Divine Creation and Evolution

Alexander R. Pruss

July 9, 2005

ap85@georgetown.edu

Department of Philosophy

Georgetown University

Washington, DC 20057

Divine Creation and Evolution

July 9, 2005

Abstract

It is widely accepted that divine creation of human beings, except maybe for the souls thereof, is compatible with neo-Darwinian evolution. I shall argue that, on the contrary, neo-Darwinian evolution is at least rationally and perhaps logically incompatible with a plausible understanding of what it means to say that God intentionally created human bodies. I will do this by examining three different accounts that try to make compatible the idea of creation with a neo-Darwinian account: determinstic evolution with God setting initial conditions, stochastic evolution with God having middle knowledge and stochastic evolution without God having middle knowledge.

Divine Creation and Evolution

Alexander R Pruss

July 9, 2005

1. Introduction

It is widely accepted that the idea of divine creation of human beings is compatible with evolutionary theory, except perhaps in regard of the human soul, and that evolutionary theory provides an explanation of speciation and of complex features of organisms that undercuts Paley-style teleological arguments, whether or not the evolutionary mechanisms are truly random or deterministic. I will argue that this “orthodoxy” is wrong on a plausible reading of the doctrine that God intentionally created human beings.

On the contrary the doctrine of the creation of human beings, read as I will recommend, is logically incompatible with deterministic evolutionary theory. How compatible it is with indeterministic evolutionary theory will depend on whether God in creating makes use of “generalized middle knowledge” (GMK) which would allow God to know what outcome a stochastic process would have were it initiated. If God is understood to make use of GMK, then creation will be logically incompatible even with indeterministic evolutionary theory. If God does not make use of GMK (e.g., because he lacks GMK), then there will be logical but not rational compatibility: one ought not believe that God does not make use of GMK, that God created human beings (in the sense I will define) and that indeterministic evolutionary theory is true.

What is particularly surprising about the argument is that I am willing to grant that all the first-order claims of evolutionary theory are rationally compatible with western monotheism, with the possible exception of the issue of the special creation of the human soul. Thus, I accept as compatible with western monotheism the claim that human bodies, as well as all non-human organisms, are genetic descendants of a unicellular common ancestor of all organisms on earth (“common descent”) and the claim that the differences between successive generations (apart perhaps from the possible exception of the production of the human soul) are caused by naturalistically explainable genetic mutations and recombinations (“no miracles in genetic history”).

2. The creation doctrine

I will need to elucidate the creation doctrine. I take the claim that God created human beings to imply that God intentionally and specifically brought it about, immediately or mediately, that human beings exist in such a way that the overall design of the human species, including human bodies, can be attributed to that intention, in the way that the design of an artifact can be attributed to the craftsman, to borrow the analogy in Isaiah 64:8 and Romans 9:21. This implies that the existence of a species having the mental and physical features of the human species is explained by God’s intentional activity. This is quite compatible with a naturalistic story about human beings came into existence (leaving aside considerations of the soul—a qualifier that in the interests of brevity I will now stop making). God might have, after all, carefully arranged the initial conditions in such a way that eventually human beings would be determined to arise, and carefully picked these initial conditions so that the they would have precisely the features God wanted them to have. The notion of creation thus does not beg the question against theistic evolutionary account, since it is compatible with God using any chain of intermediate causes he might wish to in order to fulfill his plan.

But this doctrine of creation is meant to rule out some alternative stories that are prima facie compatible with God’s being the first cause of the universe. On one story, God created an initial mix of stuff without any intention that the human species should arise from it. He then observed what arose, deterministically or not, and when human beings evolved, he, as it were, applauded. On this view God would indeed the first cause of human beings, in that he would have initiated the chain of causes leading to the existence of human beings, but he would not have been the intentional cause of the existence of human beings. The fact that he might foreknow that human beings would arise from the initial conditions does not change this, since, even without committing oneself to the full Principle of Double Effect, it is clear that one can intend something to happen without caring at all about other effects further on down the line. Thus, one might know as one sets out to write something that one’s hand motions will cause agitation in the air, without in any way caring about this agitation or intending it.

Nor is the doctrine of creation as I understand it compatible with the idea that God produced an evolutionary process merely intending that it would lead to some interesting result or other, and then when he saw that the interesting effect would in fact be the arising of something like human beings, he put some minor finishing touches on their bodies, say miraculously intervening to make them hairless. On this story, God did initially intend human beings to arise, since he intended that something or other interesting should happen and human beings in fact were that interesting thing, but he did not specifically intend that human beings should arise. The finishing touches would be a sign of a subsequent specific intentional activity, but this intentional activity would not count as responsible for the overall design of the human body.

I will call the doctrine I am elucidating “the creation doctrine” for brevity. Strong predestinationist views will imply the creation doctrine. If God intended that specifically Paul should be saved, then God intended that specifically Paul should exist. And since Paul is essentially human, God intended specifically that the human race should exist. Moreover, arguably, if God were not responsible for the overall design of the human body, then God’s intention to predestine, and hence create, Paul would not be prior in the order of explanation to all events in the created universe, in the way predestination requires.

But one does not have to accept strong views of predestination to find the creation doctrine plausible. The doctrine seems necessary in order to do justice to the notion that God is our author, that his relation to us is like that of a craftsman or artist to what is produced.

There is still an ambiguity in the creation doctrine’s use of the term “human”. On a strong reading, “human” means precisely that: a biological human being, at least a member of the genus homo. On a weaker reading, “human” means any animal capable of love, intelligence and sophisticated rationally-guided engagement with the physical world.

