Draft: William Lane Craig on the Evidential Argument from Evolution
This is a draft article I’ve been working on. Any feedback would be appreciated.
Abstract: Paul Draper defends what may be called an “evidential argument from evolution” against theism, viz., an argument which purports to show that evolution constitutes strong evidence against theism. In response to this argument, William Lane Craig argues that Draper’s argument depends upon three “dubious” probability estimates. I examine one by one Craig’s objections to these estimates and show how they miss the mark.
The idea that evolution is somehow a threat to “religion” is nothing new. Ever since the publication of Charles Darwin’s Origin of Species, there have been allegations of a “war” between “science” and “religion,” with evolution arguably constituting one of the war’s front lines. For most of this “war’s” history, the philosophical “fighting” has focused on questions of logical compatibility, such as whether evolution is compatible with Christian theism (specifically, with a literal interpretation of the Biblical book of Genesis) or, more broadly, whether it is even compatible with “mere” or “generic” theism; no one had bothered to make a serious effort to consider, apart from questions of logical compatibility, whether the truth of evolution might constitute evidence against theism even if it is consistent with it. This changed in 1997. Philosopher of religion Paul Draper, well-known for writing what is widely considered one of the best versions of the argument from evil (1996), developed what may be called the “evidential argument from evolution.” It takes the following form:
(1) Evolution is antecedently much more probable on the assumption that naturalism is true than on the assumption that theism is true.
(2) The statement that pain and pleasure systematically connected to reproductive success is antecedently much more probable on the assumption that evolutionary naturalism is true than on the assumption that evolutionary theism is true.
(3) Therefore, evolution conjoined with this statement about pain and pleasure is antecedently very much more probable on the assumption that naturalism is true than on the assumption that theism is true. (From 1 and 2)
(4) Naturalism is at least as plausible as theism.
(5) Therefore, other evidence held equal, naturalism is very much more probable than theism. (From 3 and 4)
(6) Naturalism entails that theism is false.
(7) Therefore, other evidence held equal, it is highly probable that theism is false. (From 5 and 6) (Draper 1997)
Strictly speaking, the argument is both an evidential argument from evolution and an evidential argument from evil: (1) appeals to the fact of evolution, whereas (2) appeals to facts about pain and pleasure (a type of so-called “natural evil”).
The argument includes several propositions in the relevant background knowledge:
B1: Pain and pleasure, if they exist, have intrinsic moral value.
B2: A physical universe—which operates according to natural laws, is intelligible, and which supports the possibility of intelligent life—exists.
B3: Living things, including sentient beings, exist on Earth. These sentient beings include, but are not limited to, human beings.
B4: Some (Earthly) sentient beings are not moral agents but are biologically very similar to (Earthly) embodied moral agents.
B5: Humans are goal-directed organic systems, composed of parts that systematically contribute to the biological goals of these systems.
So the argument can be restated as follows:
(1) Pr(E| N & B) >! Pr(E | T & B).
(2) Pr(P | E & N & B) >! Pr(P | E & T & B).
(3) Pr(E & P | N & B) >!! Pr(E & P | T & B). (From 1 and 2)
(4) Pr(|T|) =< Pr(|N|).
(5) Pr(N | E & P & B) >!! Pr(T | E & P & B). (From 3 and 4)
(6) Naturalism entails that theism is false.
(7) Therefore, Pr(T | E & P & B) <!! 1/2. (From 5 and 6)
In the twenty years since it was published, the evidential argument from evolution has attracted the attention of several philosophers, including William Lane Craig (2003, pp. 548-550), Alvin Plantinga (2011), and Daniel Howard-Snyder (2017). In this paper I want to critically assess Craig’s objections. Now if Craig claimed no more with respect to the evidential argument from evolution than the truism that “one man’s modus ponens is another man’s modus tollens,” namely, that one’s degree of belief in the falsity of an argument’s conclusion can exceed one’s degree of belief in an argument’s key premise (Craig 2003, p. 549), then the defender of the evidential argument from evolution would have no dispute with Craig, pointing out that the argument’s “other evidence held equal” clause preempts Craig’s punting to theistic arguments. Fortunately for discussion’s sake, Craig’s appraisal of the evidential argument from evolution is mostly independent of his appeal to theistic arguments. According to Craig, “Draper’s argument hinges on three probability estimates which seem dubious in light of our discussion” (Craig 2003, p. 549). In this response, I hope to show that the argument is, in fact, considerably stronger than Craig acknowledges.
