The Evidential Argument from Moral Agency Revisited: A Reply to Jerry Coyne

1. Introduction

Biologist Jerry Coyne recently blogged about an argument I’ve called the “evidential argument from moral agency” (EMA). The argument was formulated by (then-agnostic, now-atheist) Paul Draper in an article in the American Philosophical Quarterly. (See here for a free copy.) I’ve blogged about the argument twice: see here and here.  It appears that Coyne has only seen the first blog post, which he says he finds easier to read than Draper’s essay, but since he doesn’t actually provide his readers with a link, it’s hard to know for sure which blog post(s) of mine he’s read.

Following the standard jargon in the philosophy of religion for arguments from evil, I called the argument from moral agency an “evidential” argument because the argument does not claim that moral agency is logically incompatible with metaphysical naturalism. Rather, the argument claims that embodied moral agency is evidence against metaphysical naturalism and for theism.

It’s important to remember that the EMA is NOT, as some commenters on Coyne’s site seem to think, a moral argument for God’s existence. In other words, the EMA doesn’t claim, “Morality exists, therefore (probably) God exists.” In fact, the EMA is perfectly compatible with the fact that morality does not (ontologically) depend on God.(To avoid any doubt, I think morality does not ontologically depend on God.) Rather, the focus of the EMA is on moral agency, i.e., the ability to make moral choices and to be morally responsible for those choices.

In my initial blog post about this argument, I wrote, “I’m inclined to believe this is the strongest argument–by far–for theism I have ever read.” At the risk of stating the obvious, when I wrote, “this is the strongest argument for theism,” I was not claiming that theism is probably true or even that embodied moral agency is evidence favoring theism over naturalism. Instead, I made a relatively  modest claim: in my opinion, the EMA is the strongest argument for theism.

2. The State of the Philosophy of Religion

Jerry Coyne’s critique includes both objections to the argument and some very harsh criticisms of the philosophy of religion.I will very briefly address the latter and then address the former in depth.

Since Coyne is a professional biologist, not a philosopher, he is not an expert on philosophy. Like any other non-philosopher, he has the right to state his opinion regarding the quality of argumentation in philosophy as a whole or a sub-disciple of philosophy, such as the philosophy of religion. But his opinions do not carry the weight of an expert, so it would be fallacious to make the following argument from authority: Jerry Coyne thinks the philosophy of religion is dead; therefore, it’s dead. While some arguments from authority can be logically incorrect, this one is not. Non-philosophers do not have philosophical expertise, so the opinion of non-philosophers, including Coyne, provides no evidence at all for the claim that the philosophy of religion is dead. In other words, there is no logically correct argument from authority for the claim that the PoR is dead when the “authorities” are actually non-authorities. The philosophy of religion may or may not be dead as a discipline, but, if it is, that is for philosophers to determine, not non-philosophers.

If this seems harsh, consider the following words written by philosopher Doug Krueger, in another context, about the sort of expertise required to be competent as a potential debate opponent of William Lane Craig.

Moreover, as Mark Smith, an atheist with college debating experience, points out, if Craig is concerned about the academic credentials of his oral debating opponents, he should also be just as concerned about their speaking skills. According to Smith, “To be good at debating one needs years of training and practice. Would anyone hire a chemistry professor to build a house?” Or as Krueger asks, “What biologist would put a concert pianist up on stage to defend evolution? What mathematician would give an English teacher the job of lecturing on the four color map problem? Of course, there are experts in more than one field, and I’m sure that some English teachers would do a better job than some math teachers, but wouldn’t the math field be a more likely area to get experts in that area? Wouldn’t philosophy be a more likely field from which to recruit people to speak on philosophical issues, such as god’s existence?”

Krueger was pointing out some rather obvious flaws in Craig’s choices of debate opponents, but his comments apply equally to the absurdity of Coyne’s dismissal of the philosophy of religion.