I will thus talk of a stronger and a weaker creation doctrine, and there are differences in the conclusions of this paper for the two. Both doctrines are logically incompatible, I shall show, with evolutionary theory on either a deterministic reading or a GMK-based indeterministic reading. Both doctrines are, as far as this argument goes, logically compatible with an indeterministic no GMK version of evolutionary theory. However, the stronger doctrine is not rationally compatible with this theistic evolutionary variant. Currently, the weaker version is rationally compatible with it, but this may change with additional empirical data. The conclusions of this paper are strengthened by the fact that I will also give an independent argument against the idea of genuinely random stochastic processes in a theistic cosmos without GMK. If this argument is sound, then we are left only with the two versions of theistic evolutionary theory that are logically incompatible with even the weaker creation doctrine.[1]

3. What in evolution is fully compatible with the creation doctrine?

I will use the term “the full ancestral history of x” to denote a proposition reporting the complete first-order causal history of the descent of an individual organism x going all the way back to a very simple (admittedly a vague term that obscures the fact that the simplest organism may still be a complex being) organism, or group of very simple organisms, that does not itself have an ancestor. This history includes such conjuncts as that ancestor x₁₁₂ resulted from the sexual union of ancestors x₁₁₃ and x₁₁₄ whose DNA was recombined and/or mutated in such-and-such a way under the influence of such-and-such causal factors, as well as conjuncts reporting the first-order causal facts causally responsible for various ancestors surviving to reproduce, e.g., that ancestor x_29949431 killed a saber-tooth tiger with her tusks at t₄₄₀. I shall further assume that if humans are composed of body and soul, the ancestral history only concerns the body. If an organism is not ultimately descended from very simple organisms, then I shall say it lacks a “full ancestral history”.

What I claim to be logically compatible with the creation doctrine (in either the strong or the weak form) is the claim that all organisms now alive have full ancestral histories and the events in these histories are naturalistically caused and such as one would expect given evolutionary theory.

The first-order claims of naturalistic evolutionary theory are causal stories about all the currently extant organisms entailing that these organisms have full ancestral histories that consist of reports of events that are naturalistically caused and such as one would expect given evolutionary theory. These first-order claims are compatible with the strong creation doctrine, I take it. The argument for this compatibility is quite simple: God being all-powerful and all-knowing could surely have created a very simple common ancestor and arranged the initial conditions in such a way that the events that would eventuate would be such as a naturalistic evolutionist would accept, and there is nothing in the notion of a creator God to logically rule out that this in fact happened.

A powerful objection to this claim is that it would be deceitful for God to act in such a way, since it would mislead people into naturalism. But this objection would fail if God in fact took safeguards against such misleading (e.g., by speaking in the hearts of the people who might be deceived, or by producing other evidence for design), and it is rationally compatible with both western monotheism and the first-order claims of naturalistic evolutionary theory that he took such safeguards. Moreover, there might have been other benefits in God’s acting this way, e.g., because it might allow for more uniformity in the laws of nature.

4. Where the incompatibilities lie

4.1. The deterministic and generalized middle knowledge cases

4.1.1. The argument

What, then, is problematic in evolutionary theory vis-à-vis the creation doctrine? It is the ambitious claim that evolution provides a true explanation of why such marvelously complex and adapted animals as horses, pine trees and frogs exist, with complex organs such as equine eyes and human brains, and why intelligent animals like humans exist, an explanation whose possibility competes with, and undercuts, Paley-type teleological arguments. If evolutionary theory undercuts Paley-type teleological arguments, then plainly it does so not by showing that Paley was mistaken as to the empirical fact that there are intelligent animals or that organisms have much complexity, but rather by giving a naturalistic explanation of the same facts that Paley and his fellow teleological arguers sought an explanation for.

That evolutionary theory in fact claims to provide such an explanation is, I think, clear from the history of the debate on the evolutionary theory. Of course, the complexity and well-adaptedness of organisms and the existence of intelligent life is not all evolutionary theory purports to explain. It also purports to explain, say, facts about the geographic distribution of species, the diversity of the species, and so on. But it is the ambitious claim to explain such things as intelligence, the mammalian eye or the elephant’s trunk that make evolutionary theory particularity interesting to our culture. And it is this ambitious claim that is incompatible with even the weak creation doctrine, given determinism. As we shall see, however, it is quite possible that evolutionary theory could be weakened to make correct and less ambitious explanatory claims while at the same time not changing its first-order claims. It could well be that the resulting theory would (a) be compatible with the strong creation doctrine and (b) still be sufficiently powerful explanatorily in this less ambitious realm so as to justify belief in itself through an inference to best explanation. But such a weakened evolutionary theory would not compete with theistic teleological explanations.

However, even though a monotheist accepting the creation doctrine cannot consistently hold that a deterministic evolutionary theory making the ambitious explanatory claims is true, she can still reasonably hold that the epistemic availability of such a theory undercuts Paley-type arguments. I will not argue that there is empirical evidence of divine design. However, the weak creation doctrine logically entails the presence of a kind of design. This yields non-scientific evidence for Intelligent Design if (as I believe) there is non-scientific apologetic evidence for a monotheism accepting the creation doctrine. I also hope that once a philosophically-inclined theist sees that she must, if she is to remain consistent to her religious beliefs, abandon the truth of the ambitious explanatory claims, she will be motivated to explore whether there might not be empirically-grounded arguments against the truth of the ambitious explanatory claims, arguments perhaps like those that the proponents of Intelligent Design offer, though it is also epistemically possible at this point that no such arguments will succeed.

Enough has been said about the incompatibility claim and its implications. It is time to show that the claim is true. Now, evolution can only purport to fulfill what I have called the “ambitious” explanatory goal through statistical explanations. It is clear that there is no hope arguing that the mammalian eye had to evolve. No doubt there are multiple possible solutions to any one evolutionary problem. Not even the weaker claim that the mammalian eye or something equally adaptive to the circumstances had to evolve is true. After all, evolution is grounded in the existence of random or quasi-random mutation and recombination processes as well as the fact that the descendants with genotypes that produce more adaptive phenotypes are more likely to survive. There are no guarantees here, even in the deterministic case. For in the deterministic case, while it is possible in principle to explain physically why such-and-such a feature now exists by showing how the feature had to arise from the state of the universe in the past, such an explanation would not be a neo-Darwinian one. It is essential to an ambitious neo-Darwinian evolutionary explanation of a complex feature of an organism that the explanation be statistical in nature. The explanation will involve the claim that it is not improbable that the mammalian eye or something about as adaptive would evolve over the millions of years available for nature to experiment. In other words, the ambitious explanatory goal is supposed to be satisfied through statistical explanation.

