Ethics and Theism

Abstract

In this essay I argue that there are necessarily true synthetic a priori moral propositions whose truth does not depend on the existence of God. To make my case, I appeal to an analogy with arithmetic truths such as 2 + 2 = 4 whose truth does not depend on the existence of God. I criticize views like Peter Railton’s that hold that moral truths are like truths about natural kinds such as water and heat, and non-cognitivists who hold that there are no robust moral truths. The point of my criticisms is to answer challenges to my view that there are necessarily true synthetic a priori moral propositions and, in the case of Railton, to block an argument by Robert Adams for a Divine Command Theory of ethics. Second, I argue by example that there can be conflicts between what is best for me and those for whom I care and what is morally required that cannot be reconciled by a theistic ethics. It can be rational to violate moral requirements that have the same contents as the commands of a loving God even if there would be most reason to adhere to those requirements IF God exists, just as it can be rational to leave your umbrella at home even if there would be most reason to take it IF it rained. This will be true regardless of whether the reason to adhere to God’s commands, IF God exists, is because our greatest good is the love of God (and that requires adhering to his commands) or because God will punish you if you do not and reward you if you do. The problem of evil is the primary reason to believe that God does not exist, and so to believe that there are no divine commands that there would be most reason to follow if God did exist.

1. Some Metaphilosophy

2 + 2 = 4. A circle is a set of points equidistant from a given point. A brother is a male sibling. What all these statements have in common is that they are necessarily true. They are true in every logically possible world.

Someone might argue that “2 + 2 = 4” is true in every possible world because God believes that it is true, and God exists in every possible world and believes in all those worlds that 2 + 2 = 4. How would we determine whether necessarily true arithmetic propositions like this one depend on what God believes about them? In general, to determine whether Y depends on X, we should assume not-X and then see whether Y still exists or is true. For instance, to determine whether the existence of some building depends on the existence of some particular cement block in its foundation, we should ask whether it would remain standing if that block were removed. If it would, then the existence of that building would not depend on the existence of that block. So, to determine whether mathematical truths depend on the existence of God, we should ask what would be the case if God did not exist. Would these mathematical propositions still be true? Here is an argument to show that the truth of mathematical propositions does not depend on the existence of God.

• If the truth of mathematical propositions depends on the existence of God, then mathematical propositions would not be true if God did not exist (perhaps they would then all lack truth-value, or perhaps they would then all be false).
• But mathematical propositions would still be true if God did not exist (e.g., 2 + 2 = 4 would still be true if God did not exist).
• So, the truth of mathematical propositions does not depend on the existence of God.

(2) is intuitively obvious, and that is why we are justified in believing that it is true. The same is true of (1). It is based on the meaning of “depends.” When I say that some proposition is intuitively obvious, I mean that it seems true based solely on understanding what it asserts.1 In this sense, it is not intuitively obvious that if you drop a stone, it will fall because that proposition does not seem true based solely on understanding what it asserts. It seems true partly based on the empirical evidence that we have.

This method of argumentation is common in analytic epistemology. People argue that if having a justified true belief is sufficient for knowledge, then in such-and-such a case, a person would have knowledge. But intuitively, the person lacks knowledge despite having a justified true belief. So, having a justified true belief is not sufficient for knowledge. Or they might argue, if someone has a belief produced by a reliable cognitive faculty (one that generally produces true beliefs in the relevant circumstances), then they have a justified belief. But in certain cases, it is intuitively obvious that a person does not have a justified belief even though it was reliably produced (say, where the person has no idea that his belief was reliably produced). So, having a reliably produced belief is not sufficient for having a justified belief.

This form of argumentation is employed by both theists and atheists. It is found throughout Alvin Plantinga’s writings and especially in his Warrant: The Current Debate (Plantinga 1993). By “warrant”, Plantinga means what must be added to true belief to obtain knowledge. He defeats one proposal after another about the nature of warrant by producing counterexamples to them. What justifies us in believing that the statements about the examples are true is that they are intuitively obvious. Suppose Edmund Gettier was an atheist. That would not affect whether his famous counterexamples to the justified true belief account of knowledge were good ones. We are justified in believing the statements about his examples, namely, that they are cases of justified true belief but not of knowledge (and of the many examples that followed in their wake), because those statements are intuitively obvious. Similar remarks apply to examples produced by Keith Lehrer and Laurence BonJour to the view that having a reliably produced belief is sufficient for having a justified belief. Derek Parfit was an atheist, and perhaps the greatest moral philosopher of the late 20th and early 21st century, and he is famous for creating hypothetical cases that elicit intuitions that count against a rival view and in support of his own.