In my opinion, Coyne is partially wrong: the philosophy of religion is not “dead,” but it is in serious condition, if not on life support. This can be shown by counting the number of philosophy departments at secular colleges and universities which have faculty lines for philosophy of religion. (They are very rare.) Why is this? I think that one contributing factor to this state of affairs is the blatant partisanship which is very much the norm in the philosophy of religion. Many philosophers of religion, including both atheists and theists, function as natural theologians (if theists) or natural atheologians (if atheists). In other words, they act as if their job description says, “If you’re a theist, defend theism; if you’re an atheist, defend atheism.” It’s rare for philosophers of religion to engage in genuine inquiry and to spend equal amounts of time defending theism and defending atheism. But, if a philosopher of religion is going to act like a philosopher, not an apologist, they should be engaging in inquiry.[1]

It is for this reason that Coyne’s dismissal of the philosophy of religion, based on Draper’s article, is most unfortunate. Draper’s article on the problem of evil and moral agency is one of the most non-partisan articles in the philosophy of religion I have read in a long time. Moreover, not only is not partisan, it advances the discussion in very interesting ways. First, it shows that the arguments from evil and moral agency are parallel problems for theism and for naturalism. Second, he provides a very interesting analysis of the “multiple universes” objection against (many) arguments from evil, an objection which had received very little attention prior to his article. Indeed, he convincingly shows that the multiverse hypothesis fails as an objection to the EMA, whereas it succeeds as an objection to some versions of arguments from evil (for atheism). Third, he shows how the fine-tuning data can be used to support a theistic argument in a way which avoids one of the major problems with traditional fine-tuning arguments. Coyne, however, seems oblivious to these virtues of Draper’s paper and, instead, dismisses without argument part of Draper’s argumentation as ‘mental masturbation.’ This is simply embarrassing for Coyne. Or at least it should be!

3. The Evidential Argument from Moral Agency Revisited

Following Draper, let’s define a “moral agent” as “a being that has moral duties to others and that is, at least in some cases, morally praiseworthy or blameworthy for succeeding or failing to perform those duties.” If we abbreviate “much greater than” as “>!”, “there exist embodied moral agents” as E, “theism” as T, and “metaphysical naturalism” as N, then the argument as formulated by Draper can be summarized as follows.

(1) E is known to be true.

(2) Pr(E | T) >! Pr(E | N).

(3) N is not intrinsically much more probable than T.

(4) Other evidence held equal, Pr(T) > Pr(N).

Note that this is precisely the formulation I provided in my two previous blog posts on this argument.

In light of the fact that both Draper and I explicitly state the argument’s logical form, it’s surprising, not to mention disappointing, to find Coyne misrepresenting the argument and so tearing down a straw man of his own creation. Compare the above, actual form of the argument with Coyne’s version.

1. There are moral agents in the world, i.e., us. By “moral agents,” Draper means that humans have a code of morality and can freely make moral choices.

2. A naturalistic theory of our origins is less likely to explain our status as moral agents than is the existence of God, who made us moral agents.

3. Moral agency requires moral responsibility.

4. To be morally responsible, one must have libertarian free will, that is, at any time one must be able to choose between moral actions and immoral or neutral ones.

5. Such libertarian free will is much more likely to exist under theism than under naturalism.

6. Therefore, moral agency is a strong argument for God.

A quick comparison of the two arguments should make it obvious that they are not identical. Draper’s argument has four premises, whereas Coyne’s argument has six. Draper’s argument is formally valid, whereas Coyne’s argument is not. Furthermore, in addition to misrepresenting the argument, Coyne also misrepresents the definition of “moral agent.” In short, Coyne’s summary is a blatant straw man.

3.1. The Definition of “Moral Agent”

Let’s start with the definition of “moral agent.” Coyne defines a “moral agent” as a “human who has a code of morality and can freely make moral choices.” A careful reading of Draper’s article (and my blog posts), however, should make it obvious that that is not what Draper had in mind. Again, Draper explicitly defined a “moral agent” as a “a being that has moral duties to others and that is, at least in some cases, morally praiseworthy or blameworthy for succeeding or failing to perform those duties.” Draper’s definition makes no reference to human beings nor does it say anything about whether humans “have a code of morality.”

