Objective Moral Facts Exist in All Possible Universes

Abstract

The question of whether a God is needed to justify or ground moral facts is mooted by the fact that true moral facts exist in all possible universes that contain rational agents. This can be demonstrated in three stages. First, it is necessarily the case that true moral facts can only be described as the imperatives that supersede all other imperatives. Second, it is necessarily the case that for any rational agent there will always be true hypothetical imperatives that supersede all other imperatives. And third, if there are true hypothetical imperatives that supersede all other imperatives, they are then, necessarily, the only true moral facts. As this follows for any rational agent in any possible universe, the presence of God is irrelevant to the existence of moral facts. God could be more capable of identifying those true moral facts, but he cannot author or ground them. And though a God could casuistically alter moral imperatives by altering the corresponding physics, he is constrained in what he can make true this way by moral fundamentals that are always necessarily true. God is therefore not necessary for there to be moral facts.

1. Introduction

Many theists claim a God is necessary to ground or justify moral facts (Kratt 2023; Moreland and Craig 2017, pp. 490–99; Copan 2012; Flannagan 2012; Linville 2009; cf. Craig et al. 2020). This claim has been repeatedly found to be implausible (Mayberry 1972; Murray 2010, pp. 93–100; Morriston 2012; Carrier 2012; Sterba 2024; cf. Craig et al. 2020). But it can also be shown to be impossible. If true moral facts exist in all possible universes that contain rational agents, then God cannot be necessary to ground or justify moral facts. This captures the gist of all critiques of the claim that if moral facts exist, they must be true independently of God’s desires (even if those desires derive from his nature), because otherwise God himself can have no ground or justification for affirming them as true. They would then be arbitrary, negating their status as distinctly moral facts.

This is a reformed Euthyphro dilemma, which defeats not just divine command theories, but divine nature theories as well: one must decide whether God’s nature is good simply because it is God’s, or because it is actually good. Since it must be one or the other, and the former excludes the latter, moral truth cannot be grounded in God’s nature (so, Mayberry et al., cited above). This cannot be evaded by declaring that, for example, ‘God is both good and goodness itself’ as that still leaves the same question of whether that declaration is true and thus what would make it true. There must be some other reason to conclude God’s nature is good, other than that it is merely God’s (and likewise, to conclude God is ‘goodness itself’ other than merely that he is God). Otherwise, for example, a God whose nature was omnimalevolent would entail that causing needless harm was good—such as one might argue the Bible’s authors actually believed (e.g., Genesis 7:4, 17–23; Leviticus 20:13, 24:11–16, 25:44–46; Numbers 15:32–36, 31:15–18; Deuteronomy 2:31–34, 3:1–6, 7:1–3, 12:1–13:16, 21:10–12, 22:13–30; Joshua 10:40, 11:19–20; Judges 11:29–40; 1 Samuel 15:2–3; 2 Samuel 24:1–17; 1 Kings 20:35–36; 2 Kings 2:23–24; Psalms 137:8–9; Job 1:8–2:10; Nahum 1:2–8). Of course there are competing views on this (see Avalos 2005, 2013, 2015 vs. Copan 2011; Bergmann et al. 2013), but the present point is hypothetical not declarative: if these passages did manifest malevolence (in the way many believe), then we could not simply declare them ‘good’ because they were the acts of God. Moral people recoil at the things there depicted for a reason—whatever excuse may be given for them. Indeed, the very need to come up with an excuse for them illustrates the point. Since objective reasons can be given to disobey malevolent moralities, God’s nature being omnimalevolent would not commend it to us (we would have no rational reason to admire or emulate such a nature, least of all to call it ‘goodness itself’). So what commends whichever actual nature God has? Merely being God’s nature will not suffice. It therefore must be something else. But that something else would then have to be true without God.

For example, Linville (2009) argues that morality must derive from human dignity, which must derive from humans being made in God’s image, and that image ‘is’ good, and therefore should be valued. But why is it good? Merely declaring it good does not make it so. Linville’s argument is circular: it presumes the truth of the very thing it is supposed to be proving. If humans ‘have’ dignity in any sense that we ‘ought’ to value it, and not only value it, but above all other things, there must be some reason why we ought to do that—some reason other than ‘God’, because it would have to be the reason that warrants our valuing God’s nature and choices in the first place. ‘Because we are a poor copy of a cosmic tyrant’ does not answer that question. It merely replaces that question with a new one, “Why care about the poor copies of a cosmic tyrant?” Which is just the same question, reworded. So Linville is not actually grounding morality. He is only burying under elaborate emotivist verbiage the fact that he has not (Carrier 2011, p. 428, n. 47).

Two responses can and have been made to these observations.

The first option is to reject the existence of true moral facts and settle upon some other conception of morality (emotivism, prescriptivism, error theory, etc.). But that approach struggles to make sense of why anyone should behave in any particular way at all (the entire point of discerning moral facts). It may be true that moral facts do not exist. Theists making this argument cannot demonstrate that they do because by grounding them in an unobservable being, they remove all access to any relevant evidence that their purported facts exist. What, empirically, is God’s nature? From what we do have empirical access to, it does not look readily commendable (Sterba 2019; Tooley 2019); and while theists have hypotheses by which to fix that, they never have evidence any of those hypotheses are true (Schoenig 2024; Schellenberg 2024; Carrier 2023a). But again, even knowing confidently what God’s nature was would still not suffice to commend that nature to us. So that would still not establish that it entails any moral facts. Some other reason to care about that remains needed.

