Home » Library » Modern Library » Robin Collins Design Defended

Robin Collins Design Defended

Clarifying the Case for Cosmic Design (2008)

Robin Collins


          Clarifying My Argument

          Draper’s “Understated Evidence” Objection

          The Idea of Probabilistic Tension

          Beauty of the Laws of Nature

          Final Comments: The Intrinsic Probabilities of Theism and Naturalism

Clarifying My Argument

First, I would like to thank Quentin Smith and Paul Draper for participating in this debate and pointing out some weak points in my presentation of my opening case. This has helped me clarify the nature of the argument that I presented, which I believe is the primary benefit of a debate such as this. I agree with several of Draper’s points, especially since having further thought about the argument in the two years since I originally wrote the opening case. I agree, for instance, that the evidence I cite (e.g., the fine-tuning evidence), gives us no reason to believe in a generic design hypothesis. Indeed, I now think that the language of design is the wrong way to frame the argument, for the very reasons Draper cites: namely, a generic design hypothesis gives us no reason to expect a universe with the features I discuss since it tells us nothing about the motives of the designer. Therefore, by the likelihood principle, the features I cited provide no support for this hypothesis.

I also agree with Draper that my opening case by no means establishes the existence of God, and contrary to the impression one might get from my essay, I did not intend it to. Rather, I was thinking of the data I cited as offering evidence for the existence of God. The approach I was taking is that of what I call the restricted version of the likelihood principle: given two non-ad-hoc hypotheses, H1 and H2, and a body of data E, if E is much more surprising (that is, has a much smaller epistemic probability) under H2 than H1, then E offers significant evidence in favor of H1 over H2. I then claimed that the evidence I cited was much more surprising under the naturalistic single-universe hypothesis (NSU) than under theism, and thus counted as evidence for theism over the NSU.[1]

The restricted version of the likelihood principle is simply the standard likelihood principle of confirmation theory with the restriction that the hypothesis being confirmed is not ad hoc. A sufficient condition for a hypothesis not being ad hoc is that there are independent motivations for believing the hypothesis apart from the data E itself. To illustrate the need for the restricted version, suppose that I roll a die twenty times and it comes up in some apparently random sequence of numberssay 2, 6, 4, 3, 1, 5, 6, 4, 3, 2, 1, 6, 2, 4, 4, 1, 3, 6, 6, 1. The probability of this sequence coming up by chance is one in 3.6 x 10^15, or about one in a 3.6 million billion. To explain the die coming up in this sequence, suppose I invented a hypothesis that there is a demon whose favorite number is just the above sequence of numbers (i.e., 26431564321624413661), and that this demon had a strong desire for that sequence to turn up when I rolled the die. Now, under the assumption that the demon hypothesis is true, it would not be surprising (i.e., epistemically improbable) for the die to come up in this sequence. Consequently, by the unrestricted likelihood principle, the occurrence of this sequence would strongly confirm the demon hypothesis over the chance hypothesis. But this seems counterintuitive: given a sort of common sense notion of confirmation, it does not seem that the demon hypothesis is confirmed.

Now consider a modification of the demon case in which, prior to my rolling the die, a group of occultists claimed to have a religious experience of a demon they called “Goodal,” who they claimed revealed that her favorite number was the number 2643156432162441366, and that she strongly desired that number to be realized in some continuous sequence of die rolls performed by Robin Collins in the next month. Suppose they wrote this all down in front of many reliable witnesses days before I rolled the die. Certainly, it seems, the sequence of die rolls would count as evidence in favor of their “Goodal hypothesis” over the chance hypothesis, given that the die was a fair die. Of course, in this circumstance, the Goodal hypothesis was already advocated prior to the rolling of the die, and thus the restricted likelihood principle implies that the sequence of die rolls confirms the Goodal hypothesis.

Even though the sequence of die roles would count as evidence in favor of the Goodal hypothesis, it would not establish that it is likely: we might have other strong reasons to reject it, such as philosophical arguments against the existence of such disembodied beings. So, this likelihood approach to arguing for theism leaves open the question of whether, when all the evidence is considered, theism is objectively more likely to be true than the NSU. In order to do this within the likelihood framework, one would have to assess the prior epistemic probability of theismthat is, its epistemic probability before the evidence E that I cited is consideredsomething I believe is difficult to do. (I briefly discuss this issue at the very end below.)

