept equally or reject equally.
- A naturalistically inclined philosopher or scientist tends to accept independent mathematical facts because science is saturated with mathematics and it is hard to know what it could mean to believe physical theories but not math as independent reality, while simultaneously rejecting independent moral facts because science seems to have no need for them — moral norms being explicable through evolutionary stories about why establishing conduct norms benefited our ancestors.
“science is up to its ears in math and it's hard to even know what it could mean to say something like, I believe in our best physical theories, but I don't believe in math as an independent reality. So, on the other hand, science seems to have no need for independent moral facts.”
- Although mathematics and morality seem totally different on the surface, under close examination there are surprising similarities between the case for realism about each — the case for the existence of independent facts of the matter — which becomes an important question for naturalists deciding whether one can consistently be a mathematical realist while being a moral anti-realist.
“superficially, these areas can seem totally different, on the other hand, under a microscope, there's surprising similarities between the case for realism about one or the other, that is the case for the existence of independent facts of the matter”
- When it comes to realism, moral realism is on every bit as good footing as mathematical realism, because the obvious dis-analogies do not survive scrutiny: that math settles things once and for all with proof while ethics argues endlessly, that math gets applied to the physical world with no moral analog, and that we can explain our moral beliefs via evolution without appealing to their truth while having no such explanation in the mathematical case — none of these marks a deep dis-analogy.
“all the obvious dis-analogies I argue, don't hold up to scrutiny like that, well, we settle things in math once and for all with proof, whereas in ethics just people argue endlessly or in math, our theories get applied to the physical world where there's no analog in the moral case”
- Realism is the claim that there is an independent fact about a thing (the nature of the reality — whether mind-dependent, convention-dependent, or fully independent), whereas objectivity is the claim that to a question only one answer can be right (the uniqueness or richness of reality); the parallel postulate shows these come apart, since a Gödel-style realist takes geometric facts to be independent once meanings are fixed, yet holds there is no objective fact about whether the parallel postulate is true as pure mathematics.
“By being real or realism, I just mean that there's an independent fact about the thing. By being objective, I mean that to a question from the area only one answer can be right”
Moral realism, unlike mathematical realism, cannot settle the deliberative question of what to do because no factual moral claim determines action.
- The view does not imply nihilism (everyone lighting cats on fire), because all the action is had at the non-cognitive level — a separate domain of deliberation prior to action, investigated by Allan Gibbard via 'hyper-plans' (idealized plans), where what to do is settled; descriptive facts about whether something causes happiness or suffering are relevant, but once all descriptive facts are settled there are no further facts, moral or otherwise, bearing on the question we must resolve.
“the claim is that all the action is had at what you might call the non-cognitive level, there's a whole another domain of deliberation, it's the final deliberation we have to do before action, and it has nothing to do with facts”
- Extending Hume's point that you cannot derive an ought from an is, Clarke-Doane argues you cannot even derive what to do from a question of ought: settling that something is the right or good thing to do does not settle the question of action, because one must still decide whether to do what is good-sub-one or good-sub-two, so questions of action are objective but totally not real — they are what remain once all the facts are settled.
“Hume famously said, 'You can't derive an ought from an is,'... I'm saying that you can't even derive what to do from a question of ought.”
- It is not merely that one could hold moral realism plus the claim that it does not tell us what to do; rather, no moral realist view could tell us what to do, because moral realism mixes two issues — the cheap reality and independence of some truths, and deliberation and action — and the action issue cannot be packed into the objects, since we can always ask whether to consult those objects rather than others, and that further question cannot itself be another factual ought-claim without infinite regress.
“the claim is supposed to be, for better or worse, that no moral realist view could have that effect... that issue cannot be kind of packed in to the objects out there 'cause we can always ask ourselves whether to consult those objects or other objects, and that can't be another factual claim on pain of regress”
- Although moral realism and even moral pluralism come cheaply, ethics — unlike math, physics, or geography — is supposed to tell us what to do, so we cannot rest content saying 'it's morally good-sub-one to kill the one to save the five and morally too-bad-sub-two to do so,' because the question of action then simply gets pushed back to which subscripted ought to act on; the deliberative question of what to do is therefore not a question of fact even if moral facts exist.
“ethics, unlike math or physics or geography, is supposed to tell us what to do. If I'm wondering... whether to kill the one to save the five, I can't say, well, it's morally good someone to kill the one to save the five, it's morally too bad to kill the one to save the five, and that's all there is to it”
- There are three distinct questions: what I ought to do (a factual question about ethics), what I will do (a descriptive question of predictive psychology), and an intermediate question of what to do — whose answer is an imperative ('do X') rather than a factual claim and which cannot be true or false; weakness of will explains how one can judge X is the thing to do yet not do X, distinguishing all three.
“One question is, what ought I do? Understood as like a factual question about ethics. Another question is what I will do. Understood as a descriptive question about predictive psychology. I'm claiming there's a question in between which is what to do”
The realist move of identifying a natural-kind moral property that 'glows' as the true answer fails because it regenerates the pluralism problem at the meta-level.
- There is probably no determinate natural-language semantics for moral claims — people mean different things, and emotivists at the turn of the 20th century held 'good' merely expresses 'yay' or 'boo,' which cannot be true or false — but this question of what people are doing when they utter moral words is independent of the metaphysical question: one can stipulatively introduce a language about how we ought to live independent of anyone's opinion, and belief in those facts is on as good footing as belief in numbers, yet still does not settle practical questions.
“my feeling is that there's probably no determinate natural language of moral claims, people probably mean lots of different things”
- Moral realism is actually on better footing than mathematical realism because, unlike math, morality has no special ontology of its own — it concerns ordinary things like people and societies and merely predicates new properties of them — and being clear about what it means to be a realist about some properties over and above others is enormously difficult unless one becomes a Platonist about properties (which no naturalist wants), so it is hard to see how moral realism could fail to be true.