2. There Are Synthetic A Priori Necessary Moral Truths

This essay began with the example, “A brother is a male sibling.” That is an analytic necessary truth, but there are other kinds of necessary truths. Here are some examples of synthetic a priori necessary truths. No object can appear red all over and appear green all over at the same time (Bealer 1998, p. 211). Something has size if and only if it has shape (Bealer 2002, p. 75). If y is downstream from x and z is downstream from y, then z is downstream from x (cf. Boghossian 2022, p. 3). Analytic truths can be reduced to logical truths by substituting synonyms for relevant terms. For instance, the definition of “brother” is “male sibling.” So, after substitution, “A brother is a male sibling” becomes, “A male sibling is a male sibling,” which has the form of a logical truth, namely, A is A.2 But the necessary truths stated in this paragraph cannot be reduced to logical truths by substituting synonyms for relevant terms. These truths are synthetic a priori truths, which are non-analytic necessary truths that can be known a priori.

Here are some candidates for necessarily true synthetic a priori evaluative and moral statements. Happiness is an intrinsic good; suffering is intrinsically bad. Slavery is wrong. Rape is wrong. Betrayal is wrong. It is wrong to fail to save an innocent person’s life if you can easily do that (e.g., by pushing a button, throwing a switch, throwing a life preserver, etc.). Of course, all these statements (with perhaps the exception of the one about suffering) require further qualification for them to be true. Happiness that comes as the result of harming or exploiting others is not intrinsically good. It could be morally permissible to enslave people if the alternative were that some evil ruler would slowly torture them all to death if you did not. It could be morally permissible to rape someone if the alternative were that she would be raped along with 100 other women if you did not rape her. It could be morally permissible to fail to save five strangers if you could easily save them or your own child, but not all six.

The original claim that necessarily, rape, slavery, betrayal, etc., are wrong is false. But the following claim is necessarily true: rape, slavery, betrayal, etc., are wrong unless there are overriding or undercutting defeaters involved. An overriding defeater is some moral consideration that outweighs relevant wrong-making consideration, such as rape, slavery, etc., so the action is not wrong all things considered. In the example about rape, it is the consideration that by raping the woman, you would prevent her and 100 other women from being raped. In the example about slavery, the overriding defeater is that by enslaving the people, you will prevent them from being slowly tortured to death by an evil ruler. Let us assume that if there were no defeaters, it would be wrong to break a promise. But suppose you promised your mafia boss to torture to death the daughter of his hated enemy. The moral weight of that promise would be undercut by the nature of what was promised, namely, to perform some morally reprehensible action. In this case, the moral weight of the promise is undercut, nullified. It is not the case that the promise has some moral weight (a “smidgen” of weight) that is outweighed by some beneficial outcome.3 It has none. It is not necessarily true that happiness is intrinsically good, but it is necessarily true that happiness that is not produced by harming others or acting immorally is intrinsically good. There seem to be necessarily true moral and evaluative statements, even if they are not analytic but are instead synthetic a priori truths. They are necessary truths about what is prima facie wrong or about the intrinsic value of happiness, love, friendship, etc., when appropriately qualified.

There may be a few necessarily true statements about what is all things considered wrong. A favorite of philosophers is, “It is wrong to torture a child to death just for the fun of it.” We may be able to construct an example where torturing a child to death is not all things considered wrong, but the addition of “just for the fun of it” rules out examples where the wrongness of the action is overridden by other morally relevant considerations. Examples in Dostoevsky’s The Brothers Karamazov might be thought to be examples of a type of action that is necessarily wrong. One of his examples is of soldiers tossing babies in the air and then catching them on their bayonets. That would be necessarily wrong if they were doing it just for the fun of it. Another example is of a general ordering his hounds to tear apart a child right in front of its mother because the child threw a stone that hurt the paw of one of his dogs. Assume that the general is not setting his dogs on the child just for the fun of it. Still, inflicting serious harm on an innocent person is prima facie wrong and all things considered wrong if there are no defeaters. In this example, there are no defeaters, so the action is all things considered wrong.

That there have been people who have believed that these candidates for necessary truths were not true, let alone necessarily true, does not mean that they were right. Stephen Stich quotes from Henry Coppée’s Elements of Logic published in 1875 where he states that if you threw a die and some face had not come up for very many throws, it approached certainty that this face would come up in the next throw since it had not come up already.4 But, of course, Coppée was wrong. His reasoning commits what is now called the “Gambler’s Fallacy.” If it is a fair die, the probability remains one-sixth that any face, including the one that had not yet appeared, will turn up on the next throw.