When Draper writes that moral agents “have moral duties to others,” he is not talking about whether they “have a code of morality.” I’m quite sure that Draper is very familiar with — and accepts — evolutionary explanations for morality, viz., explanations for empathy, altruism, reciprocity, rules, etc. Rather, Draper is talking about moral duties which are objective in the sense that they are independent of human opinion. This distinction matters because it could be the case that humans evolved to “have a code of morality,” but that code of morality is nothing but the subjective opinions of human beings.

Like other non-philosopher biologists, Coyne has totally confused the distinction between moral ontology (the nature of moral properties, such as whether they are objective or subjective), moral epistemology (how moral claims and beliefs can be known and justified), and moral psychology (the nature and source of moral beliefs and moral emotions).  Biology and related fields like evolutionary psychology can (and do) inform moral epistemology and psychology, but it’s a category mistake to confuse (and conflate) them with moral ontology. This is important because Draper’s reference to “moral duties” is an implied reference to moral ontology.

3.2. Coyne’s Assessment of Draper’s Premises

(1) There are embodied moral agents, i.e., E is known to be true.

As we’ve seen, Coyne uses a different definition of “moral agent” than Draper. So when Coyne writes, “Yes, people have a moral code and consider themselves moral agents,” it doesn’t follow that Coyne agrees that there exist embodied moral agents. In fact, Coyne denies (1). He writes, “

I don’t believe in moral responsibility because, as Draper notes correctly (and contra Dennett and others), I think that true moral responsibility requires libertarian free will.

Coyne goes on to state his belief that “We don’t have libertarian free will.” Thus, Coyne denies there exist beings who are morally responsible for their actions. This entails that Coyne denies that (1) is true.

(2) Embodied moral agents are much more probable on the assumption that theism is true than on the assumption that naturalism is true, i.e., Pr(E | T) >! Pr(E | N).

As I read him, Draper offers three reasons in support of (2).

(2.1) First, moral agents, including embodied moral agents, have a distinctive dignity or worth. The dignity or worth of moral agents does not raise the probability of their existence if naturalism is true (and hence blind nature has neither the ability nor the desire to select for the evolution of beings with such dignity or worth). If theism is true, however, then such considerations are relevant to the probability of moral agents (since God has both the ability and desire to create such beings).

(2.2) Second, moral agency requires responsibility, which in turn requires moral libertarian free will (LFW). LFW is more likely on theism than on naturalism. (i) Moral responsibility is likely on theism. (ii) LFW requires mental substances. Mental substances are much less likely on naturalism than on theism.

As Draper notes, given that there are moral agents, the fact that they are embodied is more probable on naturalism on theism. He states that they are not much more probable, however, and so the fact that they are embodied does not offset or outweigh the evidence of moral agency for theism.

(2.3) Draper argues that the fine-tuning data “greatly strengthen[s]” the evidence of moral agency for theism because, he says, the fine-tuning data shows that “moral agency is extremely improbable given naturalism.” The argument seems to go like this. Moral agency does not entail life; if moral agency is true, there may or may not be life. Likewise, life does not entail moral agency; if life exists, moral agents may or may not exist. If naturalism is true, however, moral agency “almost certainly depends on the existence of living beings. Thus, moral agency is extremely unlikely given naturalism.” If theism is true, however, moral agency does not require embodied moral agents. (If theism is true, moral agents could exist in the form of unembodied beings, such as angels, demons, spirits, ghosts, and the like.) Not only are there more ways for there to be moral agents if theism is true, but, as Draper argues, theism gives us antecedent reason to expect that God would create moral agents.

What does Coyne have to say about the above reasons?

Regarding (2.1), so far as I can tell, Coyne offers no response.

Turning to (2.2), Coyne seems to agree. He writes, “If we had it [LFW], it would indeed be a kind of miracle, defying the laws of physics, and therefore would require a metaphysical explanation.” As we saw earlier, Coyne denies that we have LFW. But that says nothing about the antecedent probability of LFW on theism and on naturalism.

