Moral Anti-Realism Sequence #1: What Is Moral Realism?

Last up­date: 6/​8/​2019.

This is post one of a se­quence of posts I am writ­ing on moral anti-re­al­ism and moral rea­son­ing.


To start off this se­quence, I want to give a short de­scrip­tion of moral re­al­ism; I’ll be ar­gu­ing against moral re­al­ism in later posts, and I want to clearly ex­plain what it is I’m ar­gu­ing against.

When I’m ar­gu­ing against moral re­al­ism, I will de­liber­ately set aside some moral re­al­ist views and fo­cus on those forms of moral re­al­ism that I find most rele­vant – in the sense that the “rele­vant” ver­sions, if cor­rect, would be the most rele­vant to effec­tive al­tru­ism and to peo­ple’s lives in gen­eral. I will call these ver­sions of moral re­al­ism strong moral re­al­ism. Thus, I don’t claim that all ver­sions of moral re­al­ism dis­cussed in the aca­demic liter­a­ture are mis­taken.

The goal of this in­tro­duc­tory post is three­fold:

  1. to give a quick overview of metaethics[1] and differ­ent ver­sions of moral realism

  2. to ex­plain why I find many of these ver­sions of moral re­al­ism only mod­estly rele­vant to eth­i­cal practice

  3. to out­line what I take to be strong moral realism

Overview and summary

  • Two defi­ni­tions of moral realism

    • Mo­ral re­al­ism has two com­mon defi­ni­tions: the se­man­tic defi­ni­tion and the on­tolog­i­cal one. I con­trast these to illus­trate how moral claims can be dis­cussed at a lin­guis­tic level (“What do peo­ple mean when they make moral claims?” and a sub­stan­tive level (“Given the ob­jec­tivist as­sump­tion that moral claims re­fer to a speaker-in­de­pen­dent moral re­al­ity, are they some­times true?”). Po­si­tions that are some­times referred to as ‘moral re­al­ism’ are not always con­se­quen­tial (i.e., their truth or falsity does not have ac­tion-guid­ing im­pli­ca­tions for effec­tive al­tru­ism).

  • Si­de­note: Sub­jec­tivism and intersubjectivism

    • Sub­jec­tivism and in­ter­sub­jec­tivism are usu­ally not counted as moral re­al­ist po­si­tions. I dis­cuss them mainly for the sake of com­plete­ness and be­cause I think they are fruit­ful frame­works to think about moral­ity.

  • Ob­jec­tivist moral realism

    • When peo­ple talk about moral re­al­ism, they of­ten mean what I dis­cuss un­der ‘ob­jec­tivist moral re­al­ism.’ We can fur­ther split this cat­e­gory into moral nat­u­ral­ism and moral non-nat­u­ral­ism. I illus­trate these po­si­tions with a few ex­am­ples. I ar­gue that nat­u­ral­ism is at risk of be­ing too triv­ial (e.g., Peter Rail­ton’s moral re­al­ism does not seem to have ac­tion-guid­ing im­pli­ca­tions for peo­ple already in­ter­ested in effec­tive al­tru­ism), while non-nat­u­ral­ism is at risk of be­ing too rad­i­cally strange.

  • Strong moral re­al­ism: Two proposals

    • In this fi­nal sec­tion, I at­tempt to out­line – in­spired by the nat­u­ral­ism vs. non-nat­u­ral­ism dis­tinc­tion – two ver­sions of moral re­al­ism that would have im­por­tant im­pli­ca­tions for effec­tive al­tru­ism if they were true. One of them is moral non-nat­u­ral­ism in its essence of ir­re­ducible nor­ma­tivity, the other is a nat­u­ral­ism-in­spired po­si­tion I will call One True Ax­iol­ogy. In up­com­ing ar­ti­cles, I will ar­gue why I dis­agree with both those pro­pos­als of strong moral re­al­ism.

Two defi­ni­tions of moral realism

Mo­ral re­al­ism has been defined in differ­ent ways by differ­ent au­thors. I will start by dis­cussing two differ­ent defi­ni­tions, both of which are broad in that they al­low for many differ­ent po­si­tions to count as ‘moral re­al­ism.’

1. The se­man­tic definition

The first defi­ni­tion is from Ge­offrey Sayre-McCord’s Es­says on Mo­ral Real­ism (1988, p. 5). It is meant to serve as a defi­ni­tion for re­al­ism, not just about moral­ity, but about any do­main of claims un­der scrutiny.

Wher­ever it is found, [...] re­al­ism in­volves em­brac­ing just two the­ses: (i) the claims in ques­tion, when liter­ally con­strued, are liter­ally true or false (cog­ni­tivism), and (ii) some are liter­ally true. Noth­ing more.

Sayre-McCord’s defi­ni­tion illus­trates a con­fus­ing fea­ture of the de­bate be­tween moral re­al­ists and moral anti-re­al­ists: the dis­cus­sion can hap­pen si­mul­ta­neously on two differ­ent lev­els. At the first level (I will call it the lin­guis­tic level), peo­ple dis­agree about the na­ture of first-or­der moral claims – about what com­pe­tent speak­ers mean when they say things such as “Mur­der is wrong.” On the sec­ond, sub­stan­tive level, peo­ple dis­agree about whether some moral claims (prop­erly in­ter­preted) are true, or whether all first-or­der moral claims are false. (Se­cond-or­der moral claims, such as “all first-or­der moral claims are false,” can still be true even if all first-or­der moral claims are false.)

We can now see that differ­ent views re­gard­ing the lin­guis­tic level lead to differ­ent types of moral re­al­ism. Con­sider two moral re­al­ists. One thinks that moral claims such as “X is wrong” just mean the same thing as, e.g., “X re­duces net hap­piness,” and she thinks that some of these claims are true. Another thinks that moral claims such as “X is wrong” re­fer to an ir­re­ducible prop­erty of wrong­ness, and he thinks that some of these claims are true. While these are both forms of moral re­al­ism, they are quite differ­ent.

Merely se­man­tic ver­sions of moral realism

The se­man­tic defi­ni­tion al­lows for the situ­a­tion that whether one en­dorses re­al­ism or not can de­pend solely on one’s views about lan­guage rather than one’s views about moral­ity. Speci­fi­cally, some ver­sions of moral re­al­ism are grounded in idiosyn­cratic or min­i­mal­ist ac­counts of what it means for a claim to be true (see prag­ma­tism[2] or min­i­mal­ism about talk of truth[3]). I won’t ad­dress these views fur­ther both be­cause they are rarely ex­plic­itly defended and be­cause any moral re­al­ism that is en­dorsed on merely se­man­tic grounds is go­ing to be in­con­se­quen­tial: whether we con­sider it true or not only has con­se­quences for how we would speak (i.e., whether to call some­thing moral re­al­ism or not), not for how we would act.


Some moral anti-re­al­ists hold that moral claims are not best in­ter­preted as claims that can con­ceiv­ably be true or false (that is, are not truth-apt). This po­si­tion is called non-cog­ni­tivism (as op­posed to cog­ni­tivism); it is the most rad­i­cal form of anti-re­al­ism.

One non-cog­ni­tivist view is ex­pres­sivism, which holds that moral claims are best in­ter­preted as ex­pres­sions of an eval­u­a­tive at­ti­tude (veiled ex­pres­sions of, e.g., the speaker’s ap­proval or dis­ap­proval). Ac­cord­ing to the ex­pres­sivist in­ter­pre­ta­tion, when some­one says “Mur­der is wrong,” the best in­ter­pre­ta­tion of that state­ment is not some­thing that can liter­ally be true or false. Rather, the state­ment em­ploys similar lan­guage to that used in truth-apt state­ments to ex­press dis­ap­proval. “Mur­der is wrong,” that is to say, is just a non-cog­ni­tive ex­pres­sion of dis­ap­proval of mur­der; it can’t be true or false any more than it can be true or false to say “Ouch!” when you hit your­self with a ham­mer. “Mur­der is wrong” looks like a claim, but the ap­pear­ance is de­cep­tive.

Non-cog­ni­tivists have a great deal of ex­plain­ing to do. They need to ac­count for why moral dis­course has the ap­pear­ance of as­sert­ing truths, not just when we make emo­tion­ally loaded state­ments like “This is just wrong!”, but also in the con­text of care­fully rea­soned philo­soph­i­cal dis­cus­sions. We read­ily use logic to an­a­lyze moral claims, or we talk about moral ‘be­liefs’ and moral ‘knowl­edge’, and we have de­voted an en­tire branch of philos­o­phy (nor­ma­tive and ap­plied ethics) to figur­ing out which moral claims are true.