Here is another example that may parallel Coppée’s example but concerns morality: many people do not believe that it is necessarily true that it is wrong to raise and kill an animal simply for the gustatory pleasure one receives from eating it. That does not mean that they are right. This is a case where harm, or death, is inflicted just for the pleasure it gives, which seems comparable to harming, or killing, someone just for the fun of it.

I have offered a Rossian account of what it is for some types of action to be necessarily prima facie wrong, and in Section 1, an intuitionist epistemology to explain how we know they are necessarily wrong.

3. An Argument to Show That Necessary Moral and Evaluative Truths Do Not Depend on the Existence of God

An argument can be given that parallels the argument that mathematical truths do not depend on the existence of God to show that necessarily true normative and evaluative truths also do not depend on the existence of God. It goes like this:

• If the truth of necessarily true normative and evaluative propositions depends on the existence of God, then those propositions would not be true if God did not exist (perhaps they would then all lack truth-value, or perhaps they would then all be false).
• But such propositions would still be true if God did not exist (e.g., it would still be true that it would be prima facie wrong to rape, enslave, betray, etc.).
• So, the truth of necessarily true moral and evaluative propositions does not depend on the existence of God.

This conclusion follows regardless of the nature of the dependence involved. It does not matter whether right and wrong depend on God’s commands or on his goodness or on anything else. If God does not exist, then, if right and wrong depend on his existence, his non-existence would be like removing the keystone from an arch: the truth of all the normative and evaluative propositions would fall to the ground.

4. Are There Some Necessary Normative and Evaluative Truths That Are Neither Analytic nor Synthetic A Priori?

There are necessary truths that are neither analytic nor synthetic a priori truths. It is a necessary truth that water is composed of hydrogen and oxygen and that heat is a function of kinetic molecular activity. Peter Railton has argued that the normative terms like “what there is most reason to do” might have a similar semantics. Railton builds his positive methodological and ontological views about normative matters on an analogy with the methodological practices and ontological commitments in science. Crucial to his approach is the notion of a “job description” or “complex conceptual role.”

Derek Parfit initially says the following about the meaning of “heat.”

Heat is the property, whichever it is, that can have certain effects, such as melting solids, turning liquids into gases, causing certain sensations, etc.

(OWM II, Parfit 2011b, p. 329)5

Railton gives a slightly expanded version of this account of heat. When it comes to the concept “possessing a certain degree of heat,” he thinks that having a degree of heat is connected to the ability to melt solids, turn liquids into gases, etc., (as Parfit first states), but, more fully stated, also includes a connection to the ability to do work, to produce certain characteristic sensations (‘feeling hot’ and ‘feeling cold’) in observers under certain standard conditions, to correspond to temperature difference and changes (‘heat flow’), etc. He says,

Such a job description need not be a conceptual analysis properly so called, since it involves a great deal of substantive empirical content and since any of its particular elements–detailing conceptual interrelations, causal or constitutive connections, paradigm cases, etc.–might in turn be revised or abandoned without leading us to abandon the phenomenological concept “heat” altogether.

(Railton 2017, p. 49; Railton’s italics).

By empirical inquiry we discovered that heat is molecular kinetic energy. So, the degree of heat in something is the degree of that energy in that thing.

Parfit gives a different account of what “heat” means when fully stated. He says, it is:

…the property, whichever it is, that has the different second-order property of being the property that can have certain effects, such as those of melting solids, turning liquids into gases, etc.

(OWM II, Parfit 2011b, pp. 329–30; Parfit’s italics).

The crucial difference between Parfit and Railton on the full job description of heat is that for Parfit it will include reference to a second-order property, but not for Railton. For Railton, there are two concepts, <heat> and <kinetic molecular activity>, which refer to one property, namely, kinetic molecular activity. For Parfit, there are two properties: the first-order properties that constitute the causal basis of heat, namely, kinetic molecular activity, and the second-order property, which is the fact that they have this causal power.