Less ambitious evolutionary explanations can afford to be simply causal in nature, simply listing the causal factors in the ancestral history of an organism. But such less ambitious explanations are not sufficient for undercutting teleological arguments in the way that ambitious evolutionary theory claims to do through its explanatory account. To see this, suppose that the naturalistic history of the development of the human body was that first dogs developed, and then a bunch of cosmic rays struck a dog and caused its DNA to mutate into the DNA of human beings. While this would yield a causal explanation of the existence of human beings, still, given the astronomical unlikelihood of the event in question, the causal story here would do nothing to undercut a teleological argument. Given the data that the above sequence of events in fact happened, it would be quite reasonable to suppose that an intelligent agent aimed the cosmic rays in such a way as to produce the DNA of human beings, just as it would be reasonable to suppose an intelligent aiming agent to be involved should cosmic rays striking an empty memory chip produce all of Shakespeare’s Hamlet.[2]

That is why evolutionary theorists emphasize the great amount of time that was available for evolution and the large number of organisms in which random mutation occurs. For instance, Darwin writes about the evolution of the eye that

we must suppose that there is a power, represented by natural selection or the survival of the fittest, always intently watching each slight alteration in the transparent layers; and carefully preserving each which, under varied circumstances, in any way or in any degree, tends to produce a distincter image. We must suppose each new state of the instrument to be multiplied by the million; each to be preserved until a better one is produced, and then the old ones to be all destroyed. In living bodies, variation will cause the slight alterations, generation will multiply them almost infinitely, and natural selection will pick out with unerring skill each improvement. Let this process go on for millions of years; and during each year on millions of individuals of many kinds; and may we not believe that a living optical instrument might thus be formed as superior to one of glass, as the works of the Creator are to those of man?[3]

It is precisely the genius of Darwin to have realized that random variation over millions of years in millions of organisms is not unlikely to produce very new phenotypes and that the better of these will be selected for. Without this amount of time and without the number of organisms involved, while the first-order causal stories plainly could still have happened, the probabilities of the emergence of something as complex like the eye would have been astronomically small, and hence the ambitious explanatory goals would not have been met.

Or take Dawkins’ claim:

We have seen that living things are too improbable and too beautifully ‘designed’ to have come into existence by chance. How, then, did they come into existence? The answer, Darwin’s answer, is by gradual, step-by-step transformations from simple beginnings, from primordial entities sufficiently simple to have come into existence by chance. Each successive change in the gradual evolutionary process was simple enough, relative to its predecessor, to have arisen by chance.[4]

The probabilistic explainability of the successive steps, if not of the sequence as a whole, seems to be crucial here.

No claim is made here that all neo-Darwinian explanation proceeds through natural selection. But a claim is made that ambitious neo-Darwinian explanations involve natural selection and genetic variation to some degree, and involve them in a statistical way.

The details of how the statistical explanation works in the Darwinian case are of course controversial. Perhaps it is the case that there is a high probability that complex solutions to problems raised by the environment would evolve. In this case, the explanation would be like the statistical explanation of the fact that a thousand coins thrown from a high building landed about half heads-up and half tails-up: it is highly probable that such an arrangement would happen. Or perhaps the probability is relatively low, but at least not astronomically low, and so the explanation of why intelligence evolved is like saying that Fred got paresis as a result of having had syphilis, where paresis is a not very probable but still not very improbable result of syphilis. It does not matter, however, which story we take. It will still be logically incompatible even with the weak creation doctrine, if we are either working with a deterministic evolutionary theory or if we assume God has GMK.

One might wonder how a deterministic version of evolutionary theory could involve statistical explanations. There is a real difficulty here, akin to the question of how 19th century deterministic thermodynamics could yield statistical explanations. I suspect that any successful answer would involve in both cases something like an assertion that we can do statistical explanation provided we impose some canonical probability distribution (e.g., a uniform one) on the state space of initial conditions. The details are difficult, but since scientists did in fact propose statistical explanations on the basis of deterministic thermodynamics, the analogous procedure in the case of evolutionary theory seems at least epistemically possible.[5]

So far we have made various simple observations about evolutionary theory. Consider now the weak creation doctrine that God intentionally and specifically brought it about that intelligent life exists, in such a way as to be the author of much of the design of the life forms. Now, plainly, God could have designed his creatures through causal processes reported by full ancestral histories that are in fact compatible with evolutionary theory. Nor is the problem just that the creation doctrine purports to provide one explanation of an event (the arising of intelligent life) while the evolutionary theory purports to provide another. For a given event can have more than one explanation: “Fred died because his heart stopped beating. It is also true that Fred died because he was shot to death.” Rather, the problem is that the design claims, if true, would undercut the statistical explanations that evolutionary theory makes use of to make good on its most ambitious claims.

To see this, suppose that Fred has dropped a thousand coins one by one from a high building, and 485 of them landed heads-up. Bob now wonders: “Why was the percentage that landed heads-up between 45% and 55%?” There is a simple statistical explanation here, it seems:

(1) The probability that between 450 and 550 of the coins would land heads-up is about 99.9%, and this is why somewhere between 45% and 55% of the coins landed heads-up.

