Dawkins essentially argues that if one observes some event that seems to be designed because it seems very improbable to have occurred, positing God as a hypothesis doesn't make sense, since God is more complex, and given His nature, He seems to be the most improbable hypothesis ever.

But in the case of designed events, those events have a beginning. There was a time that we can point out where that event did not exist, such as the case of life to have arised by chance. There was a time where life did not exist. But we can't say that about God since many conceptions of God are posited to be eternal. And if God is eternal, how does it make sense to say that He is the most improbable when He never really arose by chance?

Secondly, if God is a hypothesis that we haven't actually confirmed to exist, how can we say that He is improbable instead of simply saying we don't know? For example, if He is a necessary being, then by definition He is not the most improbable but rather has a probability of one.

I wanted to illustrate this with an example. Suppose one has an event occur in their life that seems to be a sign from God. Imagine if someone has a predictive dream where in that dream God appears and says "You will see your missing daughter tomorrow". You then see your missing daughter tomorrow.

Let's say you're deciding between two hypotheses that produced that dream: chance and God. There may be other possible hypotheses, but let's assume for the purposes of this particular hypothetical, we're just comparing the likelihood of these two. Now, the probability of you having that particular dream may be very low by chance. One however might say that God sending this down as a sign is even more improbable, since through Dawkin's logic, God's very existence is the most improbable of improbable events. One can then say it's more rational to believe that the dream occurred by chance. But why? What basis do we have of assuming this?

Of course, if God doesn't exist, then it quite literally is impossible for Him to send down a sign. In that case, the dream must have occurred by chance or some other hypothesis. But if He does exist, then how can we know His probability of sending down a dream? More importantly, we don't know if He exists, and hence, we don't know P(God sending down this sign).

If we don't know if He exists, how can we say that P(God sending down this sign) is > or < P(predictive dream occurring by chance)? Also, given that these are binary events (i.e. God either sent down a sign or He didn't; this happened by chance or it didn't), can we even assign probabilities here? Ultimately, that P value seems to be either 0 or 1 for both hypotheses.

  • 3
    I would put it this way: Dawkin's argument fails because the argument from complexity only applies to things that had a beginning. If something has a beginning, then you need an explanation for how it came to be, and complex things are hard to explain without design. However, God did not come to be, so our intuitions about probability and explanations don't apply. Alternatively, one could point out that in traditional Christian theology, God is not complex; he is the opposite of complex: he has no parts. Commented Oct 7, 2022 at 14:30
  • 1
    First, Dawkins is not a philosophical source. Second, God cannot be proved through the senses or logic. It is a futile argument to try and prove or disprove the existence of a non-sensual Being through logical argumentation. Commented Oct 8, 2022 at 4:41
  • 1
    This relates I think to en.wikipedia.org/wiki/Pascal%27s_wager#Criticism & the wider topic of en.wikipedia.org/wiki/Miracle - as it says, Dawkins objects primarily in relation to Occam's Razor
    – CriglCragl
    Commented Oct 8, 2022 at 13:20

7 Answers 7


This question is based on a widespread but fallacious assumption. Probabilities are not inherent in things.

Probabilities are inherent in our representation of them, and they typically depend on our presuppositions, which in everyday life are typically never fully if at all made explicit.

So, there is no probability inherent in God. For a start, anything that would be inherent in God would first require that God exists, which just shows that nothing of what we say about things is inherent in things.

Thus, the improbability of God is properly understood the improbability that whatever we may say about God be true, including that it exists to begin with. This is what is really meant by the improbability of God.

The improbability that whatever we may say about God be true just inevitably follows from the fact that we know nothing about God. All that we know is our own ideas about God, and our ideas do not amount to any knowledge about God, let alone that it exists at all.

When you know nothing about something, the probability that it exists is not zero. It is just infinitely small. So, it is fair to say that the probability that God exists is infinitely small. It is not zero only because we don't actually know that God doesn't exist. However, we also don't actually know that Zeus, Pazuzu or any of the the zillions of divinities imagined by people at some point in human history do not exist, but that in itself doesn't make it reasonable to believe that they do.

The fact that the probability of God is infinitely small comes from the fact that we could potentially imagine an infinity of alternative and mutually exclusive explanations to our own existence, all of them equally probable as long as we don't have any empirical data justifying any of them.