When it comes to the normative concept of “what there is most reason to do,” Railton thinks that the relevant “job description” will include paradigm cases that show that future pain and well-being count, and that pure vengeance and acrophobia do not (Railton 2017, p. 51).6 Without these sorts of examples, Railton thinks we would have no way of testing whether someone has mastered the concept of what there is most reason to do (Railton 2017, p. 51). But he thinks that the “job description” or “complex role” is not so “thin.” The thicker parts include the fact that reasons are capable of motivating people. It also involves the claim that “ought implies can” (in some sense; Railton 2017, p. 56) and that it is inconceivable that causing prolonged agony just for the fun of it is not bad and wrong (Railton 2017, p. 56). Railton thinks that the role of “what there is most reason to do” cannot be “reduced to a naturalistically-characterizable role, and so would not afford a naturalistic reduction of the concept” (Railton 2017, p. 53). For Railton, whatever uniquely (or best) fulfills the job description of the concept “what there is most reason to do” will refer to the property of what there is most reason to do. If it were what maximizes happiness, then the proposition “what maximizes happiness is what there is most reason to do” would be necessary but not analytic, just as “the degree of kinetic molecular motion is what the degree of heat in something is” states a necessary, non-analytic truth (see Railton 2017, p. 52).

Railton gave his first comments on the second volume of On What Matters (OWM II) in Peter Singer’s anthology, Does Anything Really Matter. There was then an exchange between Railton and Parfit in the third volume of On What Matters (OWM III). At the end of that exchange, Railton wrote,

When I wrote my original comment on Parfit’s On What Matters, I didn’t see clearly enough that a certain claim about meaning was central to his main argument, and not an additional argument.

(OWM III, Parfit 2017, p. 126).

Railton goes on to say that Parfit thinks that our “’pre-scientific sense’ of ‘right’ and ‘good’ is not like this,” and the analogy with “heat” is not apt. Moral terms, and those about reasons to act, are not like natural kind terms. Why not? Well, one might hold that the empirical parts of the “job description” of “what there is most reason to do” do not give even part of the meaning of that phrase, nor do they fix its reference, but instead they are parts of the pragmatics associated with the concept <what there is most reason to do>. A tribe of rock dwellers might use the sentence, “The rocks are slick!” to warn people to stay off the rocks or to be extremely careful when climbing on them. That warning may then motivate people to stay off the rocks or to proceed very carefully if they venture out on them. On this view, the fact that using the sentence “The rocks are slick” can serve to warn and motivate people in certain ways is not part of the meaning of that sentence. The meaning of the sentence is that the rocks are smooth and can cause certain objects to slide off them. Being slick is a property of the surface of the rocks. That a person could easily slip when climbing on them is an empirical fact, but not a necessary one. If people had suction cups on their feet, they would not easily slide off the rocks. Assume that frogs have no problem hanging on. People with suction cups on their feet may not be motivated to stay off the rocks and saying that the rocks are slick would not serve as a warning to them. Still, the rocks would be slick; the other features associated with slick rocks are contingent.

Perhaps the fact that causing prolonged agony just for the fun of it is bad and wrong follows from the meaning of “wrong,” just as the truths involved in Gettier examples follow from the meaning of “knows.” The same might be true of “ought implies can” for certain meanings of “ought,” namely, those that imply that you are blameworthy if you do not do what, all things considered, you ought to do. But whatever empirical elements there are in the “job description” of “what there is most reason to do” will be like the empirical element involving the likelihood that people will slide off slick rocks. Those elements do not follow from the meaning of “what there is most reason to do.” They are not part of the concept. That the rocks are slippery follows from the meaning of slick and certain statements of the empirical facts, say, that people do not have suction cups on their feet but normal human skin. That people are motivated to avoid wrongdoing may follow from the meaning of the concept wrong and certain empirical facts about human psychology.

This may also be true if there are necessarily true synthetic a priori statements regarding motivation that, together with facts about human psychology, entail that people will be motivated to avoid wrongdoing. For instance, if it is a synthetic a priori truth that no fully rational person would want to cause agony just for the fun of it, then given the fact that some humans are fully rational people, it would follow that these people would be motivated not to cause such agony. Given that it is necessarily true that causing agony just for the fun of it is wrong, it follows that fully rational people would be motivated not to do wrong by causing agony just for the fun of it.

I have offered an alternative account of how some of the empirical elements in Railton’s reference fixing (job) descriptions might be true, given a different, more traditional semantics of normative terms coupled with certain statements of empirical facts. On the more traditional semantics, “job descriptions” are not relevant to the meanings of terms.