But suppose now we learn a further fact. Fred in fact calculated at what angle and velocity he would have to throw each coin in order that the coin should land heads-up and the angle and velocity needed to get tails. Then he carefully chose the angles and velocities so as to guarantee that between 45% and 55% of the coins would land heads-up. I submit that this extra fact falsifies (1). It falsifies (1) not by falsifying the probabilistic claim made. That claim is true, providing that we are careful to specify that we are looking at probabilities given no information on the angles and velocities the coins are tossed at. What is seen to be false now is the claim that this probabilistic fact explains the result. For it no longer does. The reason is that the probability stated in (1) is no longer conditioned on the relevant background condition, since now the relevant background condition is that Fred tossed the coins in such a way as to ensure that between 45% and 55% of them would land heads-up. The conditional probability conditioned on this piece of background knowledge is 100%, not 99.9%, and it is this conditional probability that is explanatory of the result. The same is true if one is dealing with statistical explanations involving low probability.

For an even clearer example suppose that Fred is extremely good at tossing coins in such a way as to get them to land as he wishes. He tosses the coins in such a way as to make them land all heads. However, his method for tossing coins has a small probability of failure, say 1%, and as a matter of fact not all coins land heads up. Let K₁ be the claim that 1000 coins were tossed. Let K₂ be the claim that 1000 coins were tossed by Fred who is extremely good at tossing coins to get the result he wants, with a failure probability of 1%, and who intended to have the coins land all heads-up. Let E be the event that not all the coins land heads-up. Consider the following two claims:

(2) P(E|K₁) is extremely high (P(~E|K₁) is astronomically low).

(3) P(E|K₂) is low but not astronomically so.

Both claims are true. But it would be simply false to say that (2) explains the occurrence of E. For given that the coins were tossed by a skilled coin-tossing agent with a specific intention, proposition K₁ is no longer the relevant background knowledge, even though it is still a true proposition. For it is clear that in this case the occurrence of E is explained by Fred’s failure, not by the high likelihood that a bunch of coins tossed will contain some coin that lands heads-up. The latter is irrelevant in cases where the initial conditions have been “cooked” by an intelligent agent with a specific intention.

It is thus clear that

The fact that P(E|K)=p explains the occurrence of E

can only hold if K is appropriately chosen, no matter how high p is. Alas, philosophers of science do not have a full story about what background information K counts as appropriately chosen, but we can often decide this on a case by case basis. For instance, apart from cases of self-explanatory propositions, the claim that P(E|E)=1 does not explain the occurrence of E, even though we have here the highest probability we can. Or consider the following case. Let E be the event that a die thrown landed 1, 2, 3, 4 or 5. Given just the information K₁ that we have a six-sided die that was tossed, by the principle of indifference we will say that P(E|K₁)=5/6, and might use this to explain why E in fact occurred. But if we have the additional information that the die is loaded in favor of sixes in such a way as to produce a six 90% of the time, then our previous attempt at explanation is undercut, at least in most ordinary contexts, even though it is still true that P(E|K₁)=5/6. For plainly we should make use of the more relevant information K₂ that we are dealing with a six-sided die that produces sixes 90% of the time.

In fact, in general, information about the intentional activity of agents working to tweak the initial conditions so as to produce or prevent the effect E is relevant background information. The failure to include this information in the background information relative to which conditional probabilities are computed vitiates the claim that these probabilities are explanatory of E. In particular, if the initial conditions are deliberately chosen by an intelligent agent in such a way as to guarantee the occurrence of E, then any attempt at probabilistic explanation that does not in some way involve the intelligent agent is a failure: the claim that an explanation is provided is simply false. Note that this illustrates the general claim that it is not always the case that

(p explains q) É ð ((p & q) É (p explains q)).[6]

That my analysis of relevancy applies in the case of pseudo-random but actually deterministic phenomena should be clear. If Fred set the initial conditions for the coin toss in such a way as to guarantee that 45%-55% of the coins would land heads-up, then it is false to say that there is a probabilistic explanation of them thus landing which explanation does not involve Fred.[7] This shows that if it is the case that God definitively intended to produce some animal or feature of an animal, then a naturalistic statistical evolutionary explanation of the existence of the animal or feature is incorrect, if the processes behind the evolutionary processes are really deterministic. For the probability that such-and-such a feature would arise relative to background information that does not specify the possibility of the initial conditions being arranged by God is explanatorily irrelevant in this case.

Thus, the presence of the ambitious evolutionary explanatory claims is in fact incompatible with the weak creation doctrine, provided the evolutionary processes are ultimately deterministic. The same negative result holds providing God is able to make use of something analogous to middle knowledge for indeterministic processes. Recall that according to the controversial position of the Molinist, God has middle knowledge of what a person would do were she created and placed in circumstances C that specify in as great a detail as possible everything that happens before the person’s decision. Similarly, one might think that God has generalized middle knowledge (GMK) of how a random process would turn out were it allowed to run. Thus, given an indeterministic coin flipping experiment described exhaustively under description D, God knows whether the coin would land heads up or whether it would land tails up were an experiment satisfying D actually executed.[8]

Now, if God has generalized middle knowledge and can make use of it while creating (a separate assumption!), then God knows what any given set of initial conditions would result in. Presumably, then, he would be choosing the initial state in such a way as to guarantee the existence of the animals or features that he intends to produce. But then this intentional act of God’s is again a piece of background knowledge that would undercut the evolutionary explanation.

Our intuitions are weaker here because we are not used to the idea of middle knowledge or generalized middle knowledge in our daily lives. But suppose now that Fred has generalized middle knowledge of the outcomes of indeterministic coin tosses. An indeterministic coin toss presumably involves the coin’s path being subject to indeterministic processes, though also affected by the initial conditions. Fred now knows that there is a set of initial conditions that would result in 45%-55% of the coins landing heads-up, and he knows which sets of initial conditions are such. He chooses one of them. I think it is still true that the statistical explanation of the result is incorrect. It is incorrect because the information about Fred’s knowledge and intentions is clearly relevant, and the purported statistical explanation is then based on background information that fails the test of saliency.

Consider a simpler case. Suppose that Fred goes to a casino and knows, e.g., because God told him, which ones of the slot machines would hand him a win on the next game. He chooses one of these and plays. Let us suppose there is a 30% chance of winning. It is surely incorrect to say that there is a statistical explanation of his winning in terms of the probability being low but not astronomically low. Rather, Fred wins because he picks a slot machine that he knows would give him victory. Nothing in what I have said in this paragraph is affected should the chance of winning be very high or very low. Fred’s knowledge of how to rig the initial conditions and his intention to do so must in some way enter into the background conditions in any correct statistical explanation.

