It is always flattering when your work generates commentary, so I am pleased to see that the Discovery Institute has published a series of articles responding to my discussion/debate with Intelligent Design theorist Dr. Stephen Meyer. The responses (by someone writing under what I can only assume is a pseudonym, “Elizabeth Whately”) analyze some of the arguments I made in an earlier edition of this newsletter, so it would be worth reviewing that if you want a refresher.
A good place to start would be to note that my article was not meant as a summary of all my arguments against Intelligent Design, nor even as a summary of all the points I made in my debate with Meyer. Rather, it is a series of clarifications regarding some points which came up in discussion threads in response to the initial debate (this is made clear in the opening of the article). So anyone wanting to engage in the full discussion should definitely watch the debate as well as read that article. If I have an overall critique of the responses offered by Whately, it is that they seem to proceed as if the clarifications I offered in my article were themselves the full arguments for my position, which means they miss a number of important points I made in the discussion itself.
Furthermore, I have to admit to a certain bafflement at some of the responses, given that they address matters which I think I fully consider in the article itself, and simply seem to ignore central points I have made and arguments I have offered. Overall, I honestly think these are confused objections to my arguments, but since they have been offered in print I want to take some time to explore them. In this article I will address the first two of Whately’s replies.
First Reply: “Philosophers Battle Over the God Hypothesis”
In her first reply, Whately concedes that Meyer is not going for what she calls a “statistical knockout” approach to Intelligent Design, which seeks to show that the likelihood of life arising “by chance” is so minimal as to be effectively impossible. Rather, Meyer is offering an inference to the best explanation, arguing that the work of God is the best explanation for the existence of life (and the existence of the universe, and the fine-tuning of the universe) when compared with other proposed explanations. Good - agreement to begin with!
Soon the disagreements begin, however. Whately argues that I have misstated an element of Meyer’s argument:
“It’s not just that certain characteristics we observe in biological phenomena are also observed in phenomena we know to be always and only the result of intelligent agency. It’s that intelligent agency provides a good explanation for these characteristics of biological phenomena. And in fact, it is the only known explanation for that shared feature in the case of computer code, radio communications, etc. Again, a subtle point, but worth making: It’s not the mere presence of similarity, but rather the specific character of the similarity that calls for explanation.”
I’m not exactly sure what to make of this objection. It seems to me that Meyer’s argument is essentially that it is the presence of a specific sort of similarity between things we already know to have been created by intelligent agents, and the things we are examining which may or may not have been created by intelligent agents, which makes intelligent agency a good explanation for those characteristics. We are able to draw the conclusion that life (for instance, and in Meyer’s view) is the result of intelligent agency because it shares a specific sort of similarity with things like computer code and radio communications. So, I am not sure there is really much of an objection to my view here: perhaps we are expressing a similar thing in slightly different ways.
Finally, to end this first response, Whately argues that I make an incorrect assertion regarding the explanatory burden Meyer bears in making his case. I’ll quote the objection in full for clarity:
After this distillation, Croft makes his first key assertion: This inference is “illegitimate” unless it can be shown “that the feature is only in principle a feature of designed things.” But Meyer doesn’t need to demonstrate a claim this strong. He only needs to show that designed things are overwhelmingly likely to have this feature — that the extrapolation of currently known naturalistic causes explains it poorly, while intelligence explains it well. If this can be demonstrated, then in the Bayesian probability terms Meyer employs, the probability of a feature given design (let’s call this P(F|D)) is greater than the probability of the feature given not design (call this P(F|~D)). This means that the fraction P(F|D)/P(F|~D), which Bayesians call the “likelihood ratio,” is top-heavy.