As for (2.3), Coyne offers no response. Coyne writes, “Draper and Lowder drag the “fine tuning” argument into this issue, but it’s not necessary.” But this comment seems to reveal a misunderstanding of the argument, as if the fine-tuning data were completely irrelevant to the EMA. “Fine-tuning” isn’t “necessary” in order to defend the EMA, but, Draper claims, it “greatly strengthens” the EMA. Thus, a correct response to Draper would be a denial that the fine-tuning does what Draper says it does. But Coyne contradicts himself. He writes:

… at any rate, even if you accept fine-tuning as an argument for God, it doesn’t do anything except make the ‘existence of moral agents’ claim … even less likely under naturalism.

In other words, Coyne admits that (2.3) is correct. Coyne agrees that Draper’s appeal to the fine-tuning data, if correct, performs the exact role in the EMA which Draper says it does.

Coyne also denies that fine-tuning data is extremely improbable on naturalism. He writes “Draper’s paper was written before physicists had provided a number of possible naturalistic solutions to the fine-tuning argument (see here or here, for instance).” The first link is to an audio interview of physicist Sean Carroll. Carroll notes there are four possible responses to the fine-tuning data: (1) design; (2) necessity; (3) chance (what Carroll calls a “brute fact”); or (4) there’s a selection effect. If I understand him correctly, Carroll defends (4). The second link is to a blog post by fellow Patheos blogger Bob Siedensticker which summarizes Carroll’s five objections to William Lane Craig’s fine-tuning argument. Space limitations prevent me from discussing Carroll’s five objections, so instead I’ll simply state my opinion that Carroll’s first two objections are much stronger than the last three.

(3) Prior to investigation, naturalism is not much more likely to be true than theism, i.e., N is not intrinsically much more probable than T.

Coyne denies (3). He writes:

But to me the argument falls down like a deck of cards, for its train of logic is weak. For one thing, it presumes that theism has at least a reasonable probability; that is, that there’s enough independent evidence for God that we can somehow put it into Bayesian probability statements with an appreciable value.

As an objection to (3), this is multiply flawed. First, Coyne confuses the distinction between prior probability and explanatory power. Having independent evidence for God is relevant to explanatory power, but not to prior probability.The fact, if it is a fact, that there is no independent evidence for God (i.e., evidence independent of the evidence of moral agency) says nothing about the intrinsic probability of theism. Furthermore, by itself, the lack of evidence tells us nothing about the hypothesis’s final probability. A hypothesis can be so highly probable prior to investigation that, even after investigating and finding no evidence, the hypothesis can still have a high final probability. (Note: I’m using the word “hypothesis” here to simply mean a proposition that is uncertain.)

Second, Coyne is wrong to claim that, prior to investigation, theism has a negligible probability. In the case of ultimate metaphysical hypotheses like theism and naturalism, we begin by equating “prior probability” with “intrinsic probability.” The intrinsic probability of a hypothesis is its probability independent of the evidence for or against it. Intrinsic probability is determined by two factors: (a) modesty; and (b) coherence. “Modesty” is a measure of how much a hypothesis asserts. The more a hypothesis asserts, the more ways there are for the hypothesis to be false; whereas the less a hypothesis asserts, the fewer ways there are for it to be true. “Coherence” measures the degree to which the logical implications of a hypothesis fit together. If the implications fit together well, the hypothesis is more coherent; if they count against each other the hypothesis is less coherent.[2]

Naturalism says that the physical world explains the existence (or apparent existence) of the mental world. Supernaturalism says that the mental world explains the existence (or apparent existence) of the physical world. Notice the symmetry here: so defined, naturalism and supernaturalism are equally modest and equally coherent.

Theism is a version of supernaturalism. Theism says everything that supernaturalism says, but adds on several additional claims: (a) that the non-physical mental entity which explains the natural world is a person; (b), that person created the world for a purpose; and (c) that person is omnipotent, omniscient, and omnibenevolent. Because theism entails supernaturalism but could be false even if supernaturalism is true, then, prior to examining the evidence, theism is less likely to be true than supernaturalism (and hence naturalism). In other words, the intrinsic probability of theism is less than the intrinsic probability of supernaturalism. Since the intrinsic probabilities of supernaturalism and naturalism are equal, it follows that theism is also less intrinsically probable than naturalism.