Lin­guis­tic-level dis­agree­ments are inconsequential

Fur­ther­more, it seems strange to deny that at least some peo­ple’s moral state­ments are truth-apt. After all, some peo­ple are ar­dent moral re­al­ists who make moral claims them­selves, and at least some of them will tell us ex­plic­itly that they give zero cre­dence to the non-cog­ni­tivist in­ter­pre­ta­tion of their moral claims. Whether their in­tended us­age of moral claims is the same as typ­i­cal us­age or not feels like a sec­ondary ques­tion to me. In­deed, I find it sur­pris­ing that such a sub­stan­tial por­tion of metaethics is cen­tered around con­cep­tual anal­y­sis on the lin­guis­tic level,[4] de­bat­ing what peo­ple might mean when they say things like “X is wrong.” Given that moral re­al­ists ex­ist and given that they be­lieve that their in­ter­pre­ta­tions of moral dis­course are in­tel­ligible and im­por­tant, it seems like we should be able to ad­dress their claims on their mer­its (or lack thereof), re­gard­less of what else moral dis­course may some­times be about.[5]

This makes de­bates about what moral claims mean in ev­ery­day lan­guage less rele­vant to my cur­rent pro­ject. An anal­ogy: Sup­pose that I’m dis­cussing the­ol­ogy with some philoso­phers. The philoso­phers are try­ing to in­ter­pret the Bible, mak­ing claims like, “It’s not true within the Bibli­cal sto­ryline that Noah was a woman but true that he was a man,” or, “Reli­gious claims aren’t truth-apt be­cause the au­thors of those texts were tel­ling parables to ex­press thoughts on the hu­man con­di­tion; they didn’t in­tend to say things that can be true or false in a literal sense.” And all I’m think­ing is, “Why are we so fo­cused on in­ter­pret­ing re­li­gious claims? Isn’t the ma­jor ques­tion here whether there are things such as God, or life af­ter death!?” The ques­tion that is of ut­most rele­vance to our lives is whether re­li­gion’s meta­phys­i­cal claims, in­ter­preted in a straight­for­ward and re­al­ist fash­ion, are true or not. An anal­y­sis of other claims can come later.

This is also how I look at the liter­a­ture on moral re­al­ism: Some ver­sions of moral re­al­ism would be rather in­con­se­quen­tial if true, while oth­ers would be vastly rele­vant to our lives. Whether the lat­ter stem from a typ­i­cal in­ter­pre­ta­tion of moral dis­course or not is not that im­por­tant (al­though if no one took strong re­al­ism about moral­ity se­ri­ously, that may in­di­cate ob­vi­ous flaws in the moral re­al­ism hy­poth­e­sis). So given that some ver­sions of moral re­al­ism would be vastly more rele­vant to peo­ple’s lives than oth­ers, I want to pri­mar­ily fo­cus on as­sess­ing whether strong forms of moral re­al­ism are true. Cor­re­spond­ingly, this means I won’t be satis­fied with non-cog­ni­tivist ac­counts that brush aside the pos­si­bil­ity that strong moral re­al­ism is in­tel­ligible and evaluable on its own mer­its, the mean­ing of or­di­nary moral dis­course aside. (See Ka­hane, 2013, who ar­gues that moral re­al­ism can be in­tel­ligible and defen­si­ble even if it doesn’t re­flect or­di­nary moral dis­course).

2. The on­tolog­i­cal definition

For the pur­poses of the pre­sent es­say, I’ll provide a defi­ni­tion of moral re­al­ism that works in­de­pen­dently of the de­bate about the proper lin­guis­tic in­ter­pre­ta­tion of moral claims. Stephen Fin­lay’s (2007) defi­ni­tion of what he calls ‘on­tolog­i­cal moral re­al­ism’ is just such a defi­ni­tion.

[On­tolog­i­cal moral re­al­ism is the claim that] moral claims de­scribe and are made true by some moral facts in­volv­ing moral en­tities (e.g., rea­sons, obli­ga­tions), re­la­tions (e.g., jus­tifi­ca­tion), or prop­er­ties (e.g., good­ness, right­ness, virtue). [...] This form of re­al­ism takes as its ob­jects the truth-mak­ers of moral claims, hold­ing that they in­clude moral prop­er­ties such as value (e.g., the good­ness of char­ity) and moral en­tities such as prac­ti­cal rea­sons and obli­ga­tions (e.g., rea­sons not to tell lies, obli­ga­tions to keep promises).

On­tolog­i­cal moral re­al­ism is cor­rect if moral claims are some­times true in virtue of cor­rectly refer­ring to a moral re­al­ity con­sist­ing of the “truth-mak­ers” for moral claims, the en­tities that make those claims true: moral en­tities, re­la­tions, prop­er­ties, etc. Th­ese en­tities, fur­ther, must ex­ist in­de­pen­dently of any­one’s be­liefs. What this moral re­al­ity would con­sist of is ex­plic­itly left open. On­tolog­i­cal re­al­ism is there­fore com­pat­i­ble with views ac­cord­ing to which moral facts, such as facts about value or dis­value, can be iden­ti­fied with nat­u­ral facts (e.g. facts about plea­sure or suffer­ing), and with non-nat­u­ral­ist views where the moral ‘re­al­ity’ con­sists of some­thing more ab­stract (e.g. facts about rea­sons for ac­tion pos­tu­lated by rea­sons ex­ter­nal­ism).[6] (If the differ­ence be­tween nat­u­ral­ism and non-nat­u­ral­ism seems too ab­stract now, I’ll ad­dress it fur­ther later.)

Differ­ent de­grees of ‘ob­jec­tive’

We can fur­ther dis­t­in­guish three, in­creas­ingly ‘ob­jec­tive’, con­cep­tions of moral re­al­ity. All ver­sions of on­tolog­i­cal moral re­al­ism have in com­mon that they are ob­jec­tive in the sense of be­ing about the ex­is­tence of moral facts: Facts do not change de­pend­ing on who we ask or how we look at some­thing. Then there is a sec­ond sense of ‘ob­jec­tive’ that refers to whether those facts are about some­thing that (in part) de­pends on one’s per­sonal de­sires/​goals/​prefer­ences, or whether the facts re­main the same even if one’s own de­sires/​goals/​prefer­ences are changed.

Sub­jec­tivism is the view that moral claims are made true or false with re­spect to facts about one’s own (i.e., sub­jec­tive) de­sires/​prefer­ences/​goals about the world.

In­ter­sub­jec­tivism is the view that moral claims are made true or false with re­spect to facts about both one’s own de­sires/​goals/​prefer­ences and those of other peo­ple. For in­stance, ac­cord­ing to some in­ter­sub­jec­tivist views, moral­ity is about ra­tio­nal ac­tors pur­su­ing their own ends while re­spect­ing an en­vi­sioned so­cial con­tract.

Ob­jec­tivism (not to be con­fused with Ayn Rand’s Ob­jec­tivism, which de­scribes a sub­jec­tivist moral­ity based on self-in­ter­est) is the view that moral­ity is the same for ev­ery­one and in­de­pen­dent from one’s per­sonal de­sires/​prefer­ences/​goals. (Or that one’s own de­sires at most count as “one out of thou­sands” e.g. in prefer­ence util­i­tar­i­anism.)

Ar­guably, it is only ob­jec­tivism that cap­tures the ways in which (at least some peo­ple’s) moral in­tu­itions make moral­ity out to be some­thing all-en­com­pass­ing that ev­ery per­son is bound to.

Si­de­note: Sub­jec­tivism and intersubjectivism

Nev­er­the­less, there are sub­jec­tivist and in­ter­sub­jec­tivist views that would be rele­vant to our lives if they were true. In this sec­tion, I will give two ex­am­ples of views, one sub­jec­tivist and one in­ter­sub­jec­tivist, that seem to me like cor­rect and prac­ti­cally rele­vant ways of think­ing about is­sues in moral philos­o­phy, even though they would plau­si­bly count as “not re­al­ist” ac­cord­ing to at least the se­man­tic defi­ni­tion of moral re­al­ism.


Sub­jec­tivism holds that moral value is de­ter­mined by an agent’s per­sonal de­sires. Ac­cord­ing to the sub­jec­tivist Michael Smith, for in­stance, what is good for an agent is what the agent would de­sire if they had perfect in­for­ma­tion and were perfectly ra­tio­nal. I’m in­clud­ing the fol­low­ing quote from Fin­lay (2007) be­cause the po­si­tion sounds like some of the views that have been dis­cussed promi­nently on LessWrong:

[...] Smith bases each per­son’s nor­ma­tive re­quire­ments on his or her own de­sires, sub­ject only to ra­tio­nal en­hance­ment (full in­for­ma­tion and co­her­ence). Mo­ral claims can be true, he main­tains, pro­vided that all ra­tio­nal per­sons would con­verge on a com­mon set of de­sires with a dis­tinctly moral con­tent (Mo­ral Prob­lem 173, 187–9). Richard Joyce, who largely ac­cepts Smith’s sub­jec­tivist ap­proach as an ac­count of nor­ma­tivity, rea­son­ably ob­jects that this claim on be­half of moral­ity is im­plau­si­ble. Ra­tional selves’ de­sires are reached by cor­rec­tion from ac­tual selves’ de­sires, and these start­ing points are too di­verse to sup­port the re­quired kind of con­ver­gence (89–94).

As a claim about about how moral dis­course is to be in­ter­preted, sub­jec­tivism holds that “X is good” should be treated as short­hand for “X is good ac­cord­ing to my de­sires” (Sayre-McCord, 1988, p. 18). This seems like a some­what im­plau­si­ble in­ter­pre­ta­tion of moral dis­course to me – but peo­ple’s in­tu­itions about what moral­ity is about may differ. Un­der the as­sump­tion that moral dis­course is usu­ally ob­jec­tivist, while ob­jec­tivist moral re­al­ism is false, per­haps one could re­sort to sub­jec­tivist moral dis­course as a way to sal­vage some­thing use­ful from the de­bris. In that case, sub­jec­tivism would rather count as (a con­struc­tive pro­posal within) anti-re­al­ism rather than a ver­sion of re­al­ism.