Again, I’m not really sure what to say in response to this, because it misunderstands the force of my statement. I was saying that given that I believe (and argue) that none of his probabilistic arguments work, Meyer’s only hope is to show that the features in question are only in principle a feature of designed things. I am well aware that Meyer is making a probabilistic case - this is the very thing I established at the outset by drawing the distinction between a Meyerian inference to the best explanation and a Dembskian “statistical knockout” approach. Indeed, I say this in the post to which Whately is responding: “I hope to show, [Meyer’s is] an illegitimate inference…unless you can show that the feature is only in principle a feature of designed things – something Meyer does not do and does not claim to do. “
This demonstrates the importance of viewing my follow-up piece not as a summary of the arguments presented in the debate, but as merely a series of additional considerations raised by online commentary after the fact.
Second Reply: The Problem of Background Knowledge
This is where the objections to my arguments become more substantive - good! To catch people up, I argue in the debate that one of the things which secures an inference that some phenomenon X was the result of intelligent agency on the part of agent A is that we have a lot of background knowledge about A (or at least about the class of beings to which A belongs). So, for instance, if I return to my car to find a window smashed and the glove compartment open, we can confidently infer that a human being is the cause because we have a lot of background about human beings we can bring to bear: notably that human beings actually exist; that they have the causal powers necessary to smash car windows and open glove compartments; and that they sometimes are motivated to do so. (I call these three components of background knowledge, relevant to inferences of intelligent agency, Existence, Means, and Motive.)
By contrast, we have no background knowledge about God which can secure an inference that “God did X”: the existence of God is disputed; we have no evidence about God’s causal powers; and we don’t know much about God’s motives. So, I argue, inferences to agency by God are both unusual and weak, given the lack of background knowledge. That is my argument in a nutshell.
Whately presents a number of objections to this argument. First, she writes:
The first natural point here is that we all start out in life with the same amount of background information about all entities and minds besides our own — which is to say, none. A newborn infant has no background information about the strange creature licking his face, or the shadowy figure lifting him out of the crib. Nevertheless, through gradual data-gathering, he forms his understanding of cause and effect. I doubt Croft would claim that infants are born with innate concepts of these things. The data-gathering process must begin somewhere. Croft objects that “the phenomena to be explained (the explanandum) is supposed to be part of our evidence that God even exists.” But by the same token, the wet-washcloth sensation of puppy tongue on baby cheeks is part of the baby’s evidence that Puppy exists.
This is an intriguing objection, but not a particularly damaging one. Perhaps stated more formally, the objection seems to be that, according to Whately, all our knowledge about agents is abductive, in that we have a set of experiences which we then hypothesize are the result of the actions of an intelligent agent. In the puppy case, the feeling of being licked by a puppy leads us to infer that a puppy exists; in the car break-in case, the observation of the smashed window and the open glove compartment leads us to infer that a human being exists; in the Intelligent Design case, the observation that life exists leads us to infer that God (the life-creator) exists.
It seems to me, though, that there are major distinctions to be made between the evidence we have at our disposal in each of the three cases. In the puppy case, we have direct sensory evidence of the agent itself: we can see the puppy, feel it in our arms, and feel its tongue against our face. We are not, in the usual use of these terms at least, “abductively inferring” that the puppy exists: we are directly perceiving it.
In the car break-in case, although we did not see the perpetrator break into our car directly, we are relying on wealth of direct sensory information about human beings in general in order to draw that inference and consider it secure. We have seen human beings, spoken with them, we may have even seen YouTube videos of them rifling through glove compartments (there was one on Facebook yesterday!), and it is on that basis that we draw the inference.
In the God case, we have never (or, at least, I have never!) directly perceived God, or seen God actually doing anything. So on what basis are we inferring that God did the particular thing we want to explain by referring to God? If there is a basis, it is certainly different to the basis which usually grounds abductive inferences to agency - and that’s my entire point: that the inference is unusual. It doesn’t have the normal evidence-base we use to secure such inferences.
If Whately wants to say that all beliefs about agents external to us are ultimately abductive, and are based on inferences derived from data, then OK - we can take that view. But the problem still persists: the data available is very different in each of the cases. In the first we have direct sensory perceptions; in the second we have a wealth of background knowledge which includes direct sensory perceptions; in the third we have neither of those, and only a sort of theoretical consideration about the properties of things made by intelligences per se. The third is clearly different to the first two, and the first two are clearly different from each other, and these differences are all epistemically relevant. So this is not in any way an objection to my case.