Part 1: Craig’s Objections in His Written Work
First Objection: The Argument for Pr(|T|) =< Pr(|N|)
Craig’s first objection is that Draper (1997) assumes that theism and naturalism have equal prior probabilities. In Craig’s (2003, p. 549) words, Draper assumes that
naturalism and theism are equally probable with respect to our general background knowledge (Pr (N) = Pr (T)), which we have seen reason to dispute (recall chaps. 23-24).
As an objection to Draper 1997, however, this is simply misguided.
(i) First, Craig has confused prior probability with intrinsic probability. The former is a measure of the probability of a hypothesis conditional upon the relevant, extrinsic background information, whereas the latter is the probability of a hypothesis determined solely by intrinsic factors related to the content of a hypothesis, e.g., its scope and modesty. Allow me to introduce some mathematical symbols to make this clear:
Let Pr(|X|) =df. the intrinsic probability of X
Let Pr(X | B) =df. the prior probability of X conditional upon background information B
So Craig’s objection assumes that Draper’s argument either contains (or implies) a premise which says:
(4′) Pr(T| B) = Pr(N | B).
But this is false. The actual premise in Draper’s argument is:
(4) Naturalism is at least as plausible as theism, i.e., Pr(|T|) =< Pr(|N|).
Even if Craig were correct that theism had a higher prior probability than naturalism, this would be irrelevant to (4), which states that theism is not intrinsically more probable than naturalism. So far as I am aware, Craig has never interacted with any of Draper’s work on intrinsic probability. (write a lot more here)
(ii) Even if Draper’s argument had claimed that theism and naturalism contained equal prior probabilities, Craig’s selection of background propositions—i.e., the propositions which constitute the relevant background knowledge—is biased. Again, Craig (2003, p. 49) writes:
naturalism and theism are equally probable with respect to our general background knowledge (Pr (N) = Pr (T)), which we have seen reason to dispute (recall chaps. 23-24).]
What, precisely, were the reasons offered in chapters 23 and 24? The cosmological, teleological, axiological, and ontological arguments. Here I think Craig has not expressed himself very well. What could it mean to say that a set of arguments constitutes “our general background knowledge”? I am not even sure what that means. One option would be to include the conclusions of those arguments in our background knowledge:
B6. A maximally great being exists, i.e., a maximally great being exists in every possible world including the actual world. (Craig 2003, p. 496)
B7. The universe has a cause. (Craig 2003, p. 468)
B8. The explanation of the existence of the universe is God. (Craig 2003, p. 466)
B9. The fine-tuning of the universe is due to design. (Craig 2003, p. 484)
B10. God exists. (Craig 2003, p. 495)
That can’t be right because several of those conclusions (B6, B8, and B10) either explicitly state or imply that God exists. It’s illegitimate to include, in the background knowledge of a Bayesian argument, a proposition which entails the truth or falsity of the rival explanatory hypotheses under consideration. But three of these conclusions (B6, B8, and B10) either state or imply God’s existence, which renders them unsuitable for inclusion in the relevant background knowledge of an evidential argument about God’s existence. Furthermore, B6, if true, would entail that God’s existence is metaphysically necessary. It would be very odd, I think, to include “God’s existence is metaphysically necessary” in the background knowledge of any evidential argument against God’s existence. If God’s existence were metaphysically necessary, then we wouldn’t say that fact ought to be included in an evidential argument against God’s existence. Rather, we would say that all evidential arguments against God’s existence are fundamentally misguided, since there is no possible world in which God does not exist.
Another option would be to include in our background knowledge the key evidence to be explained in each of those arguments.