Or consider a different case. A casino owner that has GMK and is omnipotent in regards to slot machine construction is about to construct a number of indeterministic slot machines from scratch that have a statistical propensity of generating a 5% payoff in the casino's favor. She then employs her generalized middle knowledge and chooses to build those slot machines which are such that she would in fact be guaranteed approximately a 5% payoff in the first fifty years of operation. I claim that the statistical explanation of the 5% payoff in terms of the great likelihood of this outcome is then undercut by the agency claims. One way to see that it is undercut is to note that even if the designs were ones that had a propensity to generate, say, a 12% loss, using the procedure of choosing to build those that would in fact guarantee the desired result would still ensure a 5% gain. The statistical propensity does not yield a statistical explanation of the outcome, though it may enter into an explanation in a different way[9], e.g., by explaining why the casino owner chose these and not other machines.

Note, however, that while the statistical propensity would not yield a statistical explanation, we might get a causal explanation of the outcome, since the processes inside the slot machines are presumably causes of the outcome. Likewise, then, the first-order naturalistic ancestral histories would give an explanation of the existence of the species. But it would not be an evolutionary explanation, since ambitious evolutionary explanations are essentially statistical in nature as we have already noted.

One might be worried that what I have said proves too much, e.g., that my argument shows that God’s making use of middle knowledge about libertarian-free actions in deciding whom to create would undercut the freedom of these actions. But that worry would be unjustified, at least in connection with, say, Chisholm’s classic agent-causation account. For just as generalized middle knowledge does not undercut naturalistic causal explanations but only statistical ones, so too the present argument would only present a danger for freedom given middle knowledge if the libertarian explanation of the free actions were a statistical one. Admittedly, some libertarians have attempted to give statistical explanations of why one action rather than another was chosen, but one need not take this view, or at least one need not take it that the availability of statistical explanations is essential to human freedom.

Note that the above arguments about evolutionary theory used only a weak version of the creation doctrine. What the argument needs is that God intentionally set things up so that Ks should exist, where K is a kind (natural or not) of relatively complex organism (say, intelligent animal or sexually reproducing organism) that evolutionary theory claims to give a statistical explanation of.

It is of course essential to my argument that one of the claims of evolutionary theory is to explain the instantiation of this kind. Some evolutionists may shy away from the claim that the existence of humans is explained in an evolutionary way, and insist instead that what is explained is that some complex and well-adapted species or other is exemplified. But insofar as that explanation is also statistical, and since it follows from the weak creation doctrine that God intended something (the existence of intelligent embodied animals) that entails the existence of what the evolutionary theory purports to explain statistically (the presence of a complex and well-adapted species), the problem persists.

There is also a variant of the above argument that was developed in a discussion Austin Dacey and I had, which variant I will briefly sketch as a more careful development would require undue space here. It is a basic assumption of neo-Darwinian theory that the probabilities of mutations leading to a particular genotype are independent of the adaptiveness of the phenotype that the genotype codes for. Adaptive mutations are no more likely simply in virtue of being adaptive than maladaptive ones are. However, this statistical independence is destroyed once one supposes that the initial conditions were chosen to produce intelligent life. For without any divine design, the probability that a series of mutations leading to a genotype coding for an intelligent life-form would occur is less than 1. But given that the initial conditions were rigged, as per the weak creation doctrine, so as to ensure the evolution of intelligent animals, the probability that a series of mutations leading to a genotype coding for an intelligent life-form is precisely 1. Thus, God’s creative design removes the independence between mutation probabilities and adaptiveness, thereby undercutting one of the pillars of neo-Darwinian explanation.

4.1.2. Objections

I. The argument proves much too much.[10] This argument seems to show that statistical explanations in general are incompatible with Western theism. After all, nothing in the argument depended on the specific way that statistical explanation works in the evolutionary case. Hence, if this account is correct, it seems that we cannot explain why the cream spreads throughout the coffee cup by alluding to standard random molecular motion considerations.

However, there is something important about the case of the arising of the human species that the argument depended on. The creation doctrine said that God brought it about that human beings exist and had a specific intention to do so. Moreover, on the deterministic and GMK readings, we assumed that God set up the initial conditions in order to guarantee this goal. If God has a specific intention to make the cream spread throughout the coffee and sets up the initial conditions in order to guarantee this goal, then it is false to say that the statistical facts explain the spread of the cream. If God did not set up the initial conditions in order to guarantee the diffusion of the cream, then it may well be quite appropriate to explain the diffusion in statistical ways. Remember that crucial to the statistical relevance argument was the claim that when the initial conditions are intentionally set up specifically in order to produce or preclude the explanandum, then this intentional set up is a part of the relevant background conditions. Without such an intentional arrangement, no such conclusion need follow.

And it is highly plausible that God does not always specifically intend everything that happens under every description true of it. Agents intend things under descriptions. As the Principle of Double Effect illustrates, one can know that something will result from one’s action without intending that outcome, and one can intend something under one description without intending it under another that one knows applies. Thus, a polio vaccine manufacturer may have intended the production of the vaccine under the description “production of a vaccine that overall would save vast numbers of lives and would be profitable”, but did not intend the manufacture under the description “production of a vaccine that would kill a small number of patients”, even though it is reasonable to expect some people to die due to side-effects of the vaccine.

In fact, it seems quite necessary to say that God does not intend everything that happens under every description that applies to it. For instance, if an evil happens to an innocent person, we probably do not want to suppose that God intended it under the description “an evil happening to an innocent person”. Rather, God presumably intended it under another description. What the details of that description would be will depend on what theodicy is in fact right. Perhaps the description will be “the event naturally necessitated by initial conditions I chosen so as to produce a good effect E” or “an event that would give Smith an opportunity to grow even more in courage”.