In any case, I have a lot of thoughts on the mer­its of sub­jec­tivist ac­counts that speci­fi­cally re­fer to what we would come to value af­ter moral re­flec­tion un­der ideal con­di­tions, but will re­serve them for later parts of this se­quence where I ex­plore op­tions within what I think of as moral anti-re­al­ism. So for our pur­poses here with re­spect to my up­com­ing posts on why I’m not a moral re­al­ist, I’ll treat sub­jec­tivism as one ver­sion of anti-re­al­ism. Not be­cause this is the ob­vi­ous way of cat­e­go­riz­ing it, but be­cause I want to re­serve the moral re­al­ism la­bel for only the most con­se­quen­tial ver­sions of moral re­al­ism.


Another po­si­tion lo­cated some­where near the bound­ary be­tween re­al­ist and anti-re­al­ist views is con­struc­tivism as a metaeth­i­cal view,[7] which holds that moral­ity is about what ra­tio­nal ac­tors would hy­po­thet­i­cally agree to (un­der cer­tain ideal­ized con­di­tions) with re­spect to how ev­ery­one, them­selves in­cluded, should act. Differ­ent ver­sions of con­struc­tivism give differ­ent ac­counts of how to think about this hy­po­thet­i­cal agree­ment: Some are based on con­sid­er­a­tions about so­cial con­tracts, oth­ers on Kan­tian uni­ver­sal­iza­tion of one’s de­ci­sion max­ims (“act­ing as though one ex­pects all other ra­tio­nal peo­ple to choose the same de­ci­sion pro­ce­dure”). A speci­fi­ca­tion of the con­di­tions un­der which hy­po­thet­i­cal agree­ment is to be de­rived is called a con­struc­tive func­tion (as ex­plained by Shafer-Lan­dau, 2003, pos. 201).

Me­taeth­i­cal con­struc­tivism is an in­ter­sub­jec­tivist po­si­tion. It is not ob­jec­tivist be­cause con­struc­tive func­tions merely con­strain what fol­lows from peo­ple’s de­sires, prefer­ences, or goals – they do not in­tro­duce any­thing that we ought to do (or is morally good or bad) that goes be­yond those de­sires.

While I am un­con­vinced by con­struc­tivism as a metaeth­i­cal po­si­tion (be­cause that would com­mit us to the claim that moral dis­course is nec­es­sar­ily all about hy­po­thet­i­cal con­tracts rather than also e.g. un­con­di­tional al­tru­ism or care), I am sym­pa­thetic to con­struc­tivism be­ing im­por­tant on prag­matic or pru­den­tial grounds. Is there a sin­gle, uniquely com­pel­ling way to choose a con­struc­tive func­tion? I think that the an­swer is not ob­vi­ously no. I find it note­wor­thy that cen­tral as­pects of (mostly Kan­tian)[8] con­struc­tivism are mir­rored in LessWrong-in­spired dis­cus­sions about the im­pli­ca­tions of non-causal de­ci­sion the­o­ries. Per­haps these con­sid­er­a­tions could be thought of as plau­si­ble ex­ten­sions of the con­cept “ra­tio­nal­ity as sys­tem­atized win­ning,” such that, by get­ting their im­pli­ca­tions right, one could in­crease one’s all-things-con­sid­ered de­gree of goal achieve­ment. In any case, whether we want to call this moral re­al­ism or not, it is worth flag­ging con­struc­tivism as a moral view ac­cord­ing to which moral­ity is po­ten­tially ac­tion-rele­vant in a sur­pris­ingly non-triv­ial and yet ra­tio­nally bind­ing way.

One rea­son to per­haps not think of con­struc­tivism as moral re­al­ism is pre­cisely be­cause it seems to be more of an ex­ten­sion of what it means to be ra­tio­nal rather than what it means to be moral. At least ac­cord­ing to some con­no­ta­tions of the word ‘moral,’ it is tied not only to no­tions of fair­ness or ra­tio­nal co­op­er­a­tion, but also to con­sid­er­a­tions of care or al­tru­ism. And while Kan­ti­anism or non-causal de­ci­sion the­o­ries may im­ply that one should care sub­stan­tially about the de­sires of other ra­tio­nal agents (in a per­haps power-weighted fash­ion), they do not im­ply any­thing about the con­tent of one’s own de­sires, in­clud­ing whether to care about the well-be­ing of sen­tient be­ings that are not (or in­suffi­ciently) ra­tio­nal.[9]

Ob­jec­tivist moral realism

If some­one talks about moral re­al­ism with­out fur­ther elab­o­ra­tion, this per­son is prob­a­bly[10] talk­ing about what I here call ob­jec­tivist moral re­al­ism: The view that there are speaker-in­de­pen­dent moral facts (on­tolog­i­cal moral re­al­ism) that hold for each per­son in­de­pen­dently of their per­sonal de­sires/​prefer­ences/​goals (ob­jec­tivism). I also think that ob­jec­tivism makes for the most straight­for­ward lin­guis­tic in­ter­pre­ta­tion of moral claims (al­though as I ar­gued above, this should in it­self not be our main crite­rion for se­lect­ing which po­si­tions to pay most at­ten­tion to).

Er­ror the­ory: ob­jec­tivism’s anti-re­al­ist counterpart

Er­ror the­ory is the moral anti-re­al­ist coun­ter­part to ob­jec­tivist moral re­al­ism. (In the­ory it seems con­ceiv­able to me that one can be an er­ror the­o­rist be­liev­ing that moral dis­course is sub­jec­tivist or in­ter­sub­jec­tivist in na­ture and is all false, but that would be un­usual.) Un­like non-cog­ni­tivists, er­ror the­o­rists agree with re­al­ists that moral claims are best in­ter­preted as say­ing things that can be true or false: as be­ing truth-apt. Nev­er­the­less, they deny the re­al­ist claim that first-or­der moral claims can be true. That is to say, er­ror the­o­rists hold that all first-or­der moral claims are, and must be, false.

What brand of ob­jec­tivism: Are moral facts nat­u­ral or non-nat­u­ral facts?

There are broadly two types of ob­jec­tivist re­al­ist po­si­tions: moral nat­u­ral­ism and moral non-nat­u­ral­ism. They differ with re­spect to what they take to be the na­ture of moral facts. In the same way it is true that a bach­e­lor is an un­mar­ried man, moral nat­u­ral­ists be­lieve that that moral terms such as ‘morally good’ just re­fer to some nat­u­ral prop­erty or nat­u­ral prop­er­ties, such as for in­stance plea­sure or de­sire fulfill­ment. By con­trast, moral non-nat­u­ral­ists think we will never be able to iden­tify moral terms with nat­u­ral prop­er­ties, i.e., they be­lieve that moral terms are ba­sic and mean more than what can be ex­pressed in non-moral lan­guage. Mo­ral non-nat­u­ral­ists be­lieve that, at best, we could only dis­cover how non-nat­u­ral­ist moral facts map onto (or ‘su­per­vene on’) nat­u­ral facts.[11] For in­stance, we could dis­cover that situ­a­tions where some­one need­lessly harms oth­ers always in­volve moral wrong­ness, but wrong­ness, thusly in­ter­preted, is not syn­ony­mous with need­lessly harm­ing oth­ers.[12]

How to dis­t­in­guish be­tween nat­u­ral­ist and non-nat­u­ral­ist po­si­tions is some­times a sub­ject of ex­ten­sive de­bate.[13] But per­haps the most salient differ­ence be­tween nat­u­ral­ism and non-nat­u­ral­ism is that the two po­si­tions tend to be sus­cep­ti­ble to differ­ent types of challenges. In gen­eral, moral re­al­ism is backed by the ‘moral ap­pear­ances’ (Fin­lay, 2007) – our re­al­ist in­tu­itions about moral­ity – and challenged by ex­ter­nal pres­sures about how to rec­on­cile re­al­ism about moral­ity with what we know about the world and about the mind. Mo­ral nat­u­ral­ism solves this challenge by mak­ing con­ces­sions (so one could ar­gue) that weaken moral ap­pear­ances, while moral non-nat­u­ral­ism stays max­i­mally close to these moral ap­pear­ances.

But stay­ing so close to the moral ap­pear­ances cre­ates difficul­ties in rec­on­cil­ing non-nat­u­ral­ism with the rest of what we know about the world. Th­ese difficul­ties were fa­mously sum­ma­rized by John Mackie in his Ar­gu­ment from Queer­ness. Queer­ness is the charge that non-nat­u­ral­ist moral facts, should they ex­ist, would be so differ­ent from all the other things we are used to in our con­cep­tual reper­toire that we had bet­ter think twice about in­cor­po­rat­ing them at all. The moral facts would be “queer” be­cause, de­pend­ing on the (usu­ally non-nat­u­ral­ist) moral re­al­ist ac­count in ques­tion, they may be causally re­dun­dant or im­po­tent, be epistem­i­cally in­ac­cessible, or have an (allegedly) mys­te­ri­ous con­nec­tion to hu­man mo­ti­va­tion (see rea­sons ex­ter­nal­ism).