B6′. It is rational to believe that it is possible that a maximally great being exists. (From Plantinga’s ontological argument)
B7′. The universe began to exist. (From the kalam cosmological argument)
B8′. The universe has an explanation for its existence. (From the Leibnizian cosmological argument)
B9′. The universe is life-permitting. (From Craig’s teleological argument)
B10′. Objective moral values exist. (From the axiological argument)
This second option–focusing on the evidence to be explained–seems to be the most favorable to Craig’s goal of boosting the prior probability of theism over naturalism.
The second option fails, however, because it violates the inductive Rule of Total Evidence. Why does it violate the Rule of Total Evidence? Because it considers only some propositions (those which Craig believes to be favorable to theism) while ignoring other propositions (those favorable to naturalism). For example:
B11. It is rational to believe that it is impossible that a maximally great being exists.
B12. The physical exists. (From the evidential argument from physicality)
B13. It is rational to believe that it is impossible for a timeless being to create anything.
B14. So much of our universe is intelligible without appeal to supernatural agency. (From the evidential argument from the history of science)
B15. Conscious states in general are dependent upon the physical brain. (From the evidential argument from mind-brain dependence)
B16. The world contains an abundance of tragedy and relatively little triumph. (From the evidential argument from triumph and tragedy)
To sum up: Craig’s first objection mistakenly treats intrinsic probability as synonymous with prior probability. Furthermore, even if premise (4) had appealed to prior probability, Craig would still have failed to show that theism enjoys a higher prior probability than naturalism.
Second Objection: The Argument for Pr(P | E & N & B) >! Pr(P | E & T & B)
Craig’s second objection appeals to what I call the “skeptical theism defense” (Lowder 2016). Craig (2003, p. 549) writes:
Second, he believes that the probability of the distribution of pain/pleasure in the world is greater on naturalism and evolution than it is on theism and evolution (Pr (P/E&N) > Pr(P/E&T)). But we have seen reason to question whether we are in an epistemic position to make justifiably this sort of probability judgement.
What reason is that?
What makes the probability [that God has no morally sufficient reasons for permitting the evils that occur] here so difficult to assess is that we are not in a good epistemic position to make these kinds of probability judgments with any sort of confidence. As finite persons, we are limited in space and time, in intelligence and insight. But the transcendent and sovereign God sees the end of history from its beginning and providentially orders history so that his purposes are ultimately achieved through human free decisions. In order to achieve his ends God may well have to put up with certain evils along the way. Evils that appear pointless or unnecessary to us within our limited framework may be seen to have been justly permitted within God’s wider framework. (Craig 2003, p. 543)
By interacting solely with Draper 1997, it appears that Craig missed the fact that Draper 1996 (p. ##) already answered this objection. To sum up: it’s possible that God has unknown reasons for allowing evil. But it’s also possible—and antecedently just as likely—that God has unknown reasons for preventing evil. So the possibilities of unknown reasons for allowing evil and unknown reasons for preventing evil “cancel out.” We’re right back where we started, namely, working with what we do know: P. In fact, this is pretty much the point of using epistemic probabilities. If we had perfect, complete information, then we wouldn’t need to use probabilities at all. So human ignorance is not a good objection to comparing Pr(P | E & N & B) to Pr(P | E & T & B).
Furthermore, as numerous philosophers (nontheists and theists alike) have pointed out, logically consistent natural theologians cannot appeal to the limitations of human cognitive abilities to defeat evidential arguments from evil (Draper 1996b, p. 188). Allow me to explain. If human cognitive limitations really did prevent us from assessing whether God has morally sufficient reasons for allowing evil, including facts about pain and pleasure, then Craig can kiss goodbye all of his arguments from natural theology for God’s existence. Consistent skeptical theists should also insist that human cognitive limitations prevent us from assessing:
- the antecedent probability of our universe beginning to exist on theism, i.e., Pr(beginning | theism);
- the antecedent probability of so-called cosmological ‘fine-tuning’ on theism, i.e., Pr(‘tuning’ | theism); and
- the antecedent probability of the Resurrection on theism, i.e., Pr(Resurrection | theism).
This is why logically consistent natural theologians, like Oxford University philosopher Richard Swinburne, don’t rely upon skeptical theism. Instead, they attempt to provide theodicies—explanations for why God, if He exists, would allow facts about the kinds, amounts, and distribution of evil in the world to obtain (Draper 2010, p. 18).