Whately then goes on to deconstruct my example of a car break-in (she says it’s Meyer’s example, but it’s mine!), offering various anecdotes about cars being broken into, and a reference to a paper by philosopher Lydia McGrew. Try as I might, I can’t determine if there is any substantive criticism of my view in there. To restate my objection, my worry (at this point in the argument, at least) about the abductive inference to intelligent agency by God is that, unlike everyday inferences to intelligent agency by human beings, we do not have the necessary background information about God to secure the inference.
In normal, everyday instances of abductive reasoning to the actions of an intelligent agent we are saying, essentially: “One of a known class of beings, which we know from independent evidence has both the causal means and the motive to do this thing we observe happened, probably did it.” In his inference to God, by contrast, Meyer is saying something like: “I hypothesize that there exists a being, almost entirely unlike any known being, with causal means and motives I am attributing to it without independent evidence, because if such a being were to exist, and if it were to have the causal means and motives I am attributing to it, that would explain what we see.”
These two epistemic operations are different - that’s my point. And the second is not how we regularly attribute some phenomenon to the activity of an intelligent agent. In the first case we are picking from among the already-known causal agents in our world-picture one which we think is a likely cause of a given phenomenon, using independent evidence about that agent to make the selection; in the second we are constructing what we think an agent with the necessary powers to give rise to the phenomenon might be like, based on nothing more than our own imagination.
It is critical to understand that the observation of this significant distinction does not, contra Whately’s assertion in her response article, amount to a refusal to consider the existence of “any new, non-human intelligent agent.” Nor does it amount to an assumption a priori that God cannot exist. Nor, further, am I “blocking” the consideration of evidence that God might exist. Rather, it is simply an observation of a critical distinction between normal, everyday inferences to intelligent agency - the sort which lead us to infer that Mount Rushmore was created by human beings, for instance - and the inference to intelligent agency by God which Meyer wants to draw. Meyer’s inference is unusual, in that it is unsupported by the background evidence which usually grounds such inferences, and he has a responsibility to address that problem. That’s my point here.
Inferring to Unknown Agents
How could someone responsibly construct a hypothesis that a new, non-human intelligent agent might exist? Does the lack of background knowledge that I’ve pointed to on the part of God also hamper attempts to abductively infer that aliens are the cause of some phenomenon, for instance? To some degree, it does - and of course it does. Let’s look at an example:
Imagine we find on Mars a floating orb of indeterminate origin and made of an unknown material. It is a perfect sphere and it is hovering, without obvious means of support, four feet off the ground.
Would we be justified in inferring that this thing was created by aliens?
Well let’s first imagine that we have already encountered an alien species: that they initiated first contact with us, landed on earth, exited their spherical spaceships, and have all sorts of technology which looks like floating, perfect spheres. In this case, surely we would be justified in inferring that this thing is indeed created by this class of agent. Why? Because we have the background knowledge required to make a secure inference.
Now, let’s imagine we remove all that background knowledge: we encounter the sphere, but have no independent evidence that any alien intelligences exist at all. Now would you be comfortable attributing the sphere to a race of intelligent aliens who make floating, spherical technologies? I think the answer is obvious: we would be much less confident in our inference in the second case, because we don’t have the background knowledge which would usually secure such an inference. We could certainly hypothesize that 1) an alien race exists; 2) that it has the capability to create floating, spherical technologies; and 3) that it has reasons to create such technologies. And it is correct that were this hypothesis true it would explain the data. But no one, I think, would be responsible to conclude that the hypothesis is true simply because were it true it would explain what we see. In this case, we would want independent evidence that the hypothesis is actually true. (Dawes draws attention to the distinction between unconfirmed explanatory hypotheses and confirmed ones of this sort by calling the first “potential explanations” and the second “actual explanations.”)