The second option is to present empirical evidence for the actual existence of moral facts—which will not be evidence of God’s nature or any other feature of God, because we have no other evidence by which to discern that. So this option creates a dilemma for the theist: any empirical argument for preferring one behavior to another simply is declaring a ground for moral facts other than God—that ground being whatever it is that you present as evidence warranting preferring one behavior to another. This is what theists are already doing: they propose that the evidence warranting preferring one behavior to another is something about “God’s nature,” or God’s commands or some other feature of God, but we have no evidence really bearing on what that nature is, or what any real God’s commands are (Carrier 2012; Schellenberg 2015; Anderson 2021). Confronted with this problem, theists will resort to appealing to objective natural facts (e.g., arguing homosexuality is immoral by appealing to some sort of real-world effects of that behavior on society or the person), but in so doing, they have surrendered the argument. They have then admitted moral facts are not grounded in God after all, but in ordinary observable facts. There appears to be no way around this (Carrier 2015a).

The atheist is in a better position. Since it is not possible to know whether God’s nature is moral, without first knowing what is moral, the atheist and the theist are in the same epistemic position with regard to determining what is moral. If an ‘ought’ cannot derive from an ‘is’ then it can no more derive from the ‘is’ of God’s nature than anything else. Whereas if moral facts can derive from God’s nature, then, necessarily, it is possible to derive an ‘ought’ from an ‘is’. But it is then no longer a given that the ‘is’ that moral facts derive from is any thought or property of God. And unlike the theist, the atheist has access to relevant evidence here: the ‘is’ that entails moral ‘oughts’ must be some observable facts of the world or existence itself. The theist could say this, too. For example, if a theist admitted God does not ground morality but only realizes it in his creation, then one need merely observe that creation for evidence of what is moral and why. The theist would then be in exactly the same position as the atheist: looking for the natural facts that entail moral facts (Carrier 2012, pp. 2, 4). God’s role in it all would then be irrelevant to the question of what is moral, i.e., what behaviors we should prefer to others. Because there are still only two options. If there is no truth of the matter, then any nature God had could be tautologically declared ‘moral’, but that would abandon the definition of ‘moral’ as that which we ought to prefer. Whereas if there is a behavior we ought to prefer, that would have to be so whether God existed or not. Because the reasons that we ought to prefer it would remain; it is either best for you to do so, or it is not.

We could conclude this debate if we discovered that moral facts necessarily exist in all possible universes that contain rational agents—that, in fact, ‘moral truths’ are inherent in the definition of ‘rational agent’. The reasons that moral behavior is ultimately best for you are then innate to rationality itself. If that is the case, then the existence of God could never matter to the ground or justification of morality. God could dictate some aspects of what was moral or immoral, by dictating the physics that shapes our circumstances. But that physics would still be dictating what is moral, wherever that physics came from. And there would still be universal principles governing what each physics would entail is moral, principles that would hold for all rational agents in all possible worlds (and thus under all possible physics), and which therefore God could never change.

I believe we can prove this. And therefore we can safely conclude that it is impossible for God to be the ground of moral facts, even if he existed. Rather, rational agency is itself the ground of moral facts. This, of course, entails that moral facts will always exist. There will always be some truth of the matter regarding which behaviors a rational agent should prefer over others.

2. Demonstration

That moral facts exist in all possible universes that contain rational agents can be formally proved by the following argument:

• Premise 1. It is necessarily the case that true moral facts can only be described as the imperatives that supersede all other imperatives.
• Premise 2. It is necessarily the case that for any rational agent there will always be true hypothetical imperatives that supersede all other imperatives.
• Conclusion. It is necessarily the case that for any rational agent there will always be true moral facts.

In other words, if there are true hypothetical imperatives that supersede all other imperatives, they are then, necessarily, the only true moral facts. And, empirically, we can demonstrate (and have demonstrated) that there are true hypothetical imperatives that supersede all other imperatives, which means we have already demonstrated that true moral facts exist (Carrier 2011; Clarke 2019; Bloomfield 2001; Smith 1994; Brink 1989). The work of correctly determining what they all are is a separate project; but that there is something to discover has already been well argued. Here we will examine the consequences of that.

First, the conclusion (that “there will always be true moral facts”) necessarily follows from the premises. Because if true moral facts are the imperatives that supersede all other imperatives, then all imperatives that supersede all other imperatives are (indeed the) moral facts. In other words, Premise 1 describes an identity relation, not a mere predicate relation: if an imperative supersedes all imperatives, then it is a true moral fact. So if Premise 1 is true (along with Premise 2), the Conclusion follows. So it remains only to confirm the premises.

2.1. Premise 1

The first premise is inescapable. That ‘true moral facts can only be described as the imperatives that supersede all other imperatives’ is a definition of moral facts, and one that cannot be superseded by other definitions. Because any other definition you assign to ‘moral facts’ will necessarily include imperatives that do not supersede all other imperatives (otherwise, your definition is analytically identical and thus not an alternative to Premise 1). And since that means other imperatives will supersede some of the ‘moral imperatives’ of your alternative definition, your imperatives cannot be ‘true’, since what it means for an imperative to be true is that you ought to follow it. But if you ought to follow some other imperative instead, then by definition it is not true that you ought to follow yours. Therefore your imperatives are simply not true imperatives. They therefore cannot be true moral imperatives in any relevant sense—because all true imperatives will supersede yours, and therefore ought to be followed instead. And what it is to be a ‘moral’ imperative (as opposed to any other kind) is exactly this: that it overrides all other imperatives in respect to what you ‘ought’ to do. So to be true, a moral imperative must supersede all other imperatives; and doing so is what it analytically means for it to be a ‘moral’ imperative in the first place. So all superlative imperatives must be moral imperatives.