Nonetheless, since theism was advocated prior to the fine-tuning and other sorts of evidence that I cited, it follows from the restricted version of the likelihood principle that this evidence counts in favor of theism, even if we cannot say at the end of the day that theism is true. This more limited claim, I submit, is nonetheless highly significant. One could argue, for example, that in everyday life and science we speak of evidence for and against various hypotheses, but seldom of their prior probabilities. If, for example, one were to ask most scientists why they (provisionally) accept general relativity’s approximate truth or future adequacy for doing cosmology, they would probably cite the evidence in its favor, along with some of Einstein’s motivations. They would probably not talk of the prior probability of his theory, either epistemic or otherwise. I can imagine them saying, “I have no idea what the prior epistemic probability of Einstein’s theory is; all I will say is that Einstein had motivations for considering it and there are at least three strong pieces of empirical evidence in its favor.” Indeed, I think it would be very difficult to estimate the prior probability of general relativity’s approximate truth or adequacy in any objective manner, since we would have to weigh incommensurable factors against each otherthe simplicity of the mathematical framework of general relativity against such things as the philosophically puzzling character of the idea of a four-dimensional space-time being curved. Theism suffers an analogous problem with prior probabilities. One can avoid the issue of prior probabilities, however, by using the restricted version of the likelihood principle to argue that the fine-tuning and other features of the universe I cited offer significant evidence for theism over the NSU, though they do not themselves show that theism is likely to be true.

Draper’s “Understated Evidence” Objection

Now let’s consider Draper’s understated evidence objection. According to Draper, “while [Collins] bases his argument on the fact that complex intelligent life of some sort exists in some universe or another, he ignores the fact that the only sort of intelligent life we know to exist is specifically human and exists specifically in this universe.” Further, Draper claims that, “given that intelligent life of some sort exists in some universe, the fact that the only intelligent life we know of is human and that it inhabits this universe is very many times more probable on single-universe naturalism than it is on theism,” a claim he labels claim D. Draper then goes on to make two major claims in support of D:

  1. “Given the existence of intelligent life, it is certain on single-universe naturalism that it exists in this universe. This is not so on theism, which as Collins admits, makes the existence of many universes likely.”

  2. “It is also true that on theism the existence of intelligent beings as unimpressive and flawed as humans is very unlikely.”

In response to (1), I consider the claim that “human life exists in this universe” tautological and hence true by definition. The reason is that we have no independent means of specifying the universe that “this universe” refers to other than “the universe we inhabit,” since we cannot stand outside our own universe to establish a basis for an independent reference of the indexical “this.” Consequently, the claim that “human life exists in this universe” simply becomes the tautology “human life exists in the universe that we [that is, human beings] inhabit.”

In response to (2), I think it is likely that God created other universes, other “systems of nature,” and other creatures. Consequently, an advocate of the fine-tuning argument need not claim that humans are more valuable than other possible beings, a point which Draper acknowledges. All one has to argue is that it is plausible to suppose that the type of creatures that we arefinite, vulnerable moral agentsadd an overall positive value to reality, thus giving God a reason to create a world in which such beings could arise. One such positive value is the ability to act virtuously in the face of temptation, such as acting with courage or self-sacrificial love; another is the ability to make a difference for good or ill in another creature’s life. A world in which these goods can be realized is likely to look a lot like our world, one in which creatures are limited, subject to natural laws, and the like. Thus I think God would have significant reason to create a world that could give rise to embodied, highly vulnerable moral agents, in addition to other types of worlds and creatures. As Swinburne argues in some detail, the significance of the body is that it functions as “a public place where our intentions are translated into basic actions, and incoming stimuli are translated into sensations and beliefs, and processes give rise to desire and thoughts. When there is such a physical object, we and others can damage or improve these processes.” Hence, he argues, “in order to have the kind of limited freedom and responsibility [for each other’s welfare] that I have analyzed, creatures need bodies” (2004, p. 127).

This response, however, raises a more serious understated evidence problem for the argument based on the fine-tuning for embodied moral agents (but not the elegance of the laws of nature) that is implicit in Draper’s response: the existence of evil in the world. Until thinking carefully about Draper’s objection, I had always assumed that one could address the problem of evil separately from the fine-tuning argument. I now realize that the type of embodiment that results in the goods mentioned above (e.g., being vulnerable to one another) will almost certainly bring with it various sorts of evil (e.g., suffering as a result of natural law or the free choices of other embodied agents), unless God constantly intervenes, which would defeat the good of embodiment mentioned above. Thus, apart from an adequate theodicy, it is not clear that a reality which contains a universe with embodied moral agents is, all things considered, better than a reality without such agents.