By con­trast, moral nat­u­ral­ism (at least in most ver­sions)[14] evades these ac­com­mo­da­tion charges be­cause nat­u­ral­ists be­lieve that moral facts are sim­ply nat­u­ral facts, and that we could ex­press moral claims in non-moral ter­minol­ogy with­out nec­es­sar­ily al­ter­ing the mean­ing. How­ever, moral nat­u­ral­ists are faced with a differ­ent co­nun­drum, sum­ma­rized in G.E. Moore’s Open Ques­tion Ar­gu­ment: Sim­ply ar­gu­ing that moral and non-moral terms are syn­ony­mous (e.g. that ‘good­ness’ and ‘de­sire satis­fac­tion’ or “hap­piness” are syn­ony­mous) is du­bi­ous. This is be­cause it seems perfectly co­her­ent to ask: ‘Sure, this is an ex­am­ple of de­sire satis­fac­tion, but is it good?’ (Or: ‘Sure, this is an ex­am­ple of hap­piness, but is it good?’ Etc.) For illus­tra­tion, note that one can­not ask these ques­tions co­her­ently about pairs of terms that are truly syn­ony­mous. For ex­am­ple, one can­not say: ‘Sure, John is an un­mar­ried man, but is he re­ally a bach­e­lor?’ This ques­tion is not co­her­ent be­cause ‘bach­e­lor’ and ‘un­mar­ried man’ mean the same thing. Thus, moral nat­u­ral­ists have some ex­plain­ing to do when they they hold that moral and non-moral terms are syn­ony­mous.[15]

Mo­ral naturalism

There are many differ­ent ver­sions of moral nat­u­ral­ism and I will fo­cus speci­fi­cally on just two ver­sions. We will no­tice that moral nat­u­ral­ist ac­counts of­ten share strong similar­i­ties with sub­jec­tivism, even though they are ob­jec­tivist views. This makes sense be­cause a main challenge for nat­u­ral­ists is to show that they have sin­gled out the right nat­u­ral facts in their anal­y­sis of what is morally good or bad. And one promis­ing way of es­tab­lish­ing that one has sin­gled out the right nat­u­ral facts is by ap­peal­ing to nat­u­ral facts that are already of con­cern to each per­son in­di­vi­d­u­ally.

Ex­am­ple1: “Ac­tion-guid­ing con­cepts”

The first ex­am­ple of an ap­proach to moral nat­u­ral­ism is ex­em­plified by Aris­totelian virtue ethics as en­dorsed by Philippa Foot (1958) and Paul Bloom­field (as de­scribed in Fin­lay, 2007). Here we are less in­ter­ested in this po­si­tion it­self, and more in the method­ol­ogy or ap­proach that mo­ti­vates the po­si­tion. Foot and Bloom­field both ap­peal to biol­ogy in de­ter­min­ing which nat­u­ral facts are moral facts; they note that cer­tain things are con­ducive to the at­tain­ment of our biolog­i­cal ends – e.g. health, well-be­ing, sur­vival – and oth­ers aren’t; and con­clude via con­cep­tual anal­y­sis that ‘good’ (and other moral terms) re­fer to things con­ducive to the at­tain­ment of these ends.

For in­stance, in Mo­ral Beliefs (1958), Philippa Foot ar­gues against the idea that ‘good’ is a non-nat­u­ral­ist con­cept that solely ex­presses some kind of pos­i­tive at­ti­tude. She points out that if we did use the term ‘good’ that way, noth­ing would pre­vent a “moral ec­cen­tric” to say that a good man is some­one who ran­domly claps his hands sim­ply be­cause he (the ec­cen­tric) ap­proves when peo­ple ran­domly clap their hands. In­stead, draw­ing an anal­ogy to the con­cept ‘in­jury,’ Foot ad­vo­cates for a spe­cific, sub­stan­tive un­der­stand­ing of the term ‘good’ in­formed by con­cep­tual anal­y­sis:

[It] may seem that the only way to make a nec­es­sary con­nex­ion be­tween ‘in­jury’ and the things that are to be avoided, is to say that it is only used in an “ac­tion-guid­ing sense” when ap­plied to some­thing the speaker in­tends to avoid. But we should look care­fully at the cru­cial move in that ar­gu­ment, and query the sug­ges­tion that some­one might hap­pen not to want any­thing for which he would need the use of hands or eyes. Hands and eyes, like ears and legs, play a part in so many op­er­a­tions that a man could only be said not to need them if he had no wants at all. That such peo­ple ex­ist, in asy­lums, is not to the pre­sent pur­pose at all; the proper use of his limbs is some­thing a man has rea­son to want if he wants any­thing. [...]

It will be no­ticed that this ac­count of the ac­tion-guid­ing force of ‘in­jury’ links it with rea­sons for act­ing rather than with ac­tu­ally do­ing some­thing. Just like our con­cept for ‘in­jury’ has both speaker-in­de­pen­dent and “ac­tion-guid­ing” fea­tures, Foot ar­gues that the same holds for the car­di­nal virtues pru­dence, tem­per­ance, and courage, and per­haps also of jus­tice. Her the­ory of moral dis­course is that it is dis­course about virtues: char­ac­ter traits and dis­po­si­tions that are benefi­cial for one’s nat­u­ral ends. In­ter­est­ingly enough, Foot notes that ac­cord­ing to her view, jus­tice only con­sti­tutes a virtue if it benefits the per­son who is be­ing just. After all, she equates good­ness with what is con­ducive to one’s nat­u­ral ends, not with e.g. any no­tion of al­tru­ism or of benefit­ing oth­ers. Given that it is un­clear whether jus­tice is even a virtue on Foot’s ac­count (as she her­self points out), any­one with the in­tu­ition that moral dis­course is also about con­sid­er­a­tions of jus­tice – or sim­ply some­one who per­son­ally val­ues jus­tice – might then ques­tion whether Foot’s ac­count re­ally cap­tures what is ‘good,’ and whether we should not rather want to seek jus­tice for its own sake.

Ex­am­ple2: Sub­jec­tivism, im­par­tially extended

For con­trast, we will now con­sider an­other nat­u­ral­ist po­si­tion: that of Peter Rail­ton as out­lined in his pa­per Mo­ral Real­ism. While Foot’s ac­count based on con­cep­tual anal­y­sis of what we mean by ‘good’ may fail to be con­vinc­ing for peo­ple with differ­ent in­tu­itions, Rail­ton in­stead tries to es­tab­lish the mean­ing of ‘good’ an­a­lyt­i­cally: What is good for a per­son is what they would de­sire if they had full in­for­ma­tion about all the rele­vant facts and moral ar­gu­ments. The ad­van­tage of this ap­proach is ob­vi­ous: Such a re­duc­tion of the term ‘good’ is per­son­ally rele­vant to us by defi­ni­tion. Rail­ton makes two claims:

  1. De­sire fulfill­ment (as­sum­ing we are fully in­formed when choos­ing what to want) is what is non-morally good for a per­son.

  2. Mo­ral­ity is about what would be non-morally good for ev­ery­one (from an “im­par­tial per­spec­tive”).

With re­gards to (1), Rail­ton’s po­si­tion re­sem­bles that of the sub­jec­tivist Michael Smith: What is good for us is the fulfil­ment of the de­sires we’d have if we were fully in­formed about our situ­a­tion. What makes Rail­ton’s po­si­tion differ­ent from sub­jec­tivism is only that he fur­ther holds (2) that there is such a thing as ob­jec­tivist moral­ity, con­cern­ing what is non-morally good for ev­ery­one. He writes:

[M]oral re­s­olu­tions are thought to be de­ter­mined by crite­ria of choice that are non-in­dex­i­cal and in some sense com­pre­hen­sive. This has led a num­ber of philoso­phers to seek to cap­ture the spe­cial char­ac­ter of moral eval­u­a­tion by iden­ti­fy­ing a moral point of view that is im­par­tial, but equally con­cerned with all those po­ten­tially af­fected.

Fix­ing the con­tent of moral­ity as some­thing “im­par­tial” is what al­lows Rail­ton to have ob­jec­tivist moral facts about what is good (for ev­ery­one) even though his con­cep­tion of what is valuable for any given per­son only de­pends on their per­sonal de­sires. In­ter­est­ingly enough, Rail­ton’s moral re­al­ism does not come with any ra­tio­nally bind­ing recom­men­da­tions for how to act. His the­ory has an ax­iolog­i­cal com­po­nent (ax­iol­ogy be­ing the study of what is valuable), pos­tu­lat­ing ob­jec­tive value. But his the­ory has no de­on­tic com­po­nent (he does not be­lieve in ob­jec­tive moral obli­ga­tions).

Rail­ton uses the slo­gan “ra­tio­nal­ity does go rel­a­tive when it goes in­stru­men­tal, but episte­mol­ogy need not fol­low.“ In other words: While there is no di­rect rea­son to act morally for any one in­di­vi­d­ual be­cause ra­tio­nal­ity – pro­ce­du­rally in­ter­preted[16] – only con­cerns it­self with draw­ing proper in­fer­ences from one’s pre-ex­ist­ing de­sires, there are nev­er­the­less facts about what would make so­ciety good or bad from a per­spec­tive of max­i­miz­ing de­sire fulfill­ment for all in­di­vi­d­u­als. Whether to act morally is op­tional and up to one’s own de­sires, but rea­son­ing about moral­ity, on a purely epistemic level, can be done with an ob­jec­tive foun­da­tion.