Because the question of moral facts precisely is which behaviors we should prefer to others, and only true imperatives can be true moral imperatives. Hence it follows that only imperatives that truly supersede all others can be true, and all imperatives that supersede all others are moral. There can still be moral imperatives that are equally true (such that there is no fact of the matter which you should prefer, so long as you prefer one of them: Carrier 2011, 425, n. 33). And this is why there can be true nonmoral imperatives: nonmoral imperatives can be true only insofar as they do not supersede but concur with moral imperatives. And they can only concur with moral imperatives by themselves being moral.

For example, it may be a true imperative that you should eat well when you can, and one would not usually describe that as a moral imperative, but it is nevertheless moral that you eat well when you can. The application to a particular question (eating) of a general imperative (what, morally, you ought to do) entails a moral imperative behind the application. It is true that you ought to eat well when you can because (presumably) it is a moral imperative that you avoid causing unnecessary harm (the general end that eating well serves). Likewise, it is not a moral imperative that you be a firefighter; but you only ought to be a firefighter insofar it is a moral imperative that you pursue some occupation, and that you ought to choose one suitable to your nature and circumstances (and not one unavailable or ill fitted). Even that you ought to enjoy life when suitable may be a moral imperative, but that imperative can be equally fulfilled in many ways, so there is no imperative in the matter which you must choose, as long as it is not immoral. In such cases the imperative is how you pursue your goals, not which goals you pursue, when there is no imperative difference between goals (or their timing, or anything else). All behavior is in some way moral or immoral (it either is permitted or not by any true set of moral imperatives), even when there are many equally moral ways to behave. What makes behavior moral is whether it adheres to what supersedes all contrary imperatives (there is nothing you ought do more).

Therefore, it is necessarily the case that true moral facts can only be described as the imperatives that supersede all other imperatives. Anything else called ‘moral facts’ will by its own definition be overridden by these ‘moral facts’ and therefore will not truthfully describe what you ought to prefer. That this is what moral facts are should not be controversial (Moreland and Craig 2017, p. 394). But it has analytical consequences. If it is the case that ‘true moral facts can only be described as the imperatives that supersede all other imperatives’, then ‘imperatives that supersede all other imperatives’ are what we are looking for when we are looking for moral facts. We therefore have an empirical standard by which to objectively determine when a proposed moral fact is true. Any correct definition of a term should analytically provide us with the means of verifying it (as argued by Ayer [1946] 1952, pp. 35–39 and pp. 59–71). And so here we have a universal description of what true moral facts are and thus how to discover them (arguing against Ayer [1946] 1952, pp. 102–19).

2.2. Premise 2

The same analysis unfolds for Premise 2 (that “there are always true hypothetical imperatives” for any rational agent “that supersede all others”). There are only two ways this premise can be false: if there can be rational agents for whom no hypothetical imperatives are true; or if there can be some other kind of imperatives that supersede hypothetical imperatives. Otherwise, if any hypothetical imperatives are true for any rational agent, then there necessarily will be ‘hypothetical imperatives that supersede all other imperatives’ for them (as that follows from the fact of there being true imperatives at all)—unless there are some other ‘truer’ imperatives that are not hypothetical. But no such alternative kind of imperative has been demonstrated to exist.

For example, it might be claimed that Immanuel Kant discovered a ‘categorical’ imperative that supersedes all ‘hypothetical’ imperatives. But Kant’s categoricals reduced to hypotheticals (Foot 1972; Darwall 1985; Carrier 2011, pp. 340–41; Carrier 2015b). A hypothetical imperative is an imperative whose truth depends on the agent having a goal that the recommended behavior perfects (such that ‘if’ you want that goal, ‘then’ you ought to behave as recommended). This means that a non-hypothetical imperative can never be relevantly true—as then by definition there is no goal the recommended behavior perfects, and therefore no reason for any agent to prefer that behavior. Accordingly, when Kant attempted to supply a sufficient reason to want to pursue his categorical imperatives, he unavoidably produced a hypothetical imperative: the only motive any man has to choose Kant’s imperatives was to find a “greater inner worth of his own person” (Kant [1786] 1964, p. 122, with Wolff 1973, p. 211) which is correct in modern scientific terms: (Bergman 2002; Garrigan et al. 2018). Therefore, only ‘if’ everyone wants a “greater inner worth of their own person” are Kant’s imperatives true, which is why Kant insisted that “there is no one, not even the most hardened scoundrel” who does not really want this. It would be more accurate to say that there is no one who would not want this, when they are sufficiently informed and reason without fallacy from that information. As with all knowledge, what is true is what rationally follows from the facts of the matter, whether we know what is true or not, and whether we are being rational or not. Kant’s categoricals are therefore all just hypotheticals: all their overriding behaviors are conditional on this overriding desire (or some other, or a congeries of desires), and anyone who does not see that is simply inadequately informed or reasoning fallaciously. This appears to be true for all imperatives. Without any reason to want to pursue categorical imperatives, rational agents would have no reason to prefer them, and thus those would be incapable of being ‘true’ imperatives—because hypothetical imperatives will always then supersede them (Premise 1).