Nonetheless, unless we have good reason to believe that the existence of evil could not be outweighed by the goods that come from embodiment, I think it is plausible to hold that the existence of a universe with embodied beings subject to natural laws is not improbable under theism (and perhaps also plausible to hold the opposite of this claim). If the argument in my opening case is correct, however, we have good reason to believe that a universe in which the constants and laws are fine-tuned to allow for such beingsand has the other features that I mentionedis epistemically very improbable under the NSU. Accordingly, we can be more confident of the improbability of such a universe under the NSU than under theism. Thus, unless we have good reason to believe that the existence of evil is very improbable under theism (see section 2 of this online book), the combination of the fine-tuning data and the existence of evil supports theism over the NSU. The beauty and elegance of the laws of nature tip the balance in favor of theism even more.

The Idea of Probabilistic Tension

Finally, part of Draper’s objection results, I believe, from not clearly understanding what I meant by the naturalistic single-universe hypothesis, which perhaps is the result of my own lack of clarity. According to Draper,

[B]y ‘naturalism’ Collins means the hypothesis that physical reality did not result from transcendent (nonnatural) intelligent design. By ‘single-universe naturalism,’ Collins means naturalism conjoined with the view that physical reality consists of nothing more than our own (expanding) cosmos [italics mine].

One major problem with this way of defining naturalism is that the phrase “physical reality consists of nothing more than our own (expanding) cosmos” builds into the hypothesis of naturalism that our particular universe exists. As I defined it, the naturalistic single-universe hypothesis (NSU) is the hypothesis that “there is only one universe and it exists as a brute, inexplicable fact.” This does not build in any hypothesis about the structure of the single universe that does exist; it also excludes any transcendent explanation of the universe, not just a theistic one. One can, of course, always reject this definition of single-universe naturalism, and instead build into one’s hypothesis that our particular universe existsthe one with the laws of nature that are structured just right for the existence of embodied moral agents. But then, one has just transferred the improbability up one level to that of this new NSU hypothesis itself.

This new NSU, which I will call the elaborated NSU hypothesis (NSUe), simply consists of the conjunction of the old NSU and the claim that the universe that exists has an embodied-moral-agent-permitting structure (EMAS): that is, NSUe = NSU & EMAS. In that case, the new hypothesis exhibits a high degree of what I call probabilistic tension, which I define as one conjunct of the hypothesis (in this case EMAS) being very improbable conditioned on the other conjunct (NSU): that is, P(EMAS/NSU) << 1.[2]

Now, probabilistic tension is a black mark against a hypothesis. To illustrate, consider cases in which a jury takes a defendant’s fingerprints being found on the murder weapon as evidence that the defendant is guilty. In such cases, the fingerprints are taken as evidence of guilt because the jurors implicitly reason that it is very unlikely that the fingerprints would have occurred on the murder weapon if the defendant is innocent, but not unlikely if the defendant is guilty. Thus, by the likelihood principle, the fingerprints strongly confirm the guilt hypothesis over the innocence hypothesis. But it does not confirm the guilt hypothesis over an appropriately formulated innocence hypothesis, say, for example, the hypothesis that the defendant did not touch the murder weapon conjoined with the hypothesis that someone else, with almost identical fingerprints, touched the weapon. The reason is that this hypothesis entails that the fingerprints will appear to match, and hence the matching will not be more surprising under this hypothesis than under the guilt hypothesis. Nonetheless, this hypothesis suffers from severe probabilistic tension. One conjunct of the hypothesis (that some other person with almost identical fingerprints touched the weapon) is very improbable on the other conjunct (that the defendant is innocent), since it is extremely rare for two people to happen to have virtually identical fingerprints. Thus, this probabilistic tension gives us strong reason to favor the guilt hypothesis over this latter innocence hypothesis, even though this new innocence hypothesis is not itself disconfirmed by the fingerprints.