But there are also reasons to reject his view of the semantics of normative terms. Perhaps the strongest reason to reject Railton’s account of the semantics of normative terms is that hypothetical examples can be used to refute, or to at least count against, certain proposals about what is morally required or what there is most reason to do. In Transplant, five people in desperate need of vital organs can be saved if (and only if) one healthy person is cut up and his vital organs are distributed to the five. Assume that you can let a healthy person die in circumstances where you could easily save him, say, by performing the Heimlich maneuver to dislodge whatever he is choking on. Suppose it will never be discovered that you intentionally let this person die. Suppose this happens in a hospital cafeteria where only you and the choking man are present, and where it will soon be discovered that the man is dead, so his organs will be viable and suitable for transplant. Suppose that transplant surgery has been perfected so that there is no chance that the vital organs will be rejected by the recipients. Suppose that you are not a medical doctor or in any way associated with the medical profession, but you know of the situation of the five and know that this man’s vital organs can be used to save them. Then it seems that Act Utilitarianism implies that you should let this man die. That is counterintuitive, and this counterintuition involving a hypothetical case is evidence against Act Utilitarianism.7

But hypothetical cases cannot provide evidence against necessarily true propositions such as “Water consists of H2O” or that “the degree of kinetic molecular motion is what the degree of heat in something is.” You cannot refute, or disconfirm, that water consists of H2O by introducing a hypothetical case where the relevant reference fixing description of water actually refers to XYZ. All that hypothetical case shows is that the reference fixing description of water that we use could refer to XYZ in some other circumstances, not to H2O. Hypothetical cases can be used against what are thought to be analytic propositions, or synthetic a priori ones, but not ones whose semantics are like that of natural kind terms.

Unlike me, Robert Adams thinks that Railton is right about the semantics of wrong (Adams 1999, pp. 15–8, 27; 1979, p. 74) but proposes that what uniquely (or best) fulfills the job description of the concept <wrong> is “being contrary to the revealed commands of a loving God” (Adams 1979, p. 76; 1999, chap. 11). So normative kind semantics could actually be the basis of a defense of a Divine Command Theory of ethics.

It is hard to tell what uniquely (or best) fulfills the job description of “wrong” because we do not know what the proper job description is. But Adams’ proposal seems ill-suited to fulfill that description, whatever it is. Why wouldn’t being contrary to the commands (not the revealed commands) of a loving God be at least as good a candidate? Then theists could make sense of their claim that what is objectively right or wrong may be beyond our ken because not all of God’s commands have been revealed to us. And why is not being contrary to the commands of a loving person as good a proposal as the other two that refer to the commands of a loving God? It is the loving part that seems to carry the moral weight in all these proposals and do all the moral work. We understand how loving people feel and act, but if God’s love is very different, the injunction to be like God would not offer much guidance. Moral principles should offer guidance. God’s revealed commands could offer guidance, but the guidance they offer will depend on our understanding of what a loving person is like, or what commands she would offer. Epistemically, we must rely on our understanding of what a loving person is like in determining what God’s commands require and prohibit or what the injunction “be good like God” requires of us.

Theists commonly say that some of God’s purposes are “beyond our ken.” If God’s love is beyond our ken, then what should we do to “be good like God”? If it is not beyond our ken, then it is because we understand God’s love to be analogous to our love but greater. God is more compassionate, caring, and loving than we are, but the basis of our understanding of this is our understanding of what it is for us to be compassionate, caring, and loving. That is why I say, “It’s the loving part that seems to carry the moral weight in all these proposals and do all the moral work,” that is, in the proposals about the commands of a loving God or the injunction to be good like a (loving) God.

The biggest problem with the first two proposals (that see wrong as contrary to God’s commands, revealed or otherwise) is that if God does not exist, then everything is permitted, which, according to Sartre, is Ivan’s view in The Brothers Karamazov.8 If there is no God, there are no commands of a loving God, nor any revealed commands of such a God. But not everything is permitted if there is no God. Those examples from The Brothers Karamazov are examples of wrong actions, whether or not God exists. No one must wait to see if the problem of evil is a decisive argument against the existence of God before concluding that what the soldiers and the general did was wrong. And it would be foolish to think, “Oh, the argument from evil is a sound argument against the existence of God, so what those people did was not wrong after all.”

Jim Sterba and I disagree about what the best version of the problem of evil is, but we agree that at least one of our versions is a sound argument against the existence of an all-knowing, all-powerful, wholly good God. So, any Divine Command Theory of right and wrong will have unacceptable implications about everything being permitted.

I have argued that fundamental moral principles, and certain propositions about what is right or wrong or about what there is most reason to do in specific cases, are necessarily true. I also argued that if they are necessarily true, their truth no more depends on the existence of God than do the truths of mathematics. I then addressed the worry that normative statements might be necessarily true in the way that statements involving natural kind terms like “water” and “heat” are necessarily true. I raised some objections to the view that there are normative kinds that are similar to natural kinds in their semantics, and then added that even if there were, that would not support a Divine Command Theory of what is morally right and wrong, nor of what there is most reason to do.