So the reduction of all imperatives to hypotheticals appears to be unavoidable. In order for an imperative to be true, it must be true that you ‘ought’ to prefer the behavior it recommends to all others, which means it must be true that you would prefer that behavior when relevantly informed and reasoning without fallacy (Smith 1994; Carrier 2011; Clarke 2019). Moral error is thus generally a product of some degree of being uninformed or misinformed, or reasoning fallaciously from information to a conclusion, which is why all arguments over what is or is not moral turn on either of those defects. But if it is never the case, even under your total perfection of knowledge and logic, that you would prefer the recommended behavior, then there can be no sense in which it is true that you ‘ought’ to. To the contrary, if under total perfection of knowledge and logic it is clear that you ought to do something else, then that is the only imperative that actually warrants your adherence. It is therefore the only imperative that is meaningfully true.

One might attempt to avoid this with the claim that you ought rather to do what is uninformed or illogical. But that would abandon any meaningful concern for what is true about what you ought to do (except when acting illogically or in ignorance is a logical and informed thing to do, such as when playing a game or controlling for bias). And this is evident from the very concern for a moral ground in the first place: if there is no logically or factually true reason to be moral given some fact, then that fact by definition cannot ground morality. This is the very point theists are making. So they cannot abandon it to avoid the conclusion now. The only way anything (even God) can “ground” or “justify” moral facts is if that purported ground is factually true and it logically follows from its being true that one behavior is preferable to another. If that does not logically follow, then the offered fact does not ground moral facts.

Hence this is what the word ‘ought’ means in practice (along with all its correlates, in any language): what you would do if you appreciated the grounds for agreeing that it is the best course of action. That you ‘should’ do a thing means you ‘would’ do that thing if you understood why even you would agree it is the best thing to do (Smith 1994), which means what you ‘will’ do ‘when’ you understand that. This is analytically the meaning of all imperative language in practice. Usually we just elide the ‘why’ and leave it assumed, such that to say ‘you should prefer eating well to eating poorly’ simply is to say that it is the best course of action for you and that you would agree if you knew better. In other words, if you fully understood why, and reached a conclusion from that rationally, then it is what you would do—and therefore if you do not, then you are either being irrational or not fully understanding the reasons. That is simply what we mean when we say that. And so it is always implied that some such reasons exist, which is what theists go on about as being the “ground” of morality. If no such reasons exist, then no morality is grounded, and there simply are no moral facts in any useful sense.

That leaves the question of whether rational agents can have no goals that would entail hypothetical imperatives to fulfill. That appears to be impossible (Carrier 2021). By definition, a rational agent must have at least one goal: to be rational. Otherwise, they are not ‘rational’ in any meaningful sense. The goal to be rational entails further goals. For example, an agent with the goal of being rational will also have the goal of ‘knowing whether anything is or is not worth preferring’, because it would be irrational to disregard preferable things if they exist, and that requires knowing whether they exist. A rational agent might for a time not understand this as their goal, but being rational, they could work it out. They thus ‘have’ that goal either prospectively (as what they would prefer if they rationally considered it) or presently (having rationally considered it). Thus even when a rational agent does not realize this, it is still ‘true’ that they would if they rationally considered it, and therefore it is ‘true’ that they ought to have this goal (whether they know this yet or not). In like fashion that goal entails yet other goals (everything necessary to pursue that goal, and any of the other ‘preferable goals’ it then discovers). This cascade of values entails virtues (as persistent goals), such that saying one ought to embody or live by certain ‘moral values’ is itself an imperative proposition, and thus a moral fact (see discussion in Section 3).

Therefore it is always the case, in all possible worlds, that a rational agent will have an array of goals, which entails a system of hypothetical imperatives that supersedes all other imperatives. Whether or when they know this or not, it will be factually true that they ought to pursue these goals. In other words, it will logically follow without fallacy from actual facts that, when they reason it through, they would (and thus ought to) pursue these goals and in the best (the truly most imperative) ways. What kind of moral facts this would entail is not relevant to the present point (see Section 3). The present point is that there will be moral facts for every rational agent in every possible world; and that the moral facts particular to each world will be grounded in certain universal moral facts that will be the same for all rational agents in all possible worlds, those being the moral facts that are true for all rational agents qua rational agent, the one property they share across all worlds (i.e., the goal to be rational, and all the goals that that alone entails).

So Premise 2 cannot be false. It is necessarily the case that for any rational agent there will always be true hypothetical imperatives that supersede all other imperatives. There are no other imperatives, since to be true, an imperative must be hypothetical in order to entail any sufficient reason to prefer it. And there is no possible way an agent can be both ‘rational’ and have no goals. And since goals (including values and virtues) logically entail true hypothetical imperatives, there is no way to have a rational agent for whom no hypothetical imperatives are true. This entails the conclusion: since those hypothetical imperatives will supersede all others, they are moral facts (by Premise 1), and therefore moral facts exist in all possible worlds.

God is therefore of no use in grounding moral facts. To the contrary, to be true, moral facts must follow from the nature of a rational agent itself, and thus cannot be grounded in anything else. Moral propositions simply are propositions about rational agents, which means moral facts will even exist in worlds without rational agents, insofar as ‘what will be true when or if rational agents appear there’ remains the same even in those worlds.

3. Discussion

3.1. Necessary vs. Contingent Moral Facts

It is typically not going to be the case that a rational agent is the only thing that exists in a world. Those worlds will necessarily have to be governed by some contingent physics and physical history, realizing and sustaining that rational agent. And it is likely that there will be many rational agents in any given world (whether by the same process as made the one or by the activity of that one); and, therefore, there will likely be ‘societies’ that each rational agent must interact with. These contingent material facts will then entail contingent moral facts.