  • I don’t see how being able to imagine an infinite number of theories that can’t be empirically justified implies that the probability of any of them must be infinitely small. This makes even less sense when considering your initial point of how probability is not inherent in things. Just because we can’t disprove something doesn’t mean that the being has a probability that is infinitely small. It could be 1. It could be 0. It could quite literally be anything. The more correct thing to say would be that it is unknown.
    – user62907
    Commented Oct 8, 2022 at 11:59
  • @thinkingman 1. "the being has a probability (...) infinitely small" I didn't say that. 3. "how being able (...) must be infinitely small" Think of an ordinary die... Before throwing the die, we can imagine six different a priori possible outcomes, all equally probable because there is nothing we know about the throw that would make one more probable. Hence, all six sides have exactly the same probability., therefore, 1/6. Now, God is just one possibility among an infinity of other, equally probable, possibilities, so the probability of God is 1/∞. Buy yourself a book on probabilities. Commented Oct 8, 2022 at 16:51
  • That doesn’t follow. We know that dice will land somewhere even if we don’t know where it will land. We also know they’re equiprobable because that’s how the frequencies turned out. It’s based on data, not speculation. But in the case of imagined entities, either of them could be possible or impossible. You don’t actually know their probabilities. You can talk about their “epistemic” probabilities but that’s also an unknown. You have no basis by which to make any probability judgment of some entity you don’t even know exists
    – user62907
    Commented Oct 8, 2022 at 18:46
  • @thinkingman "But in the case of imagined entities, either of them could be possible or impossible" Either of them is equally probable because we know nothing about any of them, and so we know nothing that would make one more probable, in the same way that either side of a die is equally probable because we know nothing that would make one more probable than the others. Buy yourself a book on probabilities. Commented Oct 9, 2022 at 10:54
  • “ because we know nothing that would make one more probable than the others” Uh no. You know from history how dice works and from its overall frequency that it lands on each side about 1/6 of the time. You’re unknowingly using a particular probability interpretation without realizing it’s just an opinion. There are multiple probability interpretations and YOU need to read a book on those interpretations. There is nothing that says when you don’t know something, you must assume that everything is equally probable. When you don’t know the distribution of something, you just..don’t know
    – user62907
    Commented Oct 9, 2022 at 17:30

Of course we can assign probabilities, the more interesting question would be how could we know that these probabilities are anywhere correct.

Also could you give the quote from Dawkins? The way you describe it sounds more like he's mocking intelligent design. So where intelligent design argues that "what is, is so complex, that it must be the result of an intelligent design and not just random chance" while he makes the tongue-in-cheek remark that "what is, is so simple that even we can grasp it so if there were to be a god, this would most certainly not meet his standards (he'd be way more complex)". So he's essentially twisting the argument to make intelligent design look ridiculous by praising god, hoping this would force his opponent to forfeit their believe in either or both of these concepts.

Otherwise in terms of the probability of god, well it first of all depends on what that god would look like in the first place. Like if you had a clear description what he looks like and what he does and why you could look for him in different places and from finding or not finding him there derive a probability of finding him elsewhere. But most often the concept of a god is so vaguely defined to begin with that neither his presence nor absence can be proven or even estimated.

  • There are pictures of the Flying Spaghetti Monster, which make it look pretty unlikely. Clear descriptions of God are probably all going to seem unlikely.
    – Scott Rowe
    Commented Oct 10, 2022 at 16:28

You are not thinking about this question correctly, and neither is Dawkins, though his approach is less incorrect.

There are multitudes of possible God hypotheses, and thousands which have been claimed.

None of them have P of zero or 1. P(true) is not based on what is real, but it is our judgement call of the likelihood of a P being true or not. What P() is based on, is the principle that we DON'T have any way to get certainty, and we must make inferences about reality. This is called Indirect Realism.

Dawkins realizes this, that we are operating in indirect realism, and what the nature is of probability evaluations of claims about reality.

Dawkins is asserting that one can use Occam's Razor to sort between claims about our reality, and discount more complex claims in favor of simpler, or more parsimonious ones. This, historically, has not been a valid principle. Copernicus, for example, agreed with Ptolemy that spheres or circles are "simple" and therefore more appropriate to use in Cosmos models, and put planets in circular orbits around the Sun. Kepler, however, showed that orbits being elliptical fits the data far better than circular orbits. Ellipses are FAR more complex than circles -- but that is how our world works anyway.

We have discovered that our world is astonishingly complex. The multitudinous branches of science have each discovered enough complexity to our world, that to understand just one specialty within that science requires that one narrow one's studies to that specialty area. Nobody has the ability to understand all of one science, much less all about our world, in a lifetime of study. If one compares this model of an astonishingly complex world revealed by science, to the model of a dreaming mind that just imagined all this world -- the dreaming mind is far simpler. Dawkin's criteria is -- not actually a wise one to use.

When one has multiple models that fit a data set, it is generally a good idea to accept the "simplest" as the best working hypothesis, but this is a fraught process, as we are notably able to rationalize what is "simpler". Dawkins argues God hypotheses are infinitely complex, while theists often argue them to be infinitely simple. At least one side of this dispute is engaged in rationalizations. Karl Popper suggested a better standard of parsimony -- which is more TESTABLE -- I.E. which makes more useful predictions we can use to make decisions with, and then potentially test and evaluate? That we are a dreamer -- is not a usefully testable model -- and it is compatible with any possible outcome, it makes no predictions.

Note, what Popper's approach encourages, is to get MORE data, such that we eventually don't have multiple models that fit our data set equally well -- some would fit it poorly as more data comes in that they have difficulty with, while others would continue to be predictively useful. As I noted, this has not historically been the "simplest" of models.

So -- the Popperian approach would be to look for test cases of God claims, vs. non-God models of the universe. For instance, many of the God claims involve the uniqueness and specialness of humans, and of our world. Both Biology and astronomy have shown that the Earth is far from unique, and that humans are clearly animals, and closely related to the rest of the animal kingdom, hence neither the earth nor humans are unique. This does not refute those models, but it is a significant evidence against them.

