The Evidential Argument from Moral Agency (AMA) Revisited

I want to revisit Paul Draper’s very interesting argument from moral agency against metaphysical naturalism.[1]

Informal Statement of the Argument

We know that moral agents exist. If we ignore for a moment the evidence for moral agents–i.e., independent of the evidence for moral agents–we have much more reason on theism than on naturalism to expect the existence of moral agents.

Let us start by considering metaphysical naturalism. If naturalism is true, then it is extremely improbable that there exist such things as unembodied minds like souls, spirits, ghosts, etc. Thus, in a naturalistic world, the existence of moral agency virtually requires the existence of living beings. Therefore, on the assumption that naturalism is true, we would expect that any world without (physical) life would also be a world without any moral agents.

Now consider theism. On the assumption that theism is true, we have several reasons to expect moral agency, reasons we do not have on the assumption that naturalism is true. First, moral agents, by definition, have moral value. That fact is irrelevant to the probability of their existence on naturalism, but it is relevant to the probability of moral agents on theism, which posits a personal God who is, among other things, perfectly morally good.

Second, moral agency requires moral responsibility, which in n turn requires libertarian free will. We have much reason to expect libertarian free will on the assumption that theism is true than on the assumption that naturalism is true.

Third, thanks to recent discoveries in physical cosmology, we now know that our universe is incredibly well “fine-tuned” for life. By itself, the fine-tuning data shows that the prior probability of life given naturalism is extremely low. But when we combine the fine-tuning data with the facts of moral agency, the argument becomes much stronger: embodied moral agency is extremely improbable on the assumption that naturalism is true.

Formal Statement of the Argument


>!: much greater than
E: there exist embodied moral agents
T: theism: the hypothesis that there exists an omnipotent, omniscient, and morally perfect person (God) who created the universe.
N: metaphysical naturalism: the hypothesis that the universe is a closed system, which means that nothing that is not part of the natural world affects it.

Argument Formulated:

(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).

Objects to the Argument from Moral Agency (AMA)

I’m  going to consider various objections to AMA’s premises, in a roundabout order.

Objections to (2)

Objection: Theism is not needed to explain the data about moral agency, including fine-tuning. This is because the hypothesis that there exist multiple or even infinite universes explains the data as good or better than theism. Since the physical laws in each of these universes are random, there is bound to be at least one, if not many, life-permitting universe. We just happen to live in (one of the very few) life-permitting universes.

Reply:  At first glance, the multiverse hypothesis (M) seems irrelevant to the argument from moral agency, since (2) compares the antecedent probability of E on T to the antecedent probability of E on N, not N conjoined with an auxiliary hypothesis about the multiverse. So how could M be relevant to (2)?

Those of you who have read my other recent postings can probably predict what I’m going to write next. Using the probability calculus, we can measure the effect that an auxiliary hypothesis like M has on Pr(E/N). In order to assess the evidential significance of an auxiliary hypothesis like M, we would simply need to consider a weighted average, as follows:

Pr(E/N) = Pr(M/N) x Pr(E/M&N;) + Pr(~M/N) x Pr(E/~M&N;)

This formula is an average because Pr(M/N) + Pr(~M/N) = 1. It is not a simple straight average, however, since those two values may not equal 1/2.

The weighted average formula above gives us some insight into what would need to be the case in order for M to be a good defeater for the argument from moral agency. I assume we all agree that the second half of the right-hand side of that equation, Pr(~M/N) x Pr(E/~M&N;), is not going to be useful for deriving a high value for Pr(E/N). (Otherwise, there would be no need to introduce M in the first place!)

So we’re stuck with the first half of the right-hand side: Pr(M/N) x Pr(E/M&N;). In order for M to be a good defeater of the FTA, then, Pr(M/N) x Pr(E/M&N;) needs to be high, the higher the better. The problem, however, is that we have little or no reason to believe that Pr(M/N) is high, i.e., we have little or no reason on naturalism (alone) to expect multiple universes. If Pr(M/N) is not high, then there is no reason to believe that Pr(E/N), as a weighted average of Pr(M/N) and Pr(~M/N), is high. So, unless there is independent evidence for M–i.e., evidence that is independent of the evidence for E–it appears that using M as a defeater against the argument from moral agency fails.

Furthermore, not only can we calculate the impact of M on Pr(E/N) as a weighted average, we can also calculate the impact of M on Pr(E/T) also as a weighted average.