For example, in our world, physics and history conspire to make us mortal and fragile, and we cannot as yet correct this defect of our world. We may one day, if we transition to live within digital simulations whose physics we can rewrite at will (Carrier 2020, 2024 with Blackford and Broderick 2014; Wiley 2014). But for now, ‘what is moral’ (the moral facts for current human beings) is causally constrained by present physics. Hence ‘murder’ is ‘immoral’ only because of these contingent facts. If we one day live in a world where killing someone always causes them to immediately rise back from the dead invigorated and healed of all ailments, ‘murder’ would no longer be immoral, as it would then have no immoral consequence (but in fact a moral one). Likewise, if we were an asexual species of civilized jellyfish, there would be no fact of the matter whether any kind of ‘sex’ was immoral for us, as no such organs or behaviors exist for us to concern ourselves with. This is simply an extension of the casuistics of all moral reasoning. For example, in ‘life boat’ scenarios, what is ‘moral’ may differ from other contexts because the physical circumstances and thus the available options differ. What is ‘moral’ to do when you are being attacked by an army is different from what is ‘moral’ to do in peacetime, because the circumstances change and thus constrain your options. This is why killing in ‘self defense’ is widely regarded as moral, and not murder: because the circumstances are different, such as to entail that in those circumstances killing is, unlike usually, the least worst and thus most moral option. This is obvious on any straightforward consequentualism, but even deontologically we would always will there to be a universal law to “never kill unless necessary to save a life,” or whatever rule would preserve the best total outcome that would still respect persons as an end in themselves, and that we would feel better about ourselves embodying (realizing Kant’s “sense of self-worth”).

But underlying (and thus grounding) all these contingent moral facts will be some universal set of moral facts, which facts are here only encountering different physical circumstances dictating how they can be realized. The most universal moral fact of the matter underlying all these cases is something like ‘produce by your choice of behavior the least unnecessary harm’. That rule implicates many other logically entailed goals, e.g., ‘unnecessary harm’ includes harm to yourself (e.g., not just physical injuries to yourself or your welfare but also injuries to your conscience and sense of self-worth) and harm to others, both directly (e.g., do not steal, murder, etc.) and indirectly (e.g., do not allow events to bring needless harms that you could have prevented), and includes every side of a circumstance. For example, how you behave toward a disabled person might rationally reflect how you would want to be treated if you were in their same circumstances. What aid you gift to others might rationally reflect the harm that giving too much, or taking too great a risk, might bring to you and others. And so on. Moral facts ‘on the ground’ will always be complicated, because reality is complicated. But reality still dictates what, in practice, is the right thing to do (the behavior you ought to prefer). Hence murder is only wrong because of what it does; yet what it does is determined by local physics, which can change.

What all these moral facts entail will be the most preferable behavior in particular is therefore going to be defined by all these changing circumstances—e.g., does killing someone harm them. Hence, killing someone to effect a successful surgery on them is not considered murder but saving their life (Murphy 2014). And these determining circumstances can include limitations of knowledge, means, situation, physics, biology, economics, even social systems that you have no option but to work with. But that does not undo the fact that there are still universal moral facts of the matter (like avoiding unnecessary harm). And those facts might even be diverse. The imperative to ‘avoid unnecessary harm’ might be a first-order universal; but there might also be universal moral facts derivable from that.

For example, all moral imperatives, indeed even all other virtues, might be subsumable under three basic moral virtues of reasonableness, compassion, and honesty, and these might be moral virtues (and thus entail moral facts, and thus moral imperatives) in every possible world. For example, we might imagine circumstances (however bizarre) in which the morally best behavior is to be unreasonable; but then that would by that very fact be the most reasonable behavior in that circumstance. Similarly, reasonableness mediates the other virtues (whereby too much or too little compassion or honesty is, in turn, ‘unreasonable’). But it may even be that the cascade of rational steps from ‘we ought to be rational’ to ‘we ought to determine what is preferable’ to ‘we would always prefer to live in a world where we gain more joys from consort with others than difficulties’ would land a rational person, in every possible world, with a superlative value for compassion; and that or other cascades may lead to a superlative value for honesty. Those would simply always be the best ways to live, to achieve the goals that we would always rationally prefer when available, in every possible world.

But these are all just working examples. My point at present is not that I claim to have solved or worked out or proved what is or is not moral here, but only that there is a fact of the matter to be solved or worked out or proved. Moral facts exist. And they exist in all possible worlds. And God has nothing to do with grounding or justifying them. We can be wrong about what they are, or unaware of what they are, or have ideas about them of mixed quality, but we know what it would take to perfect ourselves away from these defects. We know what we are supposed to be looking for, and that it does exist to be found.

For example, we once mistakenly thought it was best to deny women the vote, until we learned we were factually wrong about everything we thought warranted that; while at the same time, we remain rightly confident that those same reasons factually remain for denying toddlers the vote. We can have overlooked all manner of things, that taking back into account would change what we rationally conclude is morally best. But that would not negate the fact that there is a fact of the matter as to what is morally best. The fact that we observe this progress—more knowledge producing improved moralities (Zuckerman 2008; Pinker 2011; Shermer 2015)—supports the conclusion: there is something to find, it derives from the facts of the world, and it is grounded in the inalienable properties of rational agency.

God’s only two possible roles here are that, being supposedly all wise and all-knowing and all empowered, he could have (and morally should have, by any rational calculation of what ‘is’ moral) taught us the best moralities (and even the reasons for them, their actual grounds) everywhere, and from the earliest times (but did not: Avalos 2005, 2013, 2015; Carrier 2016). And God could have (and morally should have, by any rational calculation of what ‘is’ moral) fixed our world’s physics to have given us a far more moral and just world to live in (but did not: Carrier 2020, 2023b; Sterba 2019). But even had God done either, neither would ‘ground’ morality, but only casuistically shape its implementation. Just as we will one day be able to do when we can design our own worlds to live in (e.g., Carrier 2020).