One can compile these evidences and draw a plausibility evaluation based on them, for most God claims. Hence Dawkin's project is achievable, although his method, -- of arguing based on "simplicity", is not itself correct.

  • 1
    Simplicity was too simple to be the correct method, I guess. I think that ellipses are simpler than circles, because more likely, and otherwise you have to explain what keeps the orbit circular in the presence of perturbation.
    – Scott Rowe
    Commented Oct 10, 2022 at 16:19
  • @Dcleve. I think that I accidentally downvoted your post. Sorry. Commented Oct 14, 2022 at 17:09
  • @MarkAndrews thanks for the message. I extend Popperian empirical testing throughout philosophy, and consider it to be very applicable to religious claims. Note, I am religious myself.
    – Dcleve
    Commented Oct 16, 2022 at 0:41

Starting with the abstract concept of God, then looking for its existence probability, is backwards. You need to assign some observable as the basis, with probability 1, then attempt to assess the contingent probability of increasingly specific descriptions of that observable. For example, we know that Dawkins exists because we have a set of observable evidence, to whose formally unknowable cause(s) we assign the label Dawkins. Dawkins may be a male biologist, a crossdressing woman with a fraudulent degree, or even an elaborate hoax with no human anticedent, all with certain nonzero probabilities, but P(Dawkins exists)=1. In practice, we're sufficiently confident that P(Dawkins exists & Dawkins is a male biologist and popular proponent of atheism) is so close to 1 that it makes sense to call the cause Dawkins, not "the putative cause of Dawkinsness", but it is this latter, not the former, of which we can be certain.

If we approach the question in the opposite direction, starting with an infinite set of mutually exclusive perfectly specific evidentially indistinguishable models of the man Dawkins, we will discover that the probability of any one of them being true is also zero.

Since e.g. the factors that caused a certain holy book to be written in a certain way is the thing in dispute, assigning the label "God" to the observation and thus assigning P(God)=1 is disingenuous. It would be better to name the causes of the observable evidence something neutral, then assess various traits associated with God for their contingent probabilities. Here we could start with, e.g.

Let Z be the unknown cause(s) of a certain scripture

P(Z exists)=1

...then assess the various properties attributed to God (magic powers, immortality, omniscience, sinlessness), like the various properties attributed to Dawkins (male, human, biologist, atheist, author), for their probability of corresponding to Z or not.


The problem can be construed as one concerning whether existence claims can be probabilistic, and in what way. For usually, let's say I wake up one morning and it's cloudy, and I think, "It's likely to rain." I will have a finite background sample of cloudy and rainy days in mind, and I will be thinking that a sufficient number of cloudy days turned out rainy before, hence the relative likelihood of rain today.

By contrast, what are the preconditions for evaluating the likelihood that a given object exists? We don't omit tachyons from prevailing consideration in physical explanations because their existence is or is not likely. We omit them because their nature makes them either exceedingly difficult or outright impossible to empirically differentiate from bradyons, from our vantage. The omission of their existence from physical theories is a priori enough in a way that doesn't depend on the more generic probability (or lack thereof) of their existence. We don't think, "In a hundred explanations I came up with before, I didn't need tachyons to do any explanatory work; so next time I come up with an explanation for something, I probably won't have to bring in tachyons, either." Tachyon explanations are ruled out because they contradict the means we have available of explaining specific things (and then tachyonic explanations are ruled in, by the way, as it turns out).

Similarly, we have no occasion to think something like, "Nine times out of ten, when I saw a universe start to exist, I saw that God was the cause of their existence," or vice versa. In fact, there's nothing specific for us to compare our impression of the initial expansion to at all, in that we don't even see it in itself, but interpret it as a cause from past existence. "Every time I saw that something was an effect before, I found out that its cause was also an effect of a prior cause," does not improve the likelihood of there being further causes and effects, in either a future or past direction; this is a principle superimposed on the situation independently.

It would be like asking, "What is the probability that there would be probabilities at all?" or, "What explains the existence of explanations?"


Given Eternity, the idea that GOD is improbable is actually false.

Otherwise, no, I don't believe you can assign probabilities. That would require a systems-level thinking "above the box" in which you inhabit, so to speak.


Can we assign probabilities to God and is the argument from improbability from Dawkins valid?

I am not saying that this is the answer to the question, but I offer a way of looking at it.

This question is interesting. On the one hand, the chances of God having sent the dream are said to be infinitesimally small; on the other hand, if God is eternal, then we have a series of infinitesimally small chances multiplied by infinity.

So there would be some chance, however small, that this one dream, out of the billions each night, was sent by God. Proving that would be quite the trick, but the nominal chance would be there.

However, here the rational decision is that the dream plus its happy ending together are only a secular coincidence. Each day, there are trillions upon quadrillions of transactions where people meet each other in some unplanned way. The “dream and meeting” scenario might occur, say, 100 a day to people around the world.

You must log in to answer this question.