II. Evolutionary explanations can be non-statistical.[11] Consider the following naïve arm-chair biology account of why northern Europeans have long and narrow rather than broad and short noses. Many thousands of years ago, a comic ray hit a DNA molecule and this resulted in Helga’s having genes that code for a long and narrow rather than broad and short nose like her parents did. This was passed on to Helga’s descendants. The long and narrow noses in Helga’s descendants warmed the cold air and enabled them to stay out longer hunting and fishing in the long northern winter. And this made them be better nourished, and hence better able to survive, and their improved hunting abilities also made them be more desirable mates. Thus, Helga’s descendants did better than their broad short-nosed cousins. After a number of generations, people began to find long, narrow noses attractive, and Helga’s descendants dominated the gene pool. This story is not statistical. It is simply a description of a sequence of events together with causal explanations in terms of the phenotype coded for a novel genotype found in Helga.

As it stands, however, this story is not an ambitious one in the sense earlier indicated. Ambitious evolutionary explanations undercut Paley-style teleological arguments. The story as given above does nothing to undercut such an argument. To see this, suppose that it turned out that any long-narrow-nose-coding (LNNC) genes would have to include hundreds of precisely specified base-pairs differences from short-broad-nose-coding (SBNC) genes. If that were so, then the likelihood that a cosmic ray could change DNA including SBNC genes into DNA including LNNC genes would be so astronomically small that an inference to a designer would seem quite plausible. The claim that a cosmic ray was responsible would not by itself weaken a design argument based on the fittingness of long narrow noses in northern climes, if the likelihood of a cosmic ray hitting the DNA molecule with that precise angle and energy level it hit it with were all but infinitesimal. Rather, it would shift the design question to a query of why the cosmic ray thus hit the DNA molecule.

Ambitious neo-Darwinian evolution is supposed to make seemingly unlikely changes more likely, given appropriate redescriptions (e.g., redescribing “mammalian eye” into “high resolution sensory device”) and given sufficient time and a sufficient population pool that would make the changes not astronomically unlikely. Thus, in order to make the story given above be an ambitious one, one needs further information. One piece of further information could conceivably be that only one base-pair would need to be changed to move from SBNC to LNNC genes. But even this piece of information would not be enough to make the explanation ambitious, unless it is coupled at least implicitly with the claim that a cosmic ray causing that change is not astronomically unlikely.

Now, it is true that in many cases evolutionary stories will be unambitious, the way the original story here was. That is perfectly fine, and the problems for the creation doctrine do not arise in these cases. But the evolutionary story in the case of the existence of human beings is ambitious and hence irreducibly statistical. To see this, suppose that if it turned out that evolutionary theory did not increase the likelihood that intelligent animal would arise from unicellular organisms beyond the likelihood that a high energy solar flare striking a mineral-rich planetary surface would all at once produce a mature intelligent animal. Then evolutionary theory would not be particularly exciting. The astronomically unlikely theory that intelligent life arose through a high energy solar flare all at once producing intelligent animals would do little undercut a design argument based on the existence of life, and too the evolutionary theory would just as little. In the case of both theories, even if the theory were known to be true, the design argument would be largely intact: one would ask why the solar flare happened or why the astronomically unlikely sequence of mutations happened. Since evolutionary theory does undercut Paley-style design arguments, it must be more ambitious than that—it must include at least handwaving statistical claims.

III. The statistical claims are not a part of evolutionary explanations. One might also object that the probabilities are not a part of explanations. All the explanatory work is done by the first-order causal relations in the full ancestral histories, perhaps as in my largely made-up story of the long-nosedness of northern Europeans. However, probabilistic claims enter the theory not in an explanatory way but in order to defeat teleological arguments. Thus, if the probabilities involved were very low, the ancestral histories would call out for a further non-evolutionary explanation. But as it is, no further explanation is called for because the probabilities are not that low and so the full explanation is that given by the first-order claims. Thus, the probabilistic claims are extrinsic to evolutionary explanation, even ambitious evolutionary explanation. They only come in to defeat alternate hypotheses, namely the ones that posit design.

This interpretation of the explanatory work of ambitious evolutionary theory, however, is implausible. It is part of the explanatory attractiveness of evolutionary theory that it yields, assuming for the sake of argument that it does indeed so yield, probabilities for the development of complex organisms that are not astronomically small. Suppose cosmic rays strike a memory chip and produce junk rather than Hamlet and we want to know why it produced just rather than Hamlet. It surely should be part of the explanation that the production of junk is much more likely than the generation of Hamlet. When we realize that the production of junk is highly probable under the circumstances, we feel that we now understand why there is junk there. It may be that our understanding is not complete, but surely the probabilistic data is a crucial part of our understanding. Merely being told the causal fact that the junk in memory was caused by the cosmic ray would not be as satisfactory.

And even if the above claim about explanations were not correct, in any case, claims about mutation probabilities must enter into any robust neo-Darwinian evolutionary story. If the probability of mutation were sufficiently small, for instance because all cells implemented highly redundant error-correcting coding, then neo-Darwinian stories would have no plausibility at all as a scientific theory. It is going to be a part of the evolutionary story broadly conceived, at least, that mutations have a certain probability that is not so small that it will probably never occur in the history of the universe. And no doubt more precise claims need to be a part of the story.

Now, it might be argued that, nonetheless, these kinds of statistical claims do not enter into explanations, and hence need not have the kind of relevance on which our arguments above hinged. But this is incorrect. Take the above example that the probability of mutation cannot be too small. If the probability here is computed relative to an irrelevant specification of initial conditions, then it is useless. This is particularly clear in the deterministic case where any event E will have conditional probability 1 relative to initial conditions specified as “those initial conditions that lead to E”, and conditional probability that is can be made arbitrarily small for an appropriate choice of state space S and a description “initial conditions falling in S”, assuming a state space with an absolutely continuous probability measure and assuming that E is not probabilistically inevitable.[12] And it is plausible that relevance conditions will not be very different here from the explanatory case.