I, for my part, would certainly not want to conclude simply from the existence of this orb that an alien intelligence exists and created it: if we couldn’t find another explanation for the phenomenon, I would prefer to call it unexplained than to commit myself to an extravagant hypothesis without any independent evidence to support it.
So what about Meyer’s God Hypothesis: is it more like the first case, where we have met and interacted with the aliens we hypothesize to be responsible for the floating orb? Or is it more like the second case, in which we are inventing the existence of the perfect sort of alien which, were they to exist, would have created the floating orb? Clearly we are in the second position, and that means Meyer’s inference is extremely weak.
We can go further though: Meyer’s inference to God is even weaker than the inference to aliens in the no-background-knowledge floating orb case. Why? Because at least with the inference-to-aliens we can use our background knowledge about human beings, evolution, and astronomy to provide some measure of justification for the inference. We could make an argument like the one which follows, for instance:
We know that intelligent life can evolve on particular kinds of planets, because we did.
We know that there are lots of potentially life-permitting planets in the universe, because we have found many of them, and we can tell from our understanding of astronomy and physics that there must be many more out there.
It is reasonable to conclude, given what we know about the evolution of life on our planet, that intelligent life may have evolved on one of these other worlds.
It is plausible, given what we know about humans and other animals, that such life would also want to create technologies (as we and other animals do), and also that they might want to contact other species
Sufficiently advanced alien technologies might be inexplicable to us, just as the technologies we have today would be inexplicable to our ancestors were they to view them. And, therefore:
It is not unreasonable to postulate that this phenomenon (the floating orb) is a technology created by an alien intelligence.
Note how each step of this argument relies on relevant background knowledge: knowledge of biological evolution; knowledge of astronomy, physics, and cosmology; knowledge about human motives and the motives of other animals; knowledge about technological development. We are bringing to bear knowledge about a broad class of intelligent agents to which the hypothesized alien race might belong in order to make our abductive case (the class of “evolved, embodied, biological agents like human beings). I am not saying I would be convinced by such an argument - without independent evidence of the existence of such an alien species I don’t think I would be - but to the extent that such an argument has any force whatsoever, that force is down to the background knowledge mobilized in its support.
Note, now, that Meyer’s God Hypothesis is nothing like this. Rather than mobilizing our knowledge of actual agents to draw a responsible abductive inference, he posits the existence of an agent which is so different to known agents that none of our background knowledge is relevant to to it. All known intelligent agents are physical, embodied, and evolved, and God is supposed to be none of these. So what evidence can we use to draw an inference that God did something?
Whately tries to respond to such concerns, at the end of this second reply, by appealing to what she calls a “magic ratio”: “the likelihood ratio P(E|H)/P(E|~H)”, meaning “the probability of the evidence given the truth of a hypothesis divided by the probability of the evidence given the falsehood of a hypothesis.” Apologists increasingly appeal to likelihood ratios and Bayesian analyses in order to argue that some body of evidence is more likely given the truth of theism than the truth of atheism, and therefore is support for the God Hypothesis. In response, skeptics frequently worry that the probability of some evidence E is impossible to determine on theism, because God’s motives are inscrutable (or at least unknown), and if we are unable to predict what God would do, even if God exists, then we cannot determine the probability of E given theism, and therefore we cannot perform the comparison with naturalism.
Whately’s response is quite odd, especially for someone with a PhD in mathematics: she says that even if “our numerator P(E|H) is inaccessible…We can still say something about [the ratio P(E|H)/P(E|~H)]”. According to Whately, “One doesn’t need a highly specific high number for the numerator to know that it’s much higher than the denominator.”
This is a very strange reply, in my mind. Certainly it’s the case that we can compare a rough estimate for P(E/T) [the probability of some evidence given theism] with a rough estimate of P(E/A) [the probability of some evidence given atheism]. We can certainly determine which of two roughly-estimated quantities are larger. But if one of the probabilities is “inaccessible” (to use Whately’s term) then we cannot compare them: if we literally have no principled way to determine one of the terms in our ratio, we cannot say anything about the ratio.