5. Non-Cognitivism

I have been arguing that there are necessary moral truths and truths about what there is most reason to do, but the non-cognitivists in moral philosophy think that moral utterances and those about what there is most reason to do, do not express truths, let alone necessary ones. They hold that on the surface, they appear to express truths, but on further reflection, we can see that they do not. Sometimes this conclusion is reached on grounds of disagreement. In Wise Choices, Apt Feelings, Allan Gibbard (1990) argued that some people think that in Prisoner’s Dilemma situations, the prisoners should both confess even though they will end up with a sub-optimal outcome for each. Others think that they should both keep mum, thereby achieving a better outcome for all. Gibbard then argued that either they both mean the same thing by “what there is most reason to do” or they do not. If they do, Gibbard says that at least one of them must be obviously mistaken. They would be like people who mean the same by “square” but disagree about whether some figure is a square. On the other hand, if they mean something different by “what there is most reason to do,” they would be talking past each other. They would be like people who disagree over whether it is safe to put your money in a bank where one person means by “bank” a financial institution and the other the bank of a river.

But Gibbard thinks that fundamental moral disagreement, or disagreement over what there is most reason to do, is not like this. The disagreement does not stem from disagreement over what the fundamental terms mean, nor is one of the parties to the dispute obviously mistaken. His explanation for this disagreement is that it is a disagreement in attitude. One person subscribes to one set of principles about what to do when, and the other person subscribes to a different set of principles. Gibbard thinks that if their disagreement is a moral one, they also hold that their principles are correct and that they would be corrupt if they abandoned them. Their morality or normativity is accompanied by a meta-moral, or meta-normative, commitment.

On my view, the disputants do disagree, but they need not be talking past each other. If they are philosophers, they would try to determine who is right about what there is most reason to do. Is the one right or the other, and if so, why? The disagreement is similar to the one involving the Monty Hall problem. Monty Hall was the host of a game show called Let’s Make a Deal. One of three doors had a big prize behind it, while the other two doors contained joke prizes or prizes of much less value. A contestant would choose a door to open. But before opening that door, Monty Hall would sometimes open a door that did not contain the big prize and then offer the contestant the option of switching to the one remaining unopened door or sticking with the one she had chosen. Some very smart people thought that it was just as likely that the big prize was behind the door the contestant had not chosen as behind the one she had. Other very smart people held that it was more likely to be behind the door the contestant had not chosen. They were right, but it was not true that those who initially disagreed about the likelihoods were either talking past each other or that someone was obviously mistaken. They were not talking past each other, and it was not obvious who was correct. Reasons had to be presented to determine who was right.

A similar situation can occur when people disagree about whether the parties in a Prisoner’s Dilemma situation should do what is best for them, no matter what the other person does, or keep mum. They do not mean the same thing by “what there is most reason to do,” but once they start doing philosophy, they will not be talking past each other, and, at least initially, it will not be obvious who is right. They will try to figure out what they should mean by “what there is most reason to do.”

In Locke’s famous example about the prince and the pauper, he imagines that the soul of a prince is transferred into a pauper and the soul of the pauper is transferred into the prince. People might disagree about whether, so to speak, the prince is now the pauper and the pauper the prince. They might apply different criteria of personal identity to reach their conclusions. Nowadays, philosophers consider brain rather than soul swapping when trying to figure out what the correct principle of personal identity is that determines whether person A is identical to his body, or to his continuing stream of consciousness that now resides in B’s body, or to A’s brain, or to his stream of consciousness that is supported by his brain. They do not want to talk past each other by employing different criteria of personal identity. They want to know what the correct criteria are, which may be different from the criteria that each starts with. They are not going to be satisfied with the idea that they are just expressing different views about personal identity, and they will not hold that they would be confused or somehow mistaken if they abandoned their own current view.

I have discussed the example from Gibbard about the Prisoner’s Dilemma and the expressivist view it leads him to, because I think it is representative of the way people arrive at non-cognitivist views about the normative. Often, those views are reached by unsound arguments that appeal to false but seductive principles. Of course, I cannot go through all the arguments that have led to non-cognitivist views about the normative. All I can do here is to warn readers about arguments that lead to non-cognitivism and to advise them to scrutinize those arguments wherever they find them. But this much seems established: if it is true that there are necessary normative truths, there is no worry that the content of any objective morality depends on the existence of God and what he commands or approves.