Similarly, I argue, the elaborated naturalistic hypothesis suffers from a probabilistic tension from which even an elaborated theistic hypothesis does not suffer. The elaborated theistic hypothesis builds into the theistic hypothesis that God “desired” or “intended” to create a world with embodied, highly vulnerable moral agents. It is reasonable to think, I would argue, that this does not suffer from a similar sort of probabilistic tension as the elaborated atheistic hypothesis. To see this, note that, as Swinburne has argued (2004, pp. 99-106) the hypothesis that God is perfectly free and omniscient means that God’s desires would be entirely the result of what is objectively goodthat is intrinsically desirable. The reason is that to understand a state of affairs as objectively good in and of itself gives an agent a reason to bring that state of affairs about. Since God is perfectly free, God has no countervailing desires, and hence God would desire the good. Now as argued above, it is reasonable under the theistic hypothesis to think that the existence of limited, vulnerable moral agents is an overall good despite the evils that almost certainly would accompany the existence of such agents. Given that the existence of such agents is an overall good (for the reasons stated above), God would have some motive to bring them about, and hence God’s desire to bring such a world about would not be surprising under the theistic hypothesis. Hence it is reasonable to think that the elaborated theistic hypothesis does not suffer from the same sort of probabilistic tension, even when the existence of evil is considered.

Beauty of the Laws of Nature

Draper also questions my use of the beauty of the laws of nature as evidence for theism. What I mean by the beauty or elegance of the underlying structure of the universe is that it exemplifies simplicity with variety: our world seems to have enough complexity to sustain intelligent life (and to be endlessly fascinating!), while at the same time this complexity can be organized via an underlying mathematical structure with a relatively high degree of elegance and simplicity. Indeed, if, as physicist Euan Squires hypothesizes (1981), our universe is the simplest that could allow for complex life, this balancing would have to be quite extreme. This sort of complexity with underlying simplicity exemplifies the classical notion of beauty as defined, for example, by William Hogarth in his 1753 classic The Analysis of Beauty, where he famously used a line drawn around a cone to illustrate his definition of elegance as simplicity with variety. According to Hogarth, simplicity apart from variety, such as a straight line, is boring, not elegant or beautiful; with variety, however, it exemplifies elegance, such as illustrated by a line wrapped around a cone.

I thus agree with Draper that if all one means by the universe being elegant is that it is simple, then this elegance does not support theism: one could simply explain it by an appeal to a principle according to which the world is more likely to be simple than complex. The problem with such an explanation is that the universe is not nearly as simple as it could be, both in terms of the structure of its laws and its initial conditions. For example, a universe with only two particles and a single law that dictated that they move at a constant velocity with respect to each other would be a much simpler world. It is the balance of simplicity and complexity that makes it elegant, and thus points to a theistic explanation.

Final Comments: The Intrinsic Probabilities of Theism and Naturalism

So, I think Draper’s objections can be adequately answered. If I were an atheist, I would raise a different sort of objection. I would grant that the fine-tuning and related evidence supports theism, but nonetheless ask whether or not at the end of the day the theist is really better off than the atheist who accepts the type of universe (or multiverse) that we inhabit simply as an inexplicable fact. Against this, the theist could claim that the elaborated atheist hypothesis (labeled NSUe above) suffers from severe probabilistic tension that the elaborated theist hypothesis does not.

Although I think this response is good as far as it goes, the atheist could then press the case that theism suffers from other severe defects. These purported defects can be nicely illustrated by Draper’s comments about my previous belief (as expressed in my opening case) that the evidence for theistic design was as strong as that for common ancestry. The reason I believed this is that barring the considerations of the problem of evil above, both hypotheses render unsurprising a range of facts that would be surprising under the major contender hypotheses. Nonetheless, I recognize that one could argue that the theistic hypothesis is in some sense much more puzzling. The thesis of common ancestry involves a fairly small extrapolation from the common observation of offspring being both similar and different from their parents, and then simply extrapolating this process of minor differences occurring over a vast stretch of time. In contrast, one might argue, the God of classical theism is truly radically different from human beings or anything else we observe. For example, in classical Western theism, God not only creates the world ex nihilo, but also “decides” what to create in some timeless or “logical” moment before all time, whatever exactly that means. Because of the purported radical otherness of God, in much of the tradition this has led to claims that we can only truly say what God is not, not what God is. Richard Swinburne’s concept of God as simply an everlastingly omnipotent, omniscient, and perfectly free agent (Swinburne, 2004, pp. 93-96) is certainly less puzzling than the God of much of the classical theistic tradition, though much of the puzzle still remains.