4.2. The indeterministic no generalized middle knowledge case

4.2.1. The argument

What if, on the other hand, God lacks generalized middle knowledge or is not able to make use of it while creating (e.g., because doing so would lead to some absurdity like that raised by Robert Adams[13] for the case of middle knowledge)?

We may start off here by criticizing the very idea of this possibility as contrary to sovereignty. For while one might have to understand divine sovereignty in a way that is compatible with libertarian free will either in order to avoid the problem of evil or to avoid both the Scylla of universalism (the doctrine, rejected by all the major western monotheistic religions, that no one receives God’s ultimate punishment) and the Charybdis of thinking that a loving God predestines some to damnation, there is no similar reason to allow for random events that God doesn’t have generalized middle knowledge about. If God had generalized middle knowledge, then the threat to divine sovereignty would be less because God might then know, before (“before” understood in the order of explanation if God is outside time) deciding what to create, that things would turn out as he wished. This is perhaps not a very strong argument against the very idea of randomness, however, because God could still get whatever he wanted done simply by letting random events happen and as soon as they happened overriding the outcome if it did not fit with his plan.

Alternately, one might proceed more metaphysically. Quantum random events would then be mere brute facts. Why event A rather than B happened would neither be explained by science nor by the choice of any agent. While one might argue that libertarian free choices can be explained by the agent’s activity and/or the agent’s reasons[14], perhaps nothing like this can be said in the case of quantum random events which are not things done for a reason. Thus, they would be a violation of the Principle of Sufficient Reason (PSR). If this doesn’t bother one by itself one might further note that it threatens God’s role as the First Cause that there should be events, e.g., the occurrence of A rather than B, that are neither directly caused by God nor by anything standing in a chain of causes that goes back to God. Admittedly, as Haldane has noted[15], one might think of quantum systems as engaging in the same kind of “substance causation” that libertarian agents are often thought to engage in, and if so, then God would stand at the beginning of a chain of causes terminating in the existence of a substance followed by the event caused by that substance. But it is not clear whether the notion of substance causation makes sense once extended to cases of indeterministic non-personal causation. For instance, one might think that indeterministic quantum causation is a violation of the PSR, and that no instance of genuine substance causation is a violation of the PSR.

But let us suppose that the in-principle arguments against genuine randomness in a theistic world fail. We need to, then, examine the evolutionary proposal in detail. Let us grant the possibility of randomness without generalized middle knowledge.

It still is possible for God to intentionally guarantee the existence of the desired features of the universe without undercutting statistical explanations. To do that, God can first set up initial conditions in such a way as would make the statistical explanations provided by evolutionary theory correct. Since God does not know, or cannot make use of the knowledge, which initial conditions would result in which outcomes, he can do all this. God can ensure that the background conditions in the evolutionary statistical explanations are indeed the relevant ones.

Of course, this does not guarantee that the outcome God intends should occur. But we can now make one of two moves. First, we could weaken the creation doctrine even further to claiming only that what God intended (in the strong sense of “intend” in which whatever an omnipotent being intends happens) was not that certain features should arise, but that they should be likely to arise. Especially in connection with the existence of the biological human species this seems to be shortchanging the religious views. But, more plausibly, one could follow a modified version of a suggestion by Del Ratsch[16]. God intended that certain things should be generated. To that end, he set up processes that would be likely to lead to them, with the probabilities predicted by an ideal naturalistic evolutionary theory. In order to go from likelihood to certainty, however, God planned that should something go wrong, he would intervene and thereby ensure the desired result. It would be likely that, say, humans would evolve, but if they didn’t, God would miraculously intervene. On both accounts, to have compatibility with the first-order claims of naturalistic evolutionary theory, we need to assume that things turned out well—the biological human species in fact arose without the need for miraculous intervention.

This account is logically possible, and hence indeterministic evolutionary theory is logically compatible with a creation doctrine and the denial of GMK. But nonetheless, as we shall now see, this conjunction of views is quite unlikely. The reason for this is the high plausibility of the claim, made by Stephen Gould, that according to indeterministic evolutionary theory, it was in fact unlikely that these precise organisms would arise that did arise:

A historical explanation does not rest on direct deductions from laws of nature, but on an unpredictable set of antecedent states, where any major change in any step of the sequence would have altered the final result. This final result is therefore dependent, or contingent, on everything that came before—the uneraseable and determining signature of history.[17]

If one could turn back the clock, it would be likely that other solutions to evolutionary problems would arise, and in particular it is highly unlikely that the human species (even considered only in respect of the body and not the soul) should be exemplified, or any other particular species from the actual world, except for any species containing the initial organisms from which evolution was supposed to have started. If this is true, then the account currently under consideration presupposes God having been very lucky to get precisely what he wanted—it was very unlikely for him to have been so lucky. And this makes the account quite improbable, both on the theological grounds that God shouldn’t need to be “very lucky” and on the purely logical grounds that an account that entails that a particular event both happened and is very unlikely surely has correspondingly low prior probability.

Thus, this is an account that it is not rational to believe in. As an analogy, suppose that we believe that Fred intended that fifty coins that Bob indeterministically tossed in the air should land heads-up, and that they do indeed land thus. But suppose that we also know that Fred has psychokinetic powers that would allow him to supernaturally determine the outcome of any coin throw experiment by making subtle changes in the flight of the coins, and we believe that Fred was absolutely set on the coins landing heads-up. Then can it be rational to believe, on the evidence so far presented, that the coins simply happened to land thus? Surely not. It is very unlikely that the coins would have happened to land thus. But it is not at all unlikely that Fred would have modified their path in flight.

Going back to the evolutionary case, given God’s deliberately bringing about the existence of the human species, as per the strong creation doctrine, we have three possible scenarios if GMK is not available:

(S₁) it simply happens through the operation of natural causes that the biological human species arises;

(S₂) God intervenes in the process in some supernatural way that we would by now have empirically discerned; and

(S₃) God intervenes supernaturally in the process in some way that we would not have by now empirically discerned.