6. Why Be Moral?

Let the content of morality be anything other than doing what’s best for you and those for whom you care.9 Call the view that you ought to do what’s best for you and those for whom you care “egoism broadly construed.” Then why should we do what morality requires when what it requires conflicts with what is best for us and those for whom we care, that is, when it conflicts with egoism broadly construed? In emails, Parfit gave Peter Singer an example of such a conflict.

Suppose a man saves his own life and that of his two children by stealing medicine from a stranger who, as the man knows, needs the medicine to save her own life and that of her four children.

(de Lazari-Radek and Singer 2017, p. 283).

It seems there is at least as much reason for the man to steal the medicine as not to steal it, even though it would be wrong for him to do that.

I offer two examples to my classes in support of the idea that sometimes there is at least as much reason to do what is wrong as to do what is morally obligatory. The first one I call Farmworker.

A poor, aging farmworker has worked his whole life, being underpaid and overworked the entire time. An armored truck driver loses his way out in the country where the farmworker labors. His truck hits a bump, and the money spills out the back, unnoticed by the driver as he drives off looking for a route back to the freeway. The farmworker looks all around and sees no one in sight. He puts the money in the trunk of his car and drives to his home in Mexico where his family lives. They have also struggled all their lives, barely having enough to eat, living in cramped, squalid conditions, and being overworked and underpaid. The million dollars that has fallen from the truck will enable the farmworker to retire and his whole family to live a decent life in a comfortable home. Imagine that the money is for Toys for Tots, that this is written on the side of the truck, and the farmworker has read about its being in the neighborhood. Assume that the money was donated by people who would like to see poor children receive a nice $20 Christmas gift. So, 50,000 children will not experience the joy of opening such a gift on Christmas day. From a moral standpoint, I think the farmworker should have turned the money in; that is what was morally obligatory. His legitimate moral grievance over how he was treated in the past does not morally justify him in taking money from children. But I do not think it was irrational for him to steal it, nor that there was more reason for him not to steal than to steal it.

I call my second example Voting. While it is rare that a single vote will decide a state or national election, it is not as rare that it will decide a local election. Sometimes important issues depend on who is elected in these local elections. So, morally, I should inform myself about the candidates, think about which one will most likely support the causes I support, and then vote accordingly. Most of the time, I do not do this. Suppose I just want to watch the Golden State Warriors in the NBA playoffs instead of looking into the candidates. If I do not vote, what I do is wrong, but I do not think it is irrational, nor that there is more reason to inform myself and then vote.

I think that the example from Parfit, and my two examples, show that it can be rational to do what is wrong, whether what is at stake is someone’s life, the joy of many children, or the election of some local official. As the importance of what is at stake increases, so does what is required to make a wrong action rationally permissible: stealing life-saving medicine from others must be counterbalanced by the need to steal to save myself and my loved ones; the disappointment of many children at Christmas by the well-being of myself and my loved ones; and my failure to vote by my pleasure at watching a game. The same is true if what is at issue is what there is most reason to do rather than rationality.

Parfit thinks that sometimes moral reasons outweigh egoistic reasons broadly construed. He writes,

…it would not be rational for me to do something that would be only very slightly better for me, but very bad impartially.

(de Lazari-Radek and Singer 2017, p. 281)

Suppose I am jogging and have with me a large bottle of very cool water. I come upon a little girl who will die of heat stroke if I do not give her some of the water to drink and pour the rest on her to cool her down. However, if I do this, I will be a bit thirsty and uncomfortable for the rest of my jog. In this example, I think Parfit would say that what there is most reason to do is to give my water to the little girl, and I agree. It is also what is both morally and rationally required.10

Sometimes, impartial reasons outweigh partial ones, and moral reasons outweigh egoistic ones. Do egoistic reasons broadly construed ever outweigh moral ones? Parfit’s example seems to show that they can, but some might hold that there is no conflict because morality permits the father to steal the medicine to save his two girls, and egoism broadly construed requires it. Parfit’s Rule Consequentialism might even have this moral implication because it might contain an optimific rule that at least permits parents to harm or even kill others if that is needed to save their own children. Here is what Parfit says about love and why optimific rules should permit (or maybe even require) that parents do actions that are not optimific.

Having such love [for close relatives and some friends] and being loved are some of the greatest goods in most people’s lives. If instead we cared equally about everyone’s well-being, we would have no strong love for anyone, but only what Aristotle called ‘watery kindness’. This would not be how things could go best.