Mo­ral nat­u­ral­ism is of­ten vague. Is there a sin­gle cor­rect no­tion of what is good for some­one?

The above may feel like a dis­ap­point­ing con­clu­sion from a pa­per ti­tled Mo­ral Real­ism. Rail­ton him­self notes that peo­ple might ob­ject that his view “may not make moral­ity se­ri­ous enough.” Hav­ing said that, per­son­ally I am happy to call po­si­tions that pos­tu­late only an ax­iol­ogy (and no moral obli­ga­tions) moral re­al­ism – pro­vided that there re­ally is one uniquely cor­rect, com­pel­ling ax­iol­ogy, as op­posed to many differ­ent ways of de­ter­min­ing what is “good for ev­ery­one” de­pend­ing on differ­ent speci­fi­ca­tions of what con­sti­tutes (moral or non-moral) good­ness.

The prob­lem I see with Rail­ton’s po­si­tion speci­fi­cally is that it is in sev­eral re­spects un­der­defined. For in­stance, it could turn out to be very difficult to for­mal­ize a uniquely com­pel­ling no­tion of “de­sires” or “de­sires given full in­for­ma­tion” that cap­tures all our in­tu­itions about when de­sire satis­fac­tion is or is not valuable. Fur­ther­more, Rail­ton’s ac­count can­not eas­ily be ex­tended to tak­ing a stance about pop­u­la­tion ethics and does not spec­ify a pre­cise no­tion of what it means to take an “im­par­tial per­spec­tive.”Rail­ton con­sid­ers this to be an ad­van­tage:

By it­self, the equa­tion of moral right­ness with ra­tio­nal­ity from a so­cial point of view is not ter­ribly re­stric­tive, for, de­pend­ing upon what one takes ra­tio­nal­ity to be, this equa­tion could be made by a util­i­tar­ian, a Kan­tian, or even a non-cog­ni­tivist. That is as it should be, for if it is to cap­ture what is dis­tinc­tive about moral norms, it should be com­pat­i­ble with the broad­est pos­si­ble range of rec­og­nized moral the­o­ries.

How­ever, not com­mit­ting to any spe­cific per­spec­tive calls into ques­tion whether there even is, in the­ory, a cor­rect an­swer. If there are many differ­ent and roughly equally plau­si­ble in­ter­pre­ta­tions of “im­par­tial per­spec­tive” or “de­sire fulfill­ment” (or more gen­er­ally: of well-be­ing defined as “that which is good for a per­son”), then the ques­tion, “Which of these differ­ent ac­counts is cor­rect?” may not have an an­swer.

What Rail­ton shows is that there is at least one (vague) view about what to call “good for ev­ery­one” that is plau­si­ble or defen­si­ble. And we know there are some views about this that are ob­vi­ously false. This is already a lot to show, as it coun­ters a po­si­tion of ex­treme moral skep­ti­cism say­ing that there is no sense at all in which we can rea­son ob­jec­tively about moral­ity.

Nev­er­the­less, Rail­ton’s po­si­tion is not quite what I would be in­clined to call strong moral re­al­ism. It leaves too much open for in­ter­pre­ta­tion be­cause we can fo­cus on widely differ­ent crite­ria when try­ing to sys­tem­atize what do­ing good for oth­ers comes down to. What I am in­ter­ested in is whether there is more to it: Can we show that there is a view that is not only defen­si­ble, but uniquely cor­rect? In the last sec­tion, I will de­scribe what con­di­tions a Rail­ton-like view would have to meet for me to count it as strong moral re­al­ism.

Mo­ral non-naturalism

What­ever we think of non-nat­u­ral­ist moral re­al­ism, it is cer­tainly am­bi­tious. Fin­lay calls it the “nor­ma­tive face of moral re­al­ism” be­cause it is com­mit­ted to the ex­is­tence of ir­re­ducibly nor­ma­tive moral facts. What is at­trac­tive about non-nat­u­ral­ism is that the other ver­sions of moral re­al­ism ap­pear to be some­what wa­tered down by con­trast. Non-nat­u­ral­ism is ar­guably best able to cap­ture the ur­gency at­tached to the sen­ti­ment that some things re­ally are right or wrong.

Sup­port for moral non-nat­u­ral­ism has been grow­ing lately. Fin­lay (2007) writes:

Although long con­sid­ered an ab­surd Pla­ton­ism, [non-nat­u­ral­ism] to­day en­joys a re­nais­sance and boasts many and dis­t­in­guished cham­pi­ons. [...] Be­sides Scan­lon and Shafer-Lan­dau, con­tem­po­rary philoso­phers who defend non-nat­u­ral­ism (al­though not all un­der that la­bel) in­clude Thomas Nagel, Derek Parfit, Jonathan Dancy, Joseph Raz, Jean Hamp­ton, Philip Strat­ton-Lake, Colin McGinn, Ter­ence Cu­neo, David Enoch, Michael Hue­mer, and William Fitz­patrick.

The non-nat­u­ral­ist po­si­tion about moral facts is of­ten in­spired by Moore’s Open Ques­tion Ar­gu­ment. Shafer-Lan­dau, for in­stance, be­lieves that “Moore was cor­rect in think­ing that we could always in­tel­ligibly ques­tion the pro­pri­ety of any can­di­date nat­u­ral­is­tic re­duc­tion [of moral terms]” (Shafer-Lan­dau, 2003, pos. 738). This leaves two op­tions: Either we ac­cept that there is no speaker-in­de­pen­dent nor­ma­tivity, or we re­gard it as a sep­a­rate realm not re­ducible to phys­i­cal facts.[17] Shafer-Lan­dau and other non-nat­u­ral­ist philoso­phers have opted to go for the lat­ter (al­though what this means ex­actly can differ from ac­count to ac­count, and some­times the differ­ence be­tween non-nat­u­ral­ism and nat­u­ral­ism is sub­tle).

Next to the strong in­tu­ition that moral nat­u­ral­ism is in­ad­e­quate to deal with the moral ap­pear­ances, both Parfit and Shafer-Lan­dau also defend moral re­al­ism by draw­ing analo­gies be­tween moral­ity and other do­mains about which (allegedly con­vinc­ing) re­al­ist in­ter­pre­ta­tions have been for­warded. For in­stance, Shafer-Lan­dau points out “part­ners in crime” within the philos­o­phy of logic/​math­e­mat­ics, and philos­o­phy it­self (Shafer-Lan­dau, 2007, pos. 646):

[...] my kind of re­al­ism must seek out part­ners in crime. I would point to cor­rect log­i­cal stan­dards or phys­i­cal laws (as­sum­ing a re­al­ist con­strual of such things), and claim that there isn’t any­thing that makes such things true – they sim­ply are true.

And also from the philos­o­phy of mind (Shafer-Lan­dau, 2007, pos. 949):

The sort of non-nat­u­ral­ism that I find ap­peal­ing is one that bears a very close struc­tural par­allel to cer­tain non-re­duc­tion­ist the­o­ries in the philos­o­phy of mind. Ac­cord­ing to these lat­ter views, men­tal prop­er­ties are not iden­ti­cal to phys­i­cal ones; men­tal facts are not phys­i­cal facts; but men­tal prop­er­ties are re­al­ized by in­stan­ti­a­tions of phys­i­cal prop­er­ties. At least in wor­lds rele­vantly close to ours, there would be no men­tal life with­out the phys­i­cal stuff that con­sti­tutes it.

We can how­ever ask whether these part­ners in crime re­ally func­tion analo­gously, and whether re­al­ist ac­counts of them are even cor­rect. As Hal­l­vard Lille­ham­mer notes in his re­view of Shafer-Lan­dau’s Mo­ral Real­ism: A Defense, “Per­haps the most in­ter­est­ing thing about these alleged com­pan­ions in guilt is that none of them are ob­vi­ously in­no­cent.”

Another ques­tion is how we could come to know any­thing about the nor­ma­tive realm, since it is sep­a­rate from ev­ery­thing else (cf. the Be­nac­er­raf-Field prob­lem in the philos­o­phy of math­e­mat­ics).

Fi­nally, a third challenge is that, as­sum­ing for the mo­ment that we grant the ex­is­tence of non-nat­u­ral­ist moral facts, moral skep­tics can challenge whether these facts are re­ally ac­tion-rele­vant for them. In re­sponse, Shafer-Lan­dau ad­vo­cates moral ra­tio­nal­ism, the view that “moral obli­ga­tions are or en­tail rea­sons for ac­tion.” On this ac­count, moral be­liefs are on their own ca­pa­ble of mo­ti­vat­ing some­one, but may not always be de­ci­sive for mo­ti­va­tion.

Strong moral re­al­ism: Two proposals

I am most in­ter­ested in ac­counts of moral re­al­ism which, should they prove to be cor­rect, will be highly rele­vant to peo­ple’s lives and life pro­jects: ei­ther di­rectly be­cause they are in­her­ently com­pel­ling (they provide ‘real’ rea­sons to act), or be­cause they are com­pel­ling at least for those peo­ple in­ter­ested in pur­su­ing goals mo­ti­vated by al­tru­ism (“do­ing good for oth­ers”). With this in mind, I will now de­scribe two differ­ent ways in which I could be con­vinced of strong moral re­al­ism.