Even if we are charitable, and interpret the word “inaccessible” to mean “only vaguely determinable”, I still would worry that it is actually very difficult to determine both terms in the ratio. I don’t really know where to begin quantifying the probability that life might arise given atheism, for instance, and I am certainly not willing to conclude that the probability is low. Likewise, I do not know how we would start estimating the probability that life would arise given theism, and I am certainly not willing to concede that the probability is high. It seems to me that both terms in the ratio are at least extremely difficult to determine, even roughly - which is why in general I resist attempts to quantify these sorts of probabilities. We just don’t have enough knowledge to do so.
This is a general problem with a lot of apologetic arguments which rely on probability estimates: the estimates are almost always essentially made up. When you read or listen to apologists trying to justify the probability estimates they attribute to various pieces of evidence, they often do a terrible job. A great example can be found in the debate between Kevin Scharp and William Lane Craig, in which Scharp challenges Craig to justify his probability estimates regarding the fine-tuning of the universe. If you watch from 45:18, you’ll hear the following exchange:
WLC: The fine-tuning of the universe is more probable given theism than it is given naturalism. It seems to me that all the theist needs to do is to show that it’s not improbable that God would want to create a finely-tuned universe, and that is surely going to be much, much more probable than on naturalism: all of these constants and quantities falling by accident into the life-permitting zone.
KS: Good. I’m not going to get sucked into that. Trying to have a discussion about what God would do or what he wouldn’t do, or anything like that, that’s all divine psychology, it’s all equally murky, and unclear, and I don’t think we have any good reason to believe any of that stuff. If you want to push the fine-tuning argument, what you need to do is say that the chances that God would create the universe are actually better than the chances that it was created randomly, given its fine-tuning, and you don’t do that.
WLC: Pardon me?
KS: You don’t argue that.
WLC: Well, I…I think I do…
KS: I wanna hear that.
WLC: Robin Collins, in particular, does that in his formulations of the fine-tuning argument. What he argues is that the naturalist would have to show that there is some sort of significant improbability in the designer’s creating a finely-tuned universe. And the probability of these constants and quantities falling into the life-permitting zone is so incomprehensible that the atheist could never demonstrate that it’s less probable than that that a designer would want to create a finely-tuned world.
KS: Why not? Why assume that? Why not actually do the calculation and show me how likely it is that God would create the universe?
WLC: Well I think it’s a matter of…
KS: What’s the calculation? What’s the probability? Is in 1 in 10 to the tenth?
WLC: You can’t put numbers on this…
KS: You can’t! I agree, you can’t, and that means you can’t conclude what you want to conclude from the fine-tuning argument.
WLC: What you can say is that given an intelligent designer of the universe it’s not improbable that he would fine-tune it for the existence of life.
KS: Justify that. Why think that it’s not improbable that God would create a universe at all?
WLC: [Big pause]…We-uh, you don’t have to show that it’s not improbable, but simply that the probabilities are not as low as all of these constants and quantities falling by accident into the life-permitting zone. And that is so absurdly improbable that I think it just outstrips any sort of…ummmm…ummm…uncertainty with respect to saying that a designer of the universe would finely-tune the universe for agents.
KS: Excellent. So why isn’t it absurdly improbable that God would create the universe?
WLC: It’s not absurdly improbable because God could have good reasons for doing that.
KS: He could, but that’s not a basis for trying to calculate a probability.
The exchange continues in this vein, but it is clear that Craig doesn’t have a substantive response to the worry Scharp is raising, and it’s equivalent to the worry explored in this dialogue between Whately and me. If we have no principled basis for determining the probability that God would do a thing, we simply cannot compare the probability of a piece of evidence given the existence of God and the probability of a piece of evidence given the non-existence of God - even if the probability of the evidence given the non-existence of God is very low. Remember that the important thing here is we are comparing two probabilities, and if one of them is inaccessible, contra Whately, we cannot perform any kind of comparison at all.
And that’s it for the second response: there is really nothing here so far which troubles my challenge to Meyer’s abductive inference to agency by God. We’ll se if the next parts fare better…