(OWM III, Parfit 2017, p. 420; my comment in brackets)

Since it is best that most of us strongly love our close relatives and friends, optimific rules would reflect this fact. We could not love certain people if we would never choose to do what would be best for these people rather than giving greater benefits to strangers… The optimific rules would also require or permit us to benefit our close relatives much more than we benefit strangers.

(OWM III, Parfit 2017, pp. 420–21).

Such rules might permit parents to run over two deaf and blind children playing in the street if that is needed to get to a hospital in time to save their own child. On the other hand, Act Utilitarians might say that it is wrong to steal the medicine and that there is no conflict between reason and morality because what there is most reason for the parents to do is to refrain from stealing it. Bentham’s “Each child to count for one, none for more than one” might be a principle of practical reason as well as morality. Of course, given the counterintuitive implications of both Act and Rule Utilitarianism, there is no reason to take what they say about morality as the truth. Our moral intuitions about right and wrong are a more trustworthy source of what is right and wrong than these theories.

It is hard to construct an example where intuitions indicate that some action is wrong and there is adequate reason to do it, let alone one where there is most reason to do it. I constructed my farmworker example to do the former, but I do not think it shows that the farmworker is required by reason to steal the money. However, I think that Parfit’s example shows that reason can require a person to do the morally wrong thing. I can imagine someone saying, “If you really, really love your children, you’d be a fool not to steal the medicine.” Sometimes I think this, but not everyone will share that intuition.

Attempts to show that it is impossible for moral requirements to conflict with the requirements of egoism broadly construed seem futile. Parfit’s example and my two examples seem to show that reason does not always favor morality over egoism broadly construed. This-worldly considerations do not always favor morality over egoism broadly construed. So why be moral in such cases?

Theists may have an answer. Suppose the greatest good is loving God and that if you love God, you will obey his commands. Suppose, further, that God commands that we love our neighbor as ourselves (See Vallier 2025, p. 258 for the first two sentences). But what does that require? Well, perhaps it requires adhering to social rules that we “can jointly adopt” (Vallier 2025, pp. 253, 255) within a “rule-based framework of mutual respect” (Vallier 2025, p. 255). Presumably, those rules would prohibit stealing from another where stealing would mean that the other person or her loved ones would die and prohibit stealing money that is meant to buy many poor children a gift. So, the God-based moral rules would not differ in their implications from the apparently obvious moral prohibitions in the examples. But why should we adhere to those rules, given all that we will lose if we do? The theist has an answer: because you will not be following God’s loving commands if you do not, and so in this instance you will be giving up the greatest good you can have, which is loving God and so adhering to his commands.

At most, this shows that if God exists, there is reason not to do wrong. But it does not show that it is rational to adhere to what those commands require. That is because whether it is rational to adhere to them depends on whether (1) it is rational to believe that God exists and makes such commands, (2) that we cannot love God if we do not follow his commands, and (3) that God will not love us if we violate his commands. The problem of evil is the main challenge to (1), that is, to rational belief in the existence of God. Further, and relevant to (2), it does not seem plausible that we cannot love God if we violate his commands; many rebellious teenagers love their parents even when they knowingly violate their rules. And, relevant to (3), it does not seem plausible that God would not love us if we violated his commands. Many parents still love their rebellious teenagers when they violate their rules. Even if loving God is our greatest good, it does not follow that it is rational to follow God’s commands.

People sometimes think that what God rewards and punishes gives us reason to do the right thing. The idea is that God will punish wrongdoers and reward virtuous people, if not now, then in the afterlife. That, they may claim, gives all of us sufficient reason to be moral. Pascal’s Wager was meant to show that we should believe that God exists. This pragmatic argument is meant to show that it pays to be moral.

But what if God does not exist? Then there will be no punishment for wrongdoing or rewards for virtuous conduct. Whether it is rational to believe that it pays to be moral depends on whether it is rational to believe that God exists, on what it is rational to believe about the sort of punishments/rewards he would impose/provide for the type of wrongdoing/virtuous behavior we are contemplating performing, and on what it is rational to believe about what we will gain from wrongdoing as compared to right doing in our world. It is easy to compute the expected payoffs if Jim Sterba’s logical argument from evil is sound, because then there is no chance that God exists; it is harder to compute if I am right, and it is only very likely that God does not exist. But what is obvious is that the rationality of belief in God does not follow merely from its being true that there is more reason to do what is right and to be virtuous than to do what is wrong and be vicious IF God exists. What rationality requires and what there is most reason to do IF God exists can diverge, whether what is at issue is loving God or his rewards and punishments.