One Com­pel­ling Axiology

Draw­ing from the above dis­cus­sion, I would call my­self a moral re­al­ist if I could be con­vinced that there is One Com­pel­ling Ax­iol­ogy in the form of a more de­vel­oped and am­bi­tious ver­sion of Rail­ton’s po­si­tion.[18] Such a view, as I en­visage it, would com­bine a spe­cific, com­plete the­ory about what is ob­jec­tively in some­one’s own in­ter­est, or is good or bad for them, with a spe­cific, com­plete the­ory of what it means to do good for oth­ers from a kind of “im­par­tial per­spec­tive.” (Which be­ings qual­ify as morally rele­vant “oth­ers” is also some­thing the One Com­pel­ling Ax­iol­ogy would have to tell us; as is whether to only count peo­ple who ex­ist cur­rently or will ex­ist re­gard­less of our ac­tions or whether to also in­trin­si­cally count the cre­ation of new be­ings.)

As a loose (and untestable, at least not with cur­rent-day tech­nol­ogy) crite­rion for what makes this form of re­al­ism true, I stipu­late that I would count some­thing as the One Com­pel­ling Ax­iol­ogy if all philoso­phers or philo­soph­i­cally-in­clined rea­son­ers, af­ter hav­ing en­gaged in philo­soph­i­cal re­flec­tion un­der ideal con­di­tions,[19] would deem the search for the One Com­pel­ling Ax­iol­ogy to be a suffi­ciently pre­cise, non-am­bigu­ous un­der­tak­ing for them to have made up their minds rather than “re­jected the ques­tion,” and if these peo­ple would all come to largely the same con­clu­sions. If the re­sult was near-unan­i­mous agree­ment about a highly spe­cific view, I would count this as strong moral re­al­ism be­ing true.

Note that this pro­posal makes no claims about the lin­guis­tic level: I’m not say­ing that or­di­nary moral dis­course let’s us define moral­ity as con­ver­gence in peo­ple’s moral views af­ter philo­soph­i­cal re­flec­tion un­der ideal con­di­tions. (This would be a cir­cu­lar defi­ni­tion.) In­stead, I am fo­cus­ing on the as­pect that such con­ver­gence would be prac­ti­cally rele­vant: If max­i­mally well-equipped and well-in­formed peo­ple were all to come to the same con­clu­sion of what it means to “do what is good for oth­ers” no mat­ter the idiosyn­crasies they started out with, then – as­sum­ing I find the prospect of do­ing what is good for oth­ers ap­peal­ing – I have lit­tle rea­son to as­sume that my cur­rent thoughts on the mat­ter of moral­ity are bet­ter than the cur­rent thoughts of some­one who holds in­tu­itions I find rad­i­cally coun­ter­in­tu­itive. This would be im­por­tant to know![20]

I place no con­straints on the pos­si­ble out­comes of moral con­ver­gence, whether what is deemed good or bad for a per­son or a sen­tient be­ing in­volves ex­pe­riences, de­sire satis­fac­tion, an ob­jec­tive list of things (e.g. friend­ship, love, ex­plo­ra­tion, etc.), or some­thing we haven’t yet con­sid­ered. The im­por­tant point is that it needs to be a no­tion of well-be­ing or “good for some­one” that is widely com­pel­ling, not just as one defen­si­ble way of how to use the words “good for some­one,” but as a com­pel­ling ac­count of what is best for a per­son (or for a sen­tient be­ing). A suc­cess­ful pro­posal has to give pre­cise an­swers to ques­tions such as which be­ings (or com­pu­ta­tions) mat­ter morally (and how much?), or what the cor­rect stance is on pop­u­la­tion ethics or ag­gre­ga­tion in an in­finite uni­verse. For all these ques­tions, the po­si­tion would have to yield com­pel­ling ar­gu­ments for why to take ex­actly one par­tic­u­lar view as op­posed to other plau­si­ble views. (This may sound overly de­mand­ing, but note that ideal con­di­tions for philo­soph­i­cal re­flec­tion means hav­ing ac­cess to ev­ery­thing one can co­her­ently ask for, in­clud­ing e.g. a well-in­ten­tioned, su­per­in­tel­li­gent or­a­cle AI.)

The main challenge for strong moral re­al­ism in the form of a One Com­pel­ling Ax­iol­ogy is over­com­ing philo­soph­i­cal dis­agree­ment be­tween so­phis­ti­cated rea­son­ers. There are sev­eral the­o­ries of well-be­ing, and sev­eral no­tions of im­par­tial­ity, that are in­ter­nally co­her­ent and highly in­tu­itively ap­peal­ing with re­spect to at least some peo­ple’s in­tu­itions. Th­ese the­o­ries are mu­tu­ally con­tra­dic­tory.[21]

Irre­ducible normativity

The other way I could be­come con­vinced of moral re­al­ism in the sense that I mean it is if, in­spired by moral non-nat­u­ral­ism, I be­came con­vinced that ir­re­ducible nor­ma­tivity is a mean­ingful and some­how ra­tio­nally bind­ing (on some con­cep­tion of ra­tio­nal­ity I cur­rently find strange to en­vi­sion). This would roughly cor­re­spond to moral non-nat­u­ral­ism be­ing true. (Some peo­ple dis­t­in­guish moral rea­sons from pru­den­tial rea­sons, whereas I tend to use the ad­jec­tive ‘moral’ in a broader sense that re­lates to all one’s goals, both al­tru­is­tic and non-al­tru­is­tic, and gen­er­ally to that which mat­ters in one’s life. Since ir­re­ducible nor­ma­tivity also cov­ers ego­is­tic goals, it is broader than nar­row-sense moral­ity.) The challenges I see for a con­vinc­ing ac­count of ir­re­ducible nor­ma­tivity are three­fold:

  1. How to jus­tify the ex­is­tence of a realm of nor­ma­tive facts sep­a­rate from the physical

  2. How to re­li­ably gain epistemic ac­cess to nor­ma­tive facts, should they in­deed exist

  3. Whether ir­re­ducible nor­ma­tivity is re­ally a mean­ingful concept

My next post will fo­cus speci­fi­cally on ir­re­ducible nor­ma­tivity, where I will ex­plain in much more de­tail what I (and more im­por­tantly, oth­ers) un­der­stand un­der the con­cept of ir­re­ducible nor­ma­tivity.


[1] A note on ter­minol­ogy: Some peo­ple in my on­line net­work, par­tic­u­larly on LessWrong, seem to use the term ‘metaethics’ some­what differ­ently from stan­dard us­age. That is, they use ‘metaethics’ to re­fer to what I would call call­ing ‘nor­ma­tive ethics’ (or per­haps the best de­scrip­tion would be “figur­ing out what hu­mans value through philos­o­phy and cog­ni­tive sci­ence”). Within aca­demic philos­o­phy, metaethics is the study of moral claims: what moral claims do or don’t as­sert and whether these as­ser­tions are some­times true. The ques­tions of whether e.g. util­i­tar­i­anism is true, or whether hu­man val­ues are com­plex or not, are less likely to come up in a dis­cus­sion about metaethics. Of course, metaethics is in­di­rectly very rele­vant to all these ques­tions and in­forms, for in­stance, whether in­quiries into find­ing the ‘right’ hu­man val­ues or the ‘right’ ver­sion of con­se­quen­tial­ism are well-posed ques­tions or not. And it seems plau­si­ble to me that, ac­cord­ing to some metaeth­i­cal views, “figur­ing out what hu­mans value through philos­o­phy and cog­ni­tive sci­ence” is in­deed how we should be do­ing nor­ma­tive ethics.

[2] Ac­cord­ing to prag­ma­tism, a brand of philos­o­phy that em­pha­sizes the prac­ti­cal na­ture of ethics/​life/​ev­ery­thing and thereby – so one might ar­gue – blurs the dis­tinc­tion be­tween what is the case and what is prac­ti­cally use­ful, moral claims are ‘true’ not when they de­scribe speaker-in­de­pen­dent moral facts, rules or val­ues, but when they re­sult from “cor­rect pro­cesses for solv­ing prac­ti­cal prob­lems” (Fin­lay, 2007).

[3] Quot­ing from the SEP Mo­ral re­al­ism en­try:

Yet, with the de­vel­op­ment of (what has come to be called) min­i­mal­ism about talk of truth and fact, it might seem that this char­ac­ter­i­za­tion makes be­ing a moral re­al­ist eas­ier than it should be. As min­i­mal­ism would have it, say­ing that some claim is true is just a way of (re-)as­sert­ing the claim and car­ries no com­mit­ment be­yond that ex­pressed by the origi­nal claim. Thus, if one is will­ing to claim that “mur­der­ing in­no­cent chil­dren for fun is wrong” one can com­fortably claim as well that that “mur­der­ing in­no­cent chil­dren for fun is wrong is true” with­out thereby tak­ing on any ad­di­tional meta­phys­i­cal bag­gage.