While the above analysis shows that theism is less intrinsically probably than naturalism, it does not support the claim, implicit in Coyne’s remarks, that the intrinsic probability of theism is negligible, i.e., that it’s virtually impossible for there to be enough evidence to overcome this improbability.

(4) Other evidence held equal, theism is more probable than naturalism, i.e., Pr(T) > Pr(N).

So far as I can tell, Coyne says nothing which challenges the inference of (4) from (1)-(3).

3.3. Lowder’s Assessment of Draper’s Premises

I now want to comment on Draper’s one-by-one to provide my latest thinking about his argument.

(1) There are embodied moral agents, i.e., E is known to be true.

As noted above, Draper builds moral responsibility into the definition of moral agent. I agree with Coyne and Draper that moral responsibility does seem to require moral freedom. Unlike Draper and Coyne, however, I’m undecided about whether we have LFW. Therefore, I’m equally undecided about whether we are moral agent’s in Draper’s sense.

(2) Embodied moral agents are much more probable on the assumption that theism is true than on the assumption that naturalism is true, i.e., Pr(E | T) >! Pr(E | N).

I am inclined to agree with Draper, on the basis of (2.1). I am undecided about (2.2) and (2.3).

(3) Prior to investigation, naturalism is not much more likely to be true than theism, i.e., N is not intrinsically much more probable than T.

For the reasons given above (i.e., scope and modesty), I now believe (3) is false. It is undeniable that, as the terms are defined above, naturalism is both more modest and more coherent than theism. But that entails that, prior to investigation, naturalism is more likely to be true than theism. In fact, I think Draper may (?) now reject (3).

(4) Other evidence held equal, theism is more probable than naturalism, i.e., Pr(T) > Pr(N).

I think (4) deductively follows from (1)-(3), but since I reject (3) and am undecided about (1), I cannot accept (4) on the basis of (1)-(3).

4. Related Evidence

There are two related pieces of evidence worth mentioning.

1. Even if there are embodied moral agents (in Draper’s sense), the fact that they exist hardly exhausts what we know about them. We also knows facts about the variety and frequency of conditions which severely limit our freedom. These facts are more likely on the assumption that naturalism is true than on the assumption that theism is true. It’s not obvious that the evidence of moral agency for theism outweighs the evidence of moral agency limitations against theism.[3]

2. Likewise, if the life-permitting conditions of our universe are somehow supposed to be evidence favoring theism over naturalism, then the hostility of the universe to life is evidence favoring naturalism over theism.

5. Concluding Thoughts

In summary, I still think the EMA is the best argument for theism I have seen. It is, without a doubt, a massive improvement over cosmic ‘fine-tuning’ arguments for God. Nevertheless, I’m undecided about LFW, so I’m unable to grant that moral agency (in Draper’s sense) is a genuine item of evidence. If LFW and moral agency exist, then I’m inclined to agree with Draper that they are more probable on theism than on naturalism. But since naturalism is intrinsically more probable than theism, it’s far from obvious that the likelihood of moral agency on theism, by itself, outweighs theism’s intrinsic improbability. In other words, it’s far from obvious that moral agency make God’s existence more probable than not.[4]


[1] This point was inspired by an unpublished paper by Paul Draper.

[2] Paul Draper, “Theism, Naturalism, and the Burden of Proof,” unpublished paper.

[3] Paul Draper, “Cumulative Cases,” in Charles Talioferro, Paul Draper, Philip L. Quinn, Blackwell Companion to the Philosophy of Religion (John Wiley and Sons: 2010), 414-24 at 421.

[4] I’m grateful to Paul Draper for helpful comments on an earlier version of this essay.

ETA (7-Jul-14): Inserted footnote 2, which was missing from the original draft of this essay.