Finally, Craig is completely silent on Draper’s supporting arguments for believing that Pr(P | E & N & B) !> Pr(P | E & T & B). As I read him, Draper gives three such arguments. First, our background knowledge includes the fact many other parts of organic systems are systematically connected to reproductive success. Second, Draper points out that evolutionary naturalistic Darwinism (E & N & D) provides an antecedent reason for believing that pain and pleasure, like anything else produced by natural selection, will be systematically connected to reproductive success, which is what P states. In fact, evolutionary naturalism (E&N) entails nothing that would provide an antecedent reason for doubting that pain and pleasure will resemble other parts of organic systems by being systematically connected to reproductive success. Third, given E&T, however, P would be true only if the biological goal of reproductive success and some unknown justifying moral goal happened to coincide in such a way that each could be simultaneously satisfied. That’s a really big coincidence that E & N & D doesn’t need.
Thus, on the assumption that E&N is true, it would be extremely surprising if pain and pleasure appeared to be anything but morally random, whereas on the assumption that theism is true, a discernible moral pattern would be less surprising. Draper concludes, accordingly, that (2) is true and Pr(P & E & N & B) >! Pr(P | E & T & B).
Third Objection: The Argument for Pr(E| N & B) >! Pr(E | T & B)
Craig’s third objection seeks to undercut (1) by appealing to the (alleged) improbability of life on naturalism. In his (2003, p. 549) words:
Finally, he argues that the probability of evolution on naturalism is greater than the probability of evolution on theism (Pr(E/N) > Pr(E/T)). For if naturalism is true, evolution is the only game in town; but if theism is true, God had more alternatives. But this assessment is confused. What Draper’s argument supports is the assessment that evolution is more probable relative to naturalism and the existence of biological organisms than to theism and the existence of biological organisms (Pr(E/N&B) > Pr(E/T&B)). But we have seen from our discussion of the teleological argument (chapter 23) that the existence of biological organisms (and, hence, their evolution) is virtually impossible relative to naturalism alone and that we should therefore expect a lifeless world given naturalism, which cannot be said of theism. Without his three crucial probability estimates Draper’s evidential argument from evil founders.
As an objection to (1), however, this objection is multiply flawed.
(i) I think Craig is being uncharitable to Draper. In Draper’s writings, he does not explicitly refer to background knowledge in his probabilistic notation; thus, “Pr(E / N)” can be charitably restated in its more explicit form as, “Pr(E / N & B),” where “B” represents the relevant background information. Indeed, this is precisely how I have presented Draper’s argument in this article. The key point here is that, in Draper’s original article, “probability of evolution on naturalism” means “probability of evolution on naturalism and our background information” and “probability of evolution on theism” means “probability of evolution on theism and our background information.”
(ii) What about the possibility of biological organisms on naturalism alone? Here Craig attempts to change the subject by appealing to the teleological argument. Let’s grant, but only for the sake of argument, that the probability of a life-permitting world on theism is greater than the probability of a life-permitting world on naturalism, i.e., Pr(life-permitting world | T) > Pr(life-permitting world | N). That fact, if it is a fact, is not of obvious relevance to the evidential argument from evolution. For the evidential argument from evolution compares the antecedent probability of evolution on naturalism and on theism, i.e., Pr(E | N & B) > Pr(E| T & B). Craig seems to think that if he can show that if a life-permitting world is extremely improbable on naturalism, it somehow follows that (1) is false. In other words, Craig seems to move from:
The probability of a life-permitting world on naturalism is extremely low, i.e., Pr(life-permitting world | N) << 0.5.
It is false that the probability of evolution on naturalism (and background information) is greater than the probability of evolution on theism (and background information), i.e., it is false that Pr(E | N & B) > Pr(E| T & B).
The problem, however, is that this does not follow. For the sake of argument, it may be the case that the fact that our universe is life-permitting is more probable on theism than on naturalism, but, given that our universe is life-permitting, the fact that all living things are the gradually modified descendants of earlier living things is evidence favoring naturalism over theism. Indeed, this is precisely Draper’s (2001) position!