[4] This crit­i­cism ap­plies to many in­stances where philoso­phers do con­cep­tual anal­y­sis. See also Luke Muehlhauser’s post on con­cep­tual anal­y­sis and metaethics, or sec­tion 6 of this pa­per by David Chalmers. In short, the prob­lems with us­ing con­cep­tual anal­y­sis to es­tab­lish nor­ma­tive con­clu­sions are three­fold: Firstly, there may be no uniquely typ­i­cal set of in­tu­itions about the ‘cor­rect’ us­age of moral ter­minol­ogy. Se­condly, or­di­nary us­age may of­ten be un­der­speci­fied, be­cause most peo­ple are not rigor­ously trained moral philoso­phers. Thirdly, even if the vast ma­jor­ity of peo­ple did use moral ter­minol­ogy a cer­tain way, this would not nec­es­sar­ily mean that they would be us­ing it the most use­ful or most ‘right’ way (pro­vided the moral re­al­ist premise that there is a uniquely right way). As an an­ti­dote to ap­proaches an­chored in the tra­di­tion of con­cep­tual anal­y­sis, Chalmers makes the fol­low­ing pro­posal (“X” refers to con­cepts such as ‘knowl­edge,’ ‘moral,’ and ‘sci­ence’ that are difficult to define):’

On the pic­ture I fa­vor, in­stead of ask­ing “What is X”, one should fo­cus on the roles one wants X to play, and see what can play that role. The roles in ques­tion here may in prin­ci­ple be prop­er­ties of all sorts: so one fo­cuses on the prop­er­ties one wants X to have, and figures out what has those prop­er­ties. But very fre­quently, they will be causal roles, nor­ma­tive roles, and es­pe­cially ex­plana­tory roles.

[5] See also Ka­hane (2013) for the same point ar­gued for at length.

[6] It is im­por­tant to note that Fin­lay uses the ad­jec­tive ‘on­tolog­i­cal’ in a weak sense. Derek Parfit (2011) defended a metaeth­i­cal view he called Non-On­tolog­i­cal Cog­ni­tivism: Both math­e­mat­i­cal talk and moral talk can be ob­jec­tively true, but there are no math­e­mat­i­cal or moral en­tities. Note that, ac­cord­ing to Fin­lay’s ty­pol­ogy, this would still count as on­tolog­i­cal moral re­al­ism be­cause Fin­lay’s defi­ni­tion liber­ally counts ex­ter­nal­ist rea­sons as (ab­stract) ‘moral en­tities.’

[7] Con­struc­tivism as a metaeth­i­cal view is differ­ent from con­struc­tivism as a po­si­tion in nor­ma­tive ethics. Tim Scan­lon (2012), for in­stance, is a con­struc­tivist as re­gards nor­ma­tive ethics; how­ever, his metaeth­i­cal po­si­tion is ob­jec­tivist moral non-nat­u­ral­ism. Scan­lon be­lieves, like Parfit and Shafer-Lan­dau, that there are ir­re­ducibly nor­ma­tive rea­sons about what peo­ple ought to do. He fur­ther be­lieves that con­struc­tivism as an ap­proach to nor­ma­tive ethics helps us de­ter­mine which rea­sons are cor­rect. But the ques­tion to be an­swered in the end is which rea­sons are re­ally cor­rect, rather than which rea­sons are cor­rectly the out­put of a well-speci­fied con­struc­tive func­tion.

[8] Espe­cially the “king­dom of ends” for­mu­la­tion of Kant’s cat­e­gor­i­cal im­per­a­tive sug­gests this in­ter­pre­ta­tion. What fol­lows is my own trans­la­tion from Ger­man (Kant, 1986[1785]). Note that in pro­duc­ing this trans­la­tion, I had to make sev­eral sub­stan­tial judg­ment calls.

The idea that ev­ery ra­tio­nal be­ing is com­pel­led to re­gard it­self as an ar­biter of uni­ver­sal­iz­able norms, in or­der to eval­u­ate it­self and its ac­tions from this per­spec­tive, leads us to a re­lated and ex­traor­di­nar­ily fruit­ful con­cept, namely that of a realm of ends. Un­der a realm of ends, I un­der­stand the sys­tem­atic con­nec­tion be­tween ra­tio­nal be­ings through col­lec­tively shared norms. Be­cause ends are de­ter­mined by the uni­ver­sal val­idity of these norms, it fol­lows that, if one ab­stracts from the per­sonal differ­ences be­tween ra­tio­nal be­ings and from the con­tent of their per­sonal ends, then we can think up a sys­tem­at­i­cally con­nected whole that en­com­passes all ends (in­clud­ing both the ra­tio­nal be­ings as ends in them­selves and the ends that any ra­tio­nal be­ing may set for it­self), i.e., a realm of ends, which we can con­ceive ac­cord­ing to the afore­men­tioned prin­ci­ples.

To be clear, I am not say­ing that Kan­ti­anism is best in­ter­preted as mak­ing claims that are re­lated to cur­rent dis­cus­sions of non-causal de­ci­sion the­o­ries. I think there are sev­eral as­pects of Kan­ti­anism that go against this in­ter­pre­ta­tion. I am only say­ing that there are in­ter­est­ing par­allels, and that, if one wants to, one could make a case for such an in­ter­pre­ta­tion (or ex­ten­sion) of Kan­ti­anism.

[9] Peter Car­ruthers, for in­stance, has ar­gued on con­trac­tu­al­ist grounds against an­i­mals’ hav­ing rights (Car­ruthers, 1992). There are, how­ever, some con­struc­tivists who en­dorse an­i­mals hav­ing rights, most no­tably Chris­tine Kors­gaard (2012).

[10] Alter­na­tively, some­one may have in mind an even more re­stric­tive defi­ni­tion of moral re­al­ism – what is some­times called ‘ro­bust moral re­al­ism’ – that only refers to a sub­type of ob­jec­tivism: moral non-nat­u­ral­ism. This defi­ni­tion fo­cuses on whether there are facts that are ir­re­ducibly nor­ma­tive. (See the sub­sec­tion on moral non-nat­u­ral­ism, as well as posts 2 through 4 in this se­quence.)

[11] Note that some non-nat­u­ral­ists, such as Shafer-Lan­dau, think that moral prop­er­ties are re­al­iz­able by many differ­ent ‘con­stel­la­tions’ of nat­u­ral prop­er­ties. Mo­ral plu­ral­ism as a nor­ma­tive view is ar­guably more at­trac­tive for non-nat­u­ral­ists than it is for nat­u­ral­ists be­cause nat­u­ral­ists seem to be com­mit­ted to a one-to-one re­la­tion­ship be­tween good­ness and some other nat­u­ral prop­erty (Shafer-Lan­dau, 2003, pos. 1215).

[12] As an anal­ogy: Dis­play­ing a par­tic­u­lar image on a com­puter screen is not syn­ony­mous with dis­play­ing a spe­cific con­figu­ra­tion of pix­els on the com­puter screen, be­cause the image – at least when viewed sub­jec­tively at a macro­scopic level with hu­man-level vi­sion – is re­al­iz­able via many differ­ent pixel con­figu­ra­tions. The pic­ture “su­per­venes” on the pixel con­figu­ra­tions. The anal­ogy is im­perfect be­cause an image re­ally is noth­ing more than the sum of its pix­els, and a pic­ture that is slightly but per­ciev­ablly differ­ent from an­other pic­ture is just that, an­other pic­ture. With non-nat­u­ral­ist moral­ity, go­ing from a con­stel­la­tion of phys­i­cal facts that does not form a moral cat­e­gory to a con­stel­la­tion that does form a moral cat­e­gory must make for a sharp bound­ary some­how. How­ever, be­cause we can­not ar­tic­u­late, with refer­ence to phys­i­cal facts alone, what this sharp bound­ary sig­nifies, it seems strange or “queer” to think that such a bound­ary even ex­ists in a mean­ingful and ac­tion-rele­vant sense.

[13] The SEP en­try on moral non-nat­u­ral­ism reads: “There may be as much philo­soph­i­cal con­tro­versy about how to dis­t­in­guish nat­u­ral­ism from non-nat­u­ral­ism as there is about which view is cor­rect. [...] Per­haps the most vex­ing prob­lem for any gen­eral char­ac­ter­i­za­tion of non-nat­u­ral­ism is the be­wil­der­ing ar­ray of ways in which the dis­tinc­tion be­tween nat­u­ral and non-nat­u­ral prop­er­ties has been drawn.”

[14] An ex­cep­tion is Cor­nell Real­ism, a ver­sion of nat­u­ral­ism that holds that moral facts, al­though nat­u­ral, can­not be re­duced to other nat­u­ral facts. Cor­nell re­al­ists claim that their po­si­tion avoids the Open Ques­tion Ar­gu­ment that threat­ens other ver­sions of nat­u­ral­ism. (See also foot­note 16.)

[15] One might think that the Open Ques­tion Ar­gu­ment leaves open the op­tion of moral facts and nat­u­ral facts merely be­ing co­ex­ten­sional, i.e., that they re­fer to the same thing but via differ­ent routes. How­ever, some philoso­phers (and my­self) be­lieve that if two con­cepts are co­ex­ten­sional in ev­ery pos­si­ble world, that just means that they are syn­ony­mous.