So now the question is which of the three scenarios it is or is not rational to believe in. Suppose K contains full information about the state of the physical universe at the time of the beginning of the evolutionary process together with the claim that the creation doctrine is true and the denial of God’s use of GMK. For simplicity, add to K the claim that God only intervenes in the evolutionary process if he has to. If this claim doesn’t hold, then the credibility of S₁ will go down, so it is a fair assumption to make. Then P(S₁|K) is very small, since it is equal to the probability that naturalistic processes, starting with the K-specified initial conditions, would bring about the existence of the human race. Let us say, for definiteness, that P(S₁|K)<0.000001, though the actual number is surely much, much smaller—remember that we are talking of the probability of beings with precisely the genetic code distinctive of humans arising. Moreover, P(S₁ or S₂ or S₃|K)=1, since S₁, S₂ and S₃ exhaustively describe the three mutually exclusive ways that the creation doctrine could be true without GMK being used. Since P(S₁|K) is very small, P(S₂ or S₃|K) = P(S₂|K) + P(S₃|K) is very close to 1, indeed greater than 0.999999.

Now, let E be the event of God’s noticing that the natural evolutionary processes are not leading up to the existence of biological human beings and need to be tweaked. Both S₂ and S₃ include E, and conversely, if E happens, then given K, either S₂ or S₃ must happen, since K includes the claim that God intentionally created the human species. Now, when E happens, God has to decide whether to intervene in such a way that his action would be noticeable by the early 21st century or not. It seems not unlikely that he would act in a way that would be unnoticeable, e.g., because it seems not unreasonable to believe that God would have a prima facie preference for letting the natural causes that he has created to work as much in accordance with their natures as is compatible with God’s plans, and hence smaller miracles would be preferred over bigger ones. Consequently, it would be not unlikely that God would intervene simply by controlling some or many mutation/recombination events, and if so, we would be unlikely to have any direct evidence of this, and it is not that likely that we would have indirect evidence, either.

If we add, as we may, to K additional background information implicit in the great monotheistic religions about God’s apparent preference for often working “behind the scenes” (consider, for instance, Christianity’s assurance that God does hear prayers together with the fact that overtly observable miracles are relatively rare—few people experience more than one or two of them in their lives, while prayers seem to have to be answered much more often for the promises of the Gospels to hold), then the above point becomes even clearer. Thus it seems that P(S₃|K&E) might even be more than 0.5, but in any case, very plausibly, P(S₃|K&E)>0.001.

Now, if the advocates of Intelligent Design (ID) are right, then we actually have strong evidence in favor of an agent’s intervening in the evolutionary process. In that case, it is very easy to make the case that it is unreasonable to believe S₁, and reasonable to believe S₂. Suppose, however, as many scientists think, that ID-advocates are wrong and we do not have empirical evidence of God’s intervening in the evolutionary process. If we can show that in this case it is unreasonable to accept S₁, we will be done. Let F be the fact that we do not have any empirical data indicating God’s intervention. Now, S₃ entails F and K&F holds if and only if K&(S₁ or S₃). Hence:

(4) P(S₃|K&F) ≥ P(S₃|K) / P(F|K) = P(S₃|K) / P(S₁ or S₃|K) = P(S₃|K) / (P(S₁|K)+P(S₃|K)).

Now, P(S₃|K) ≥ P(S₃|K&E)P(E|K) > 0.001 P(E|K). Observe next that, given K, E happens if and only if S₁ does not, since God intervenes if and only if he has to, we have assumed. Thus, P(S₃|K) > 0.001 (1P(S₁|K)) > 0.000999999 as P(S₁|K) < 0.000001. Using both this estimate of P(S₃|K) and the claim that P(S₁|K) < 0.000001, we get from (4) that P(S₃|K&F) > 0.999, and hence that P(S₁|K&F) < 0.001. Therefore, even if there is no empirical evidence of God’s intervention, it is not rational to believe that it did not happen—assuming something with epistemic probability less than 0.001 is not rational to believe in.

We can look at the argument this way. Suppose we start epistemically with the acceptance of the strong creation doctrine. We are, let us suppose, first religious people, and only secondly scientists. We then realize that if God were to make use of something like indeterministic evolutionary processes to produce human beings, he would very likely have to miraculously intervene, perhaps once, perhaps more than once. Thus, should we learn on scientific grounds that something like indeterministic evolutionary processes in fact occurred, we would have reason to suppose that somewhere in the process at least one miraculous event occurred, unless there is evidence specifically against the occurrence of such an event. Since our fossil record is plainly not sufficiently detailed to provide significant evidence against such a miracle, and there does not appear to be any other relevant evidence here, the theory that God used something like indeterministic evolutionary processes together with at least one miracle to produce human beings is going to be less probable than the theory that God used just indeterministic evolutionary processes. It is thus not rational to believe the latter theory.

Similarly, if we start with science and then come to accept western monotheism, learning that the human species was originally intended by God for exemplification should make us think that there was a miracle in its full ancestral history, over and beyond any infusion of soul. It is no longer rational at this stage to accept a full evolutionary theory.

4.2.2. Objections

I. A step-by-step decision-making process. The first objection relies on a fairly literal reading of Genesis 1. Given that it relies on specifics of Genesis 1, it only applies within Christianity and Judaism. Schematically, the central portion Genesis 1 is a sequence of texts each saying something like:

God said, “Let x occur.” And x occurred. And God saw that this was good.

These claims occur successively. Without insisting on the literal specifics, one might take the text to imply that God did his planning bit-by-bit rather than all at once. He created one thing. Evaluated it. Then decided to create another. And then created it. This reading takes the “vayomer” (God said) to mean “God decided”, and to imply a sequence of decisions each made after finding out the outcome of a previous decision.

This allows for a model of creation as follows. God actualizes some initial total state S