Craig, William Lane (2003). “The External Problem of Evil,” in J.P. Moreland and William Lane Craig, Philosophical Foundations for a Christian Worldview (pp. 548-550). Downers Grove: InterVarsity.
Draper, Paul (1996). “Pain and Pleasure: An Evidential Problem for Theists.” Noûs, 23 (3): 331-350. Reprinted in Daniel Howard-Snyder (Ed.), The Evidential Argument from Evil (pp. 12-29). Indianapolis, IA: Indiana University Press.
Draper, Paul (1997) “Evolution and the Problem of Evil” in Louis Pojman (Ed.), Philosophy of Religion: An Anthology (pp. 219-230). 3rd ed., Belmont: Wadsworth.
Draper, Paul (2001). “Seeking But Not Believing: Confessions of a Practicing Agnostic” in Daniel Howard-Snyder and Paul Moser (Eds.), Divine Hiddenness: New Essays (pp. 197-214). New York: Oxford University Press.
Draper, Paul (2010). “God and Evil: A Philosophical Inquiry” (October 1, 2010). Talk presented at the University of Notre Dame Ninth Annual Plantinga Lecture, Notre Dame, Indiana. <https://philreligion.nd.edu/assets/44795/1011lecture.pdf>
Howard-Snyder, Daniel (2017). “The Evolutionary Argument for Atheism” in John-Christopher Keller (Ed.), Being, Freedom, and Method: Themes from Van Inwagen (pp. 241-62). New York: Oxford University Press.
Lowder, Jeffery Jay (1998). “Summary and Assessment of the Craig-Draper Debate on the Existence of God (1998).” The Secular Outpost blog. <https://secularfrontier.infidels.org/2016/02/20/summary-and-assessment-of-the-craig-draper-debate-on-the-existence-of-god-1998/>, site accessed December 20, 2016.
Lowder, Jeffery Jay (2016). “In Defense of an Evidential Argument from Evil: A Reply to William Lane Craig.” The Secular Web. <https://infidels.org/library/modern/jeff_lowder/pain-and-pleasure.html>, site accessed September 1, 2019.
Oppy, Graham (2016). TBD
Plantinga, Alvin (2011). Where The Conflict Really Lies: Science, Religion, & Naturalism. New York: Oxford University Press.
 I think the truth of B6′ is far from obvious. Indeed, as Graham Oppy (2016, p. TBD) points out, “opponents of the argument are bound to challenge the acceptability” of B6. He continues, “And, of course, they do. Let’s just run the argument in reverse.” Oppy then runs the argument as follows:
There is no entity which possesses maximal greatness.
(Hence) There is no possible world in which there is an entity which possesses maximal greatness.
Oppy concludes: “Plainly enough, if you do not already accept the claim that there is an entity which possesses maximal greatness, then you won’t agree that the first of these arguments is more acceptable than the second. So, as a proof of the existence of a being which possesses maximal greatness, Plantinga’s argument seems to be a non-starter.”
 Even if we assume, but only for the sake of argument, that B7′ is more probable on theism than on naturalism, this argument commits the fallacy of understated evidence. Given that the universe began to exist, the fact that it began to exist with time, not in time, is more probable on naturalism than on theism.
 Even if we grant that the universe has an explanation of its existence, it doesn’t follow that the explanation is God. Other possible explanations include: (i) an infinite regress of contingent universes; and (ii) our universe’s factual necessity. If our universe is factually necessary, then its existence would be partially explained by its own nature (which is uncaused, beginningless, and independent / free-standing) and partially explained by virtue of other things that happen to exist (i.e., nothing around it has what it takes to knock the universe out of existence). I owe this objection to Felipe Leon.
 Even if we assume, but only for the sake of argument, that B9′ is more probable on theism than on naturalism, this argument commits the fallacy of understated evidence. Given that the universe is life-permitting, the fact that so much of it is hostile to life is more probable on naturalism than on theism. Furthermore, given that the universe is life-permitting, the fact that life is the result of evolution is much more probable on naturalism than on theism.
 B10′ is not more probable on theism than on naturalism. Theism assumes, not explains, the existence of objective moral value.