So it remains necessarily the case that, for any rational agent, there will always be true moral facts, in every possible universe. And therefore God is not needed to ground them. Nor could he ground them. Because God cannot change what would be rational for a rational person to do. So there is no way God could ‘unmake’ the conclusion that it is always more rational to do less harm within the circumstances an agent is in, or the conclusion that only the ignorant, the irrational, or the insane will fail to agree with that.

3.2. Emotional Objections to Moral Facts

It will also be of no avail that you ‘do not like’ what the true moral facts turn out to be. Since they would be objectively true (they will be the only actual imperatives a rational agent truly will prefer to every other, and thus truly ‘ought’ to), your opinions to the contrary will simply be false. This means it is also of no avail to complain that, this being the case, moral facts are fundamentally ‘egoist’ (as in, deriving from what a rational agent would want most). Because if egoist moral facts are the only true moral facts, then your desire for ‘other’ moral facts is moot. It would then simply be true that morality is egoist, and therefore all other moralities are false. This objection usually trades on irrational conflations anyway. There is a fundamental distinction between selfishness (caring not for others) and self-interested selflessness (caring for others because it serves your own physical and mental wellbeing). And harming your own conscience and sense of self-worth is a self-interest (just as Kant argued and science has since confirmed, per above).

Theists cannot disagree. For if they claim moral facts are grounded in God because of how our behaviors affect our divinely mediated fate (whether the appeal be to heaven, death, hell, abandonment by God, or whatever the projected outcomes are supposed to be), then that is simply egoism—self-interest is grounding moral facts. If instead they appeal to how immoral behavior injures our minds or souls in some sense that we are supposed to care about, then that is again egoism—self-interest is grounding moral facts. And so on. Theists are stuck on the horns of a dilemma here. They can try to find some self-interested reason to care about their moral facts (thus simply replicating the egoism they had been complaining about: Carrier 2012, pp. 3–4), or they can abandon any reason to care about their moral facts (thus eliminating them from any hope of being true: Carrier 2012, pp. 1–6). We cannot rationally believe moral facts are true ‘even when’ we have no rational reason to prefer them, because that is self-contradictory. Any moral facts that we do have a rational reason to prefer would then, by definition, supersede those, and thus eliminate the theist’s moral facts from the category of being true.

But the complaint is unwarranted anyway. The fear, I suppose, is that if moral facts are egoist, then the only true morality will be ruthlessly selfish (something like the morality of Ayn Rand, say). Of course, if that is the moral truth, then it simply is the moral truth. Disliking it will make no difference to whether it is true. And all objections of this kind fall to this same point: no matter how weird you find the results, they remain necessarily true. The syllogism leading our Demonstration is valid and sound. If it then should entail the true moral facts are in some way weirder or more vexing than we thought, then we were simply wrong about what was morally true.

But this worry is not well founded. The fact that theists (just like atheists) can adduce a plethora of reasons why a ‘ruthlessly selfish’ morality (or any supposedly weird morality) would be bad suggests that the true moral facts are not ‘ruthlessly selfish’ (or in whatever way weird), but in fact are pretty much what most of us suspect they are: a rational interest in acting to produce a better world for everyone, including oneself. Moral truth cannot be unnecessarily harmful or unreasonable, because any behavior that was unnecessarily harmful or unreasonable would fail to supersede behaviors less harmful or more reasonable. The latter would then be the moral truth. And as it happens, the virtues of compassion and honesty, for example (and not a mere pretense of them), are better for oneself (in fact essential to mental and physical wellbeing), as has been established by the findings of game theory and human psychology (Carrier 2018, summarizing Axelrod 1997; Bergman 2002; Fedyk 2017; Churchland 2011; Narvaez and Lapsley 2009; Garrigan et al. 2018).

That this will be the case in every possible world can be deduced from a cascade of values derived from pure reason (Carrier 2021; Smith 1994). Purely rational agents with access to all options (e.g., agents who could rewrite all their emotional responses at will) will always choose compassion and honesty over discompassion and dishonesty because the one produces more joys and fewer miseries than the other, both internally (enjoying a more pleasant and contented conscience) and externally (enjoying a more cooperative society). You could doubt this, though it would be odd for a theist to doubt that moral facts are rational. But there are reasons to allay those doubts.

In the first case, becoming the sort of person you hate will rationally entail hating yourself (because inconsistency is irrational), leading to increasing personal discontent (a deprivation from Kant’s “sense of self worth”), whereas becoming the sort of person you like will entail liking yourself, leading to increasing personal contentment (hence Bergman 2002; and Garrigan et al. 2018). There is no other logically possible way to access these preferable outcomes, other than by self-deception; but self-deception entails false beliefs, and no true moral facts can follow from false beliefs. False premises cannot secure true conclusions. And our only concern here is with what is true; not what can be falsely believed.