Perhaps these issues can be resolved on their own terms. For example, Swinburne has extensively argued for the coherence of his definition of God (Swinburne, 1993). Here I will only briefly discuss one response that I would give to this issue.[3] The response begins by noting that the world seems to contain two radically and irreducibly distinct kinds of entities: impersonal entities such as rocks and personal, conscious agents such as ourselves. Given this, we naturally are led to initially speculate that ultimate realitythat is, what ultimately lies behind everything elseis either personal or impersonal (or more like the personal than the impersonal), or some combination of the two. Suppose that it is more like the personal. Then since agency and consciousness are the essence of the personal, if everything originates in the personal, then that ultimate reality would not be limited by any contingent thing (since it gives rise to all contingent reality), and thus plausibly could be thought to possess agency and consciousness that would only be limited or constrained by what is absolutely impossible, such as not being able to do what is logically contradictory. Unlimited consciousness implies that such an agent would know which states of affairs are objectively good (both aesthetically and morally) and which are not. Also, as argued above for God, an agent’s being completely unconstrained by anything contingent seems to imply that the only motivations such a being would have is the perceived intrinsic value of a state of affairs.[4] Thus, such a being would be perfectly goode.g., unlike humans, such a being would not have desires grounded in bodily needs, or resulting from having evolved, but only based on perceived objective value. Such an unlimited, perfectly good, personal agent is at the core of what theists have meant by God.

Now, one might legitimately find the hypothesis of such an agent puzzling and paradoxical, and be uncertain of its ultimate coherence. In reply, the theist could point out that the alternative that reality ultimately is grounded in the impersonal is also puzzling, paradoxical, and arguably incoherent. This, I believe, is the main thrust of the cosmological argument, which basically argues that assuming contingent things (such as our universe) exist without a personal cause is deeply metaphysically problematic, if not incoherent. Currently, for me, the cosmological argument draws a tie: on the basis of the mere existence of the universe (without considering its structure), I find both the personal and impersonal hypotheses deeply puzzling and worry about their coherence. When I look at the underlying structure of the universe, however, it seems to me that it provides significant evidence in favor of theism over the naturalistic alternative. Or put in terms of probabilistic tension, given that my arguments based on the fine-tuning and other features of the universe are correct, we know that naturalism suffers from a high degree of probabilistic tension, whereas even when the existence of evil is considered, we do not know this about theism.[5] Thus, given that, everything else being equal, it is rational to choose the hypothesis with the least amount of known probabilistic tension, theism is the intellectually rational choice.[6]

Proceed to the Next Debate


Collins, Robin (forthcoming). “The Argument from Fine-tuning” in The Blackwell Companion to Natural Theology, 2nd edition, ed. William Lane Craig and J. P. Moreland. Oxford, UK: Blackwell.

Hogarth, William (1753). The Analysis of Beauty. London, UK: J. Reeves.

Squires, Euan J. (1981). “Do We Live in the Simplest Possible Interesting World?” European Journal of Physics 2 (January 1981): 55-57.

Swinburne, Richard (2004). The Existence of God, 2nd edition. Oxford, UK: Oxford University Press.

Swinburne, Richard (1993). The Coherence of Theism, revised edition. Oxford, UK: Oxford University Press.


[1] In the opening case, I briefly argued that multiverse naturalism had similar problems to single-universe naturalism and thus I will not consider it further here.

[2] If P(EMAS/NSU) << 1, then since P(EMAS & NSU) = P(EMAS/NSU) x P(NSU), it follows that P(NSUe) = P(NSU & EMAS) << 1.

[3] I find another response, based on the ultimacy of value, even more powerful, but I cannot adequately develop it here. See Collins (forthcoming).

[4] For philosophers who believe in objective value, the intrinsic value of a state of affairs is typically considered a necessary property of the state of affairs, and thus not contingent.

[5] Of course, this assumes that the evidential problem of evil can be largely neutralized by a defense or a theodicy and that the sum total of other considerations does not weigh heavily against theism.

[6] I would like to thank the John Templeton Foundation for a grant that helped support this work, and David Schenk and Caleb Miller for reading through an earlier version of this response.

Copyright ©2008 Robin Collins. The electronic version is copyright ©2008 by Internet Infidels, Inc. with the written permission of Robin Collins. All rights reserved.

all rights reserved