“math has its own special subject matter, or in the lingo of philosophy, it's own ontology, but morality actually doesn't, it's about things like you and me and societies and so forth, and all it does is predicate new properties of them”
- The realist move (made by David Enoch and Tristram McPherson) of saying one moral-like property is the 'natural kind' that carves nature at the joints — the one that 'glows' as the true morality — fails by an argument analogous to Shamik Dasgupta's in natural science: either the claim that these properties are natural kinds carries no normative import (so it is like learning something is merely 'brown' — irrelevant to deliberation, à la Hume), or it does carry normative import telling us which to use, in which case the pluralism problem simply re-arises at the meta-level with natural-kind-sub-one versus natural-kind-sub-two.
“either the claim that these properties are natural kinds is in itself a normative claim... If it's not, then it's just like the original Hume case where you learn that it's natural, but who cares?... If it is a claim with normative import telling us which we ought to use, then the problem of pluralism re-arises at the level of natural kinds”
Realism and objectivity are orthogonal: some areas like arithmetic demand unique answers while algebraic and geometric areas admit multiple equally legitimate frameworks.
- Realism (whether there are independent facts) and the geometry-like character of an area are orthogonal parameters: one can be a realist who believes there are independent mathematical facts while still holding it is 'all like geometry,' and Joel David Hamkins's set-theoretic multiverse view even treats questions of consistency as parallel-postulate-like — a position Hamkins describes as Platonism on steroids, positing a set-theoretic universe for every set of axioms one might entertain, including ones that disagree about consistency.
“one could in principle be a realist, I think there are independent facts to which our mathematical theories answer, but so to speak, it is all like geometry”
- Taking an objective stand on what is consistent with what commits you to objectivity about arithmetic, because consistency claims just are arithmetic claims — that set theory is consistent means there is no natural number coding a proof of '0=1' from its axioms — and by Gödel's second incompleteness theorem, if standard arithmetic is consistent then so is arithmetic conjoined with the claim that it is inconsistent, so one cannot say 'let a thousand flowers bloom, any axioms are as good as any other' while still holding there are facts about consistency.
“if we take the claim if the axioms, then the twin primes conjecture to be what's called the material conditional... if the axioms aren't satisfied, that whole thing comes out true no matter what.”
- Mathematics divides into areas where questions have no single right answer — algebraic areas like group theory (asking whether commutativity is 'true, period' is misconceived) and geometry (asking whether the parallel postulate is 'true, period' is misconceived) — versus areas like arithmetic that do seem to demand a unique answer, a distinction Stuart Shapiro labels by calling the former 'algebraic'.
“Stuart Shapiro, I think, calls the kind of areas like geometry, algebraic areas, just like it would be clearly misconceived to ask whether the axiom of commutativity of groups is true, period. It would be misconceived to ask whether the parallel postulate is true, period.”
Cross-disciplinary dialogue and accessible exposition of specialized technical work both advance philosophical understanding of foundational problems.
- One must distinguish the truth of a sentence under some interpretation from the truth of what it expresses: one could use 'there are tall buildings' to mean 'there are numbers,' but in ordinary language the sentence concerns buildings, so what matters for the realism debate is what mathematical sentences actually express, not their truth value under some arbitrary reinterpretation.
“you gotta distinguish between the truth of a sentence under some interpretation or another and the truth of what it expresses”
- Because philosophical literatures have become highly specialized and technical, the same kinds of issues recur across ethics and philosophy of mathematics with a great deal of re-inventing the wheel, so occasional cross-disciplinary conferences — where participants explain their starting points rather than presupposing shared expertise — would reveal points of overlap that let one conversation inform another; comparing the two fields is what enabled Clarke-Doane to arrive at his positive views in each.
“it's striking how often the same kinds of issues arise, and there's a bit of re-inventing the wheel across different areas of philosophy in my experience”
- Against the idea that existence comes in different senses (e.g. quantifier variance), the standard Quinean view holds that something either exists or it does not, and the notion of existence is the thinnest possible — there simply being truths about the thing; numbers, electrons, wave functions, spacetime points, and moral values, if they exist, all exist in the very same sense, differing only in their properties, and introducing subscripted senses of 'exist' merely confuses one of the few notions philosophy had a firm grip on.
“we can introduce by stipulation a thinnest possible notion of exists, it's the sense in which if there's a God, it exists, it's the sense in which there are numbers they exist, if there are wave functions they exist... It's the same sense.”
- Beyond hyper-specialized technical work, packaging the best and most important parts in widely accessible form is itself a genuine intellectual problem — illustrated by set theorists' recognition of an 'open exposition problem' for the technique of forcing, the challenge of explaining it to other mathematicians being as real a problem as proving the twin primes conjecture — and is as important to the general scientific enterprise as the on-the-ground work.
“We need to figure out how to package forcing in a way that other mathematicians can understand what forcing is, and that is a genuine problem as much as kind of the question of whether the twin primes conjecture is true”
- Nelson Goodman showed in the 1950s that you cannot induct on just any predicate — some predicates are 'projectable' and others are not — radicalizing Hume's point: where Hume noted no deductive argument carries you from 'all Fs to date have been G' to 'the next F will be G,' Goodman showed there is not even an inductive argument for many choices of G, so the predicate must be projectable, a notion later glossed as picking out natural kinds that carve nature at the joints.
“Nelson Goodman in the '50s pointed out that not just you can't induct on any predicate. Some predicates are what he called project-able and others aren't. This was a kind of radicalization of Hume's point”