Which is what theists are already arguing when they appeal, for example, to the role of human conscience and the damage immorality inflicts upon your inner person. That is the same thing; and will be the same in all possible worlds. For example, a hypothetical rational agent capable of choosing what emotions to have will always find contentment preferable to dissatisfaction, and will always find more and easier paths to contentment by cultivating compassion and honesty. This is why psychopaths are so discontented (Love and Holder 2014; Blasco-Belled et al. 2023; Doren 1996; Reid et al. 1986; Sinnott-Armstrong 2007, pp. 119–296, 363–66, 381–82): their mental illness hinders their access to the joys of compassion and other virtues (directly, in lost affect; and indirectly, through a poverty of fulfilling or even effective relationships: Seto and Davis 2021; Martens [2006] 2014), and drives them to irrational behavior against their own self-interest (Vaurio et al. 2022; Beaver et al. 2014). This is also why it is imperative to be rational: because irrationality is always inevitably counter to self-interest (Carrier 2011, pp. 426–27, n. 36). It therefore can never be imperative in any possible world.

In the second case, the net outcome of different arrangements of social interaction upon oneself is logically the same in all possible worlds—because there are only so many ways a society can interact, and the effects are always the same. For example, strict tit-for-tat reasoning always leads to inevitable death-loops of feuding, diminishing access to the benefits of cooperation, whereas strict leniency leads to inevitable death-loops of exploitation, also diminishing access to the benefits of cooperation (Axelrod 1997; Fedyk 2017). Computer models show that the ideal behavior in all possible systems is a ‘modulated’ tit-for-tat: always start nice and only punish defectors, but add in a rationally determined amount of proactive forgiveness (rather than always punishing defectors, sometimes we should trust and cooperate instead of retaliating, i.e., sometimes we should meet hostility with kindness) and a rationally determined amount of targeted spitefulness (to defeat would-be manipulators of forgiveness by switching back to a ‘never-cooperate’ strategy with repeated defectors, i.e., sometimes we should never cooperate).

From this can be deduced the conclusion that compassion and honesty modulated by reasonableness will always produce the statistically best outcomes for you no matter what social system you are dropped into (Axelrod 1997, 2006). And habituating in yourself the sentiments of reasonableness, compassion, and honesty will maximize your compliance with this strategy, and thus ensure the best available outcomes for you, in terms of the goods you can acquire from or lose to any social system you are in. This conclusion is unchanged by converse outcomes, since the reason this strategy is more rational is that it statistically produces better results; so that it occasionally fails does not matter to the point, because there is no alternative strategy that would reduce that rate of failure. Hence, the fact that being reasonably nice ‘can’ get you killed or robbed, for example, is irrelevant to the fact that there is no other strategy with a lower risk of getting unjustly killed or robbed (or suffering some other commensurate deprivations). Any ‘tweak’ to the strategy that would mitigate that risk ends up right where computer models find it: modulated tit-for-tat. Trying to avoid worse outcomes (too many exploiters, or too many feuds) always ends at the peak available strategy, which is the one here described. And that strategy will be most reliably enacted by the reasonably compassionate and honest.

Therefore, the complaint theists have here is moot. Their own morality is either egoist or not true. And the findings of logic and science are that what we already tend to agree is moral is the best-performing egoist behavior. Compassion, honesty, and reasonableness lead both to you being more satisfied with who you are and who you have become and (more often) to better material outcomes for you in your interactions with society. That you can avoid this correlation with irrationality (such as adopting false beliefs about yourself, convincing yourself that you actually like mean and dishonest people or the poor outcomes you are producing, or that you are not cruel or dishonest or unreasonable when in fact you are) does not remove the truth of it (what you would think of yourself when your beliefs about yourself are true, and thus what is actually true about yourself). So insofar as what we want to know is what the true moral facts are, all the ways that that truth can be avoided do not matter. What the true moral facts are remains as it is.

3.3. Analytical Concerns

Might my account of moral facts still be compatible with divine command theory? The answer depends on what one means by “divine command theory.” As explained in the foregoing analysis, as a rational agent God can discover and thus could report to us what the moral facts are. But those facts would not thereby depend on God. So they do not gain their status as moral facts simply by being God’s commands. Remove God from the system, and all moral facts remain the same, just as detailed in the foregoing analysis. God is not needed to command them. They would be discoverable to any rational agent, like human beings. God would only have an advantage of superior faculties, so he has greater epistemic access to what the moral facts are. But he does not ontologically create them—except insofar as he can change the physics that entails them (and thus change what they are), but as explained in the forgoing analysis, that, too, does not make God necessary for the ensuing moral facts to be true, because those new moral facts still follow from the physics, not from God, so that if accidents, or humans, or anything else made those same changes to physics, and not God (for example in computer simulated worlds, as just discussed), then the moral facts would likewise change in exactly the same way, and remain just as true. It does not matter where any world’s physics comes from: it always entails certain moral facts regardless. Thus God is never necessary for moral facts. Even if God necessarily existed (as unlikely as that may be), his existence would still not be necessary for moral facts to exist. They would exist anyway. Because (as demonstrated) they derive from other things.

Another question is whether we could argue that there must be other ‘defining’ features of moral imperatives than what is stated in my Premise 1, such as that moral imperatives must concern the wellbeing of others. But this mistakes the definition of moral facts (how we distinguish facts as moral) for the reality of moral facts (what actually turns out to be a moral fact). Whether ‘concern for the wellbeing of others’ is moral is a question of fact, not of definition. Because if some imperative supersedes that one, then that one by definition cannot be true—and therefore cannot be a fact at all, moral or otherwise (as previously explained). What the moral facts turn out to be is an empirical question, not analytical. Moral facts cannot be arbitrarily defined into existence, and therefore ‘what is’ moral cannot be part of the definition of moral facts qua moral facts. Whether moral facts do meet such a criterion of ‘concern for others’ (or indeed any other) simply has to be discovered. I presented in the foregoing analysis some evidence that concern for others is indeed a component of moral facts in this (and probably any) world. But it is analytically possible that, for example, only selfishness is moral (and, for example, only Ayn Rand’s moral system is true). That possibility cannot be excluded by tautology. It can only be excluded by evidence. The same is the case for whether there are behaviors pertaining only to the self (like smoking or suicide) that are moral or not.