Pr(E/T) = Pr(M/T) x Pr(E/M&T;) + Pr(~M/T) x Pr(E/~M&T;)

In order to have a true “apples-to-apples” comparison, we would need to compare both Pr(E/N) and Pr(E/T) as weighted averages. On the assumption that theism is true, it is far from obvious that God would create only this universe.[1] This is significant because the higher the value of Pr(M/T), the closer the value Pr(E/T) will be to Pr(E/M&T;). Since we are ultimately interested in the ratio of Pr(E/T) to Pr(E/N), it seems far from obvious that M significantly decreases that ratio. (In fact, I wouldn’t be surprised if the “apples-to-apples” comparison ultimately strengthened the argument by increasing the ratio of Pr(E/T) to Pr(E/N), but I don’t need to show that in order to refute the objection based on M.)

Objections to (3)

Objection: (3) is false. N is intrinsically much more probable than T. N only asserts that supernatural beings do not exist, while T asserts that a very specific kind of supernatural being exists. Thus, N asserts much less than the proposition that nothing at all exists, while T says much more than the proposition that something exists. Thus, N is intrinsically much more probable than the proposition that nothing exists, while T is intrinsically much less probable than the proposition that something exists.[1]

Reply: I find this objection very persuasive. I am tempted to conclude that this objection is also decisive, but theories of intrinsic probability are extremely controversial. On the other hand, theories of intrinsic probability are extremely controversial. And I know of no non-question-begging reason to believe that (3) is true. Even without a decisive or conclusive argument for the conclusion that N is intrinsically much more probable than T, this objection does constitute a solid reason for doubting (3).

Objections to (1)

Objection: This argument depends upon the assumption that we have libertarian free will, which is dubious. Only 16% of professional philosoph

ers believe we have libertarian free will; the rest are either compatibilists or determinists.

Reply: I have called the argument from moral agency the “best argument for theism” without calling it a “good argument.” Part of the reason why is that I am undecided between libertarianism and compatibilism. If libertarianism is false, then so be it. But what interests me about the argument from moral agency is how it directly solves one of the major defects of what I call ‘standard’ fine-tuning arguments.

Standard fine-tuning arguments just assume that God would want to create embodied moral agents, but fail to provide a reason why we should believe that. In contrast, the argument from moral agency directly deals with that problem. It provides a plausible explanation for why, if theism is true, we should expect moral agents. This does not show that the existence of embodied moral agents is more probable than not on the assumption that theism is true. What it does show, however, is this. Given that moral agents of any kind (i.e., embodied or otherwise) is antecedently much more probable on theism than on naturalism, the fact that there are embodied moral agents is much more probable on theism than on naturalism. (Again, please review what I wrote earlier about the ratio of Pr(E/T) to Pr(E/N).) That is the inference that I find particularly interesting (and persuasive).

Back to the objection: I think the objection correctly points out that the argument from moral agency seems to have the concept of libertarian free will built into the definition of “moral agent.” If that is so, then we can’t use our knowledge of our own existence, by itself, as evidence for (1). We also need evidence that we possess libertarian free will. In my opinion, this is probably the strongest objection to the argument. (This assessment could obviously change in response to a robust defense of libertarian freedom.)

Objections to (4)

Objection: The fact that embodied moral agents exist hardly exhausts what we know about moral agents. We also know facts about the variety and frequency of conditions that severely limit our freedom, facts which are antecedently more probable on N than on T.[2] Thus, one can’t conclude that T is true, solely on the basis of moral agency, without committing the fallacy of understated evidence.

Reply: AMA, by itself, doesn’t commit the fallacy of understated evidence, since its conclusion is merely that moral agency provides prima facie evidence favoring T over N, not ultima facie evidence. I consider it an open whether the fully stated evidence regarding moral agency–in other words, whether the general fact of moral agency combined with the more specific facts about the variety and frequency of conditions that severely limit our freedom–favors T or N.

Conclusions and Prospects

If I were to rank the objections from strongest to weakest, I would rank the objection to (1) as the strongest objection, followed by the objections to (3), (2), and (4), respectively. Suppose, however, someone presented a robust, even overwhelming, defense of libertarian freedom and hence of (1). Where would that leave this argument?

I strongly suspect that, once there is some reasonably well-accepted theory of intrinsic probability, it will support the claim that naturalism is antecedently much more probable than theism. Assume, for the sake of argument, that prediction is correct and premise (3) is false. While the argument from moral agency, as it stands, would be unsound, we would still be left with the argument’s premise (2), which, in my opinion, is powerful. How then could a philosopher build an explanatory argument for theism, based upon moral agency, that is also inductively correct? In my opinion, the most promising option would be to construct a cumulative case utilizing multiple, independent lines of evidence, since the cumulative effect of multiple, independent lines of evidences can be enough to overcome even an extremely low prior probability.


[1] Paul Draper, “Terrestrial Suffering and Cosmic Fine-Tuning: Parallel Problems for Theism and Naturalism.”  American Philosophical Quarterly 41 (2004): 311-321.

[2] 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.