[16] Many moral re­al­ists re­ject the pro­ce­du­ral ac­count of ra­tio­nal­ity (which cor­re­sponds to the way ‘ra­tio­nal­ity’ is used on LessWrong: cog­ni­tive skills that help to achieve what­ever goals one already en­dorses) in fa­vor of what they call sub­stan­tive ra­tio­nal­ity. On the sub­stan­tive ac­count, be­ing ra­tio­nal may for in­stance en­tail hav­ing the right dis­po­si­tions to ap­pre­hend or be mo­ti­vated by ex­ter­nal­ist rea­sons for ac­tion. The be­lief in sub­stan­tive ra­tio­nal­ity there­fore tends to go to­gether with rea­sons ex­ter­nal­ism. (And one may ar­gue that this is cir­cu­lar: it defines sub­stan­tive ra­tio­nal­ity with re­spect to ex­ter­nal­ist rea­sons, and ex­ter­nal­ist rea­sons with re­spect to sub­stan­tive ra­tio­nal­ity.)

[17] As a third op­tion, we may ac­cept that nor­ma­tivity is noth­ing over and above the phys­i­cal, but that it can­not be defined in terms of the phys­i­cal eas­ily, any more than a stock mar­ket can be defined in terms of the phys­i­cal. This po­si­tion de­scribes non-re­duc­tion­ist moral nat­u­ral­ism, a view that is usu­ally as­so­ci­ated with Cor­nell Real­ism (cf. foot­note 13), but not limited to it (see, Sin­hab­abu, 2018). How­ever, just like I think there is noth­ing of rele­vance that de­pends on whether we call our­selves “re­al­ists” about the stock mar­ket or not, I fail to see how such a po­si­tion would be rele­vant to our lives if it were true. (I sup­pose it would be rele­vant in­so­far as it may come with meta­phys­i­cal bag­gage of re­ject­ing re­duc­tion­ism in gen­eral, which might change how we ap­proach philo­soph­i­cal ques­tions.)

[18] It need not be ex­actly like Rail­ton’s po­si­tion. Rail­ton en­dorses some no­tion of de­sire fulfill­ment as what is good for a per­son. Another ver­sion of moral re­al­ism, one that I would con­sider to be “strong moral re­al­ism” in the One Com­pel­ling Ax­iol­ogy sense, is moral re­al­ism based on the idea that ex­pe­riences can be in­trin­si­cally morally valuable or dis­valuable. Pro­po­nents of such views be­lieve that phe­nomenolog­i­cal in­tro­spec­tion can tell us about plea­sure’s (moral) good­ness or pain’s (moral) bad­ness (see e.g.: He­witt, 2008 & Sin­hab­abu, 2010). I will dis­cuss moral re­al­ism based on phe­nomenolog­i­cal in­tro­spec­tion in my sev­enth post in this se­quence.

[19] By “ideal con­di­tions”, I am en­vi­sion­ing a sce­nario that is perfectly suited for mak­ing progress on ques­tions of philos­o­phy. Imag­ine a setup that cov­ers ev­ery­one’s needs and also pro­vides ac­cess to all of the fol­low­ing:

  • the world’s best (and most use­fully or­ga­nized) library or on­line library

  • re­vived ver­sions of his­tory’s great­est moral philosophers

  • con­tem­po­rary philoso­phers ea­ger to dis­cuss their is­sues of expertise

  • or­a­cle ar­tifi­cial su­per­in­tel­li­gence in­tent on char­i­ta­bly (and pas­sively) helping out by an­swer­ing any well-posed questions

  • life ex­ten­sion (in case one needs more than an or­di­nary life­time to prop­erly re­flect)

  • ad­vanced nootrop­ics (so peo­ple could think faster or more ac­cu­rately)

  • mind-al­ter­ing tech­nol­ogy (to e.g. ex­pe­rience what it is like to have differ­ent moral in­tu­itions or ex­pe­rience yet-un­known states of mind)

  • etc., things in that spirit.

Fur­ther­more, there would be some mechanism in place to gen­tly break up epistem­i­cally un­healthy group dy­nam­ics (for ex­am­ple, if charis­matic peo­ple’s in­fluence on oth­ers’ opinions was dis­pro­por­tionate). Alter­na­tively, the jour­ney could also be un­der­taken in soli­tude. In gen­eral, we could imag­ine a mechanism in place to pre­vent any­thing that rad­i­cally al­ters the in­tu­itions and goals of our would-be philoso­phers in ways that are not in­tended. Need­less to say, there is no uniquely cor­rect no­tion of “ideal con­di­tions for philo­soph­i­cal re­flec­tion,” and if differ­ent plau­si­ble se­tups lead to rad­i­cally differ­ent re­sults, that would just be an ad­di­tional way in which moral re­al­ism via One Com­pel­ling Ax­iol­ogy could fail.

[20] Some­one may ob­ject that it doesn’t mat­ter to them what the vast ma­jor­ity of peo­ple would con­clude af­ter philo­soph­i­cal re­flec­tion, be­cause they have their own in­tu­itions about what it means to do good for oth­ers, and be­cause moral con­ver­gence is not nec­es­sar­ily the same as moral truth. I think this is a le­gi­t­i­mate ar­gu­ment in a situ­a­tion where peo­ple are pur­su­ing differ­ent ques­tions: If some peo­ple as­so­ci­ate moral­ity with words like ‘ex­cite­ment,’ whereas other peo­ple as­so­ci­ate it more with ‘se­ri­ous­ness,’ maybe that just means they are en­vi­sion­ing differ­ent things and are an­swer­ing differ­ent ques­tions when try­ing to sys­tem­atize their moral in­tu­itions. How­ever, in the One Com­pel­ling Ax­iol­ogy sce­nario, I stipu­late that there is agree­ment about what the ques­tion is, and that peo­ple who are rele­vantly similar to one­self with re­spect to how they ap­proach ex­is­ten­tial ques­tions also con­ver­gence on the same an­swer as ev­ery­one else. In that case, it would be weird to con­sider this fact ir­rele­vant to one’s per­sonal think­ing about what it means to do good for oth­ers.

[21] Note that just be­cause there may not be a One Com­pel­ling Ax­iol­ogy does not mean that we should not ex­pect ideal con­di­tions for philo­soph­i­cal re­flec­tion to be use­ful, or that we should think that there is no differ­ence be­tween ob­vi­ously silly views about what mat­ters and views that are plau­si­ble or defen­si­ble. I think of the differ­ence be­tween well-done and poorly-done moral rea­son­ing as a con­tinuum with differ­ent peaks, rep­re­sent­ing differ­ent ques­tions be­ing asked. Re­ject­ing One Com­pel­ling Ax­iol­ogy only means that we need to put in more leg­work up­front in or­der to de­cide what types of ques­tions we want to an­swer, but it does not mean that ev­ery­thing re­lated to moral-philo­soph­i­cal prac­tice is use­less.


Many peo­ple helped me with this post, but I want to speci­fi­cally thank Si­mon Knutsson for im­por­tant ad­vice on ear­lier drafts that greatly im­proved the di­rec­tion I went for with this post.

My work on this post was funded by the Cen­ter on Long-Term Risk.


Chalmers, D. (2011). Ver­bal Dis­putes. Philo­soph­i­cal Re­view 120(4):515-566.

Fin­lay, S. (2007). Four Faces of Mo­ral Real­ism, Philos­o­phy Com­pass 2(6):820-849.

Foot, P. (1958). Mo­ral Beliefs. Meet­ing of the Aris­totelian So­ciety.

He­witt, S. (2008). Nor­ma­tive Qualia and Ro­bust Mo­ral Real­ism. PhD the­sis. New York Univer­sity.

Ka­hane, G. (2013). Must Me­taeth­i­cal Real­ism Make a Se­man­tic Claim? Jour­nal of Mo­ral Philos­o­phy, 10(2):148-178.

Kant, I. (1986(1785)). Grundle­gung zur Me­ta­physik der Sit­ten. Stuttgart: Reclam.

Kors­gaard, C. (2012). A Kan­tian Case for An­i­mal Rights. In: An­i­mal Law – Tier und Recht: Devel­op­ments and Per­spec­tives in the 21st Cen­tury, ed. M. Michel, D. Kühne & J. Hänni. Dike: Zürich.

Lille­ham­mer, H. (2004). Mo­ral Real­ism: A Defense, Notre Dame Philo­soph­i­cal Re­views. https://​​​​news/​​moral-re­al­ism-a-defense/​​.

Parfit, D. (2011). On What Mat­ters, Vol­ume II. Oxford: Oxford Univer­sity Press.

Rail­ton, P. (1986). Mo­ral Real­ism. The Philo­soph­i­cal Re­view, 95(2):163-207.

Sayre-McCord, G. (ed.) (1988). Es­says on Mo­ral Real­ism. Ithaca: Cor­nell Univer­sity Press.

Scan­lon, T. (2012). The Ap­peal and Limits of Con­struc­tivism. In Con­struc­tivism in Prac­ti­cal Philos­o­phy, ed. J. Len­man & Y. Shem­mer, 226-242. Oxford: Oxford Univer­sity Press.

Shafer-Lan­dau, R. (2003). Mo­ral Real­ism: A Defense [Kin­dle ver­sion]. Oxford: Oxford Univer­sity Press. Retrieved from Ama­

Sin­hab­abu, N. (2010). The Epistemic Ar­gu­ment From He­donism. Un­pub­lished. (Retrieved: May 2018).

Sin­hab­abu, N. (2018). Eth­i­cal Re­duc­tion­ism. Jour­nal of Ethics and So­cial Philos­o­phy, 13(1):32-52.