One might also ask whether this entire analysis of moral facts presupposes that moral facts must be rational. It does not presuppose but entails this. Rational beliefs and desires are by definition those beliefs and desires that are arrived at from true facts without fallacy. Moral facts must be (or be capable of being) justified true beliefs. So true facts cannot be reliably ascertained by fallacy. Only rationally obtained facts can be regarded as ‘true’ (whether those be moral facts or any other). Indeed there is no way to reliably ascertain which facts are true, even to serve as premises in discovering moral facts other than to do so rationally—as in, to do so without fallacies of logic. This is why, insofar as one is able, “it is imperative to be rational” (as was stated, with citation to a demonstration, in Section 3.2). After all, if moral imperatives, to be moral facts, must supersede all other imperatives, then irrationally derived claims to moral fact can never be true, as they would always be superseded by imperatives that are true: the ones derived without fallacy. Hence “uninformed or illogical” moralities will always be superseded by informed and logical moralities (as was argued in Section 2.2).

Another question is whether this analysis conflates or ignores the distinction between moral values and moral imperatives. Are those different kinds of moral facts? This was addressed in Section 3.1 (with supporting examples in Section 3.2, including citations to corresponding science). In the cited scientific literature, moral values are moral desires, and moral virtues are habituated (persistent) moral desires (and sometimes those terms are interchangeable). But moral facts can only meaningfully be defined as what we ought to do (imperative propositions). Moral values or virtues that entailed no imperatives would fail to be describable as moral. For example, if ‘knowledge is good’ entails no behavior in respect to knowledge, then that statement becomes irrelevant to morality. Moral virtues (as dispositions of character embodying moral values) do increase the probability of adhering to the moral imperatives that they entail (e.g., as already cited: Narvaez and Lapsley 2009; Bergman 2002; Churchland 2011; Garrigan et al. 2018). But they are not analytically required for that. It is analytically possible (even if empirically unlikely) for someone to ‘behave morally’ absent any guiding values or virtues, simply by always obeying true moral imperatives. So the role of moral virtues is instrumental and not grounding. What makes a value or virtue ‘moral’ is precisely what behaviors it rationally impels (see the discussion of psychopathy in Section 3.2 and of goals as value cascades in Section 2.2). Hence, as already discussed, it may even be a moral imperative that we “habituate” moral virtues (hence moral desires or ‘values’) in ourselves (like “the sentiments of reasonableness, compassion, and honesty,” per the conclusion of Section 3.2). Thus virtue theory itself reduces to a system of hypothetical imperatives (Foot 1972; Carrier 2015b). Even when we speak of ‘moral persons’ we are speaking of persons with dispositions to behave morally. Behavior thus defines what is moral.

Another analytical concern is whether ‘imperatives’ can even be true or false in the grammatical sense (like the sentence ‘feed your child’), and so we should instead be talking about ‘obligations’ or ‘duties’ (which then entail imperatives). But as stated at the end of Section 2.2, the word ‘imperative’ here is not the grammatical but the philosophical term for ‘imperative propositions’, e.g., ‘hypothetical imperatives’ and ‘categorical imperatives’, which can be true or false, and indeed the conditions of their being true or false is discussed in the literature on them here cited. As such, duties or obligations are just tautological restatements of imperatives, not distinct assertions. The sentences ‘you ought to feed your child’ and ‘you have a moral obligation to feed your child’ are simply different ways to state the same imperative proposition. So this leaves no distinction relevant to the present analysis.

4. Conclusions

Emotional and casuistic objections to the conclusion that true moral facts exist in all possible worlds are therefore ineffectual (Section 3.1 and Section 3.2); and there are no remaining conceptual objections (Section 3.3). If true moral facts turn out to not be what we wanted them to be, we were simply wrong what the true moral facts are. And science (e.g., psychology) and logic (e.g., game theory) entail they are close to what we already suspect anyway. Moral disagreement is thus entirely a product of ignorance or irrationality. For example, the Christian belief that homosexuality is immoral is based on false beliefs about homosexuality and its effects on persons and societies; or on a fallaciously selective reliance on ancient primitive cult taboos; or both. When these ignorances and irrationalities are corrected, Christian belief moves toward factual reality: that homosexuality is not immoral (Lorens 2018; Valk 2024). This evidence of personal and societal progress suggests there is a truth of the matter, and that it derives entirely from a rational response to the material facts of the world and the people in it.

This suspicion is logically confirmed when we define true moral facts in a way that actually allows us to empirically discover them (Premise 1, “true moral facts are imperatives that supersede all other imperatives”), and then recognize that they must be hypothetical imperatives (Premise 2), because no other imperatives entail reasons sufficient to prefer the behavior they recommend to the behavior already recommended by true hypothetical imperatives. Since all rational agents in all possible worlds have a true hypothetical imperative (to be rational), which entails an array of other true hypothetical imperatives (enacting rational behaviors toward discovering a knowledge of what is preferable, and then enacting rational behaviors to obtain what is thus found to be preferable), true moral facts will exist in all worlds with rational agents, whether God exists or not. Nor could God change what a rational person will ascertain to be rational. And pursuing this fact empirically and logically appears to lead to widely promoted moral virtues like compassion, honesty, and reasonableness. So recognizing it poses no threat to individuals or society. To the contrary, it can only improve them.