Google+
Close

The Corner

The one and only.

Does Accepting Darwin Require Atheism?



Text  



Lots of scientists say yes (or more precisely, that evolution forces you either to retreat to pure mysticism or a conception of God that is so abstract as to be functional atheism), but I’m pretty sure the answer is no.

While guest-blogging for Andrew Sullivan last week, I responded to a book review by eminent biologist and prominent New Atheist Jerry Coyne in TNR in which he made this claim by showing in detail that genetic evolution can in fact be teleological. This stirred up a lot of blog commentary, including a fairly intemperate response by Professor Coyne. I replied to him at The Daily Dish, but wanted to lay out a specific case that responds to my more thoughtful critics. Here goes.

First, I am making a fact claim. My fact claim is this: The findings of the modern synthesis of evolutionary biology (MES) do not demonstrate that the universe is not unfolding according to a divine plan that privileges human beings. An informal specification of what I mean in my claim by “does not demonstrate” is not restricted to something like “does not demonstrate it because it’s possible that everything we believe we observe through sense data is an illusion” or things of that ilk, but instead is closer to the sense of “does not make it obviously unreasonable to believe it.”

Some claims I am not making include:

• There is compelling evidence of a divine plan that privileges human beings made available to us by the structure or result of evolution through natural selection (for convenience, “evolution”).

• There are no other findings or observations external to the MES that render untenable a belief in a divine plan that privileges humans.

• Darwin’s work has not undermined many beliefs that many people have held about their religious views and their place in the universe.

• Quantum mechanics (QM) demonstrates that there is irreducible uncertainty (i.e., “true randomness”) in the physical universe.

Note that I am neither asserting nor disputing the truth any of these claims in the context of this argument, merely stating that neither my fact claim nor the reasoning by which I produce it depends on any of these statements.

Let’s start with a simple observation. It seems kind of crazy (or at least to be a violation of common sense) for an omnipotent deity who desires some outcome to get there through a process that unfolds over time. If you’re all-powerful, and you want the universe to look like X, why would you have eons of change to get there? Why wouldn’t just create things the way you wanted them on the first day? In fact, any observable change of any kind creates this problem. In the case where a deity initiates or sustains a process of change for the purpose of achieving a desired end-state, this seems illogical: We should be in a world of complete stasis. I’ll call this “the argument from change.”

An omnipotent deity would therefore have to desire (again, assuming our rationality can comprehend an omnipotent being) that all changes that actually take place happen for reasons other than their causal effectiveness in creating some desired end-state. But many changes that we observe around us seem to be what we intuitively believe to be evil — the classic case is a tortured and suffering child. How could a God that comports with our idea of benevolence actively desire every evil act in the universe? As everybody knows, this is the problem of evil.

Note that both of these objections to a divine plan are independent of evolution. Both are objections that could be (and were) raised long before Charles Darwin was born. So, for us to say that the MES rules out a divine plan that privileges humans, we must assert that there is some incremental knowledge provided by the MES that rules out such a divine plan that was not available to us prior to Darwin. The argument that “it just seems crazy that your Flying Spaghetti Monster would act through such a haphazard process” may or may not be valid objection to a belief in a divine plan, but it doesn’t require knowledge of evolution; it’s just a special case of the argument from change. Further, the point that it seems hard to reconcile an omnipotent and benevolent deity that chooses to act through a plan that incorporates what seems to a normal person to be evil also doesn’t require knowledge of evolution; we are all aware of many seemingly evil actions — that must be willed by any purported omnipotent deity — independent of any knowledge of evolution.

So what then is the unique contribution of the MES in this regard? I was confronting in my original post the frequently made argument that the structure of the process of evolution is inherently undirected, or goalless, and therefore it can not be that this is a mechanism used to create any specific outcome such as humans, and therefore evolution makes belief in such a divine plan unreasonable. I did this by walking methodically through an example of a genetic algorithm (GA), and showing that the fundamental evolutionary operators of selection, crossover and mutation can comprise a goal-directed process.

The arguments that have been proposed by my various interlocutors for why this does not tell us much about evolution in nature are all variants of the idea that there are differences between a GA and evolution in nature that are relevant for the philosophical argument, and that these differences mean that evolution in nature is actually goalless. I’ll subdivide these arguments into two sub-types: (1) evolution in nature is more complex than the GA example, and (2) the pseudorandom steps in the GA correspond to “truly random” events in nature.

I’ll take the first sub-type first. All of these were addressed, at least at a high level of abstraction, in the original post, but I’ll try to go into a little detail on each. One of these arguments is that the GA will tend to home in on a “local optimum” in the solution space and not make the kind of huge changes that we see over time in evolution. But this is demonstrably false. Consider the simple case of the mutation rate, which was set in my example to 1 mutation per 10,000 genes per generation. In a limit case, if we set the mutation rate to 10,000 mutations per 10,000 genes per generation, we would, in effect, just be picking genomes arbitrarily out of a hat, and would by definition never be trapped in a local optimum. More realistically, by varying mutation rate (and actually more importantly, crossover probabilities) a specific instantiation of a GA can be made more or less “greedy” — trading speed of convergence on a solution off against the probability of premature closure on a local optimum. There is a large body of research on this. Another of these arguments is that the fitness landscape in nature is not some pre-defined factory production rate, but a constantly moving target created by the interaction of all parts of nature. But what determines fitness for any particular genome at any particular moment in the GA is a module of code that runs a simulation of the factory to estimate output for various genomes. This module of code could be anything that establishes fitness. It can be arbitrarily complex, can be in part or in total determined by reference to the existing genomes, and can create a different fitness landscape in every generation. It still remains a deterministic set of code. Another of these arguments is the idea that there are multiple parallel lineages in natural evolution. But of course one key virtue of the structure of a GA is that, mostly because of crossover, it will tend to create exactly this effect, termed “implicit parallelism.” As you watch a GA step through generations, you can see it developing clusters of genomes of particular types that represent, roughly, the examination of multiple high-fitness regions of the search space in parallel. I’ve yet to see a convincing example of a difference between my GA example and natural evolution that invalidates my argument.

The second sub-type of argument is much more interesting. The basis of it is the belief that if we accept QM, we must believe that many events in the physical universe are non-deterministic, i.e., “truly random.” That is, that there is no amount of information about physical laws plus initial conditions that can ever allow us to predict them reliably. Therefore mutation rates and crossover probabilities, as well as various other physical processes that will determine the environment in which evolution will operate, are unknowable in advance, and therefore the path that evolution will take is truly unknowable.

One big complication to this argument is that, despite repeated assertions in public debates, there are alternative interpretations of QM. What is normally termed the “causal interpretation” is the side of the argument that disputes the inherently non-deterministic nature of QM: in simple terms, the idea that there are so-called hidden variables that we just don’t yet know. So we need to consider two cases: deterministic and non-deterministic QM.

The case of deterministic QM is simple. As per the prior discussion, if the so-called “random events” in the evolutionary process are really pseudorandom, then while their complexity means that scientists appropriately proceed as if they are truly random, a being with sufficient knowledge of physical laws and initial conditions could predict them reliably, and therefore evolution can be consistent with a divine plan. If you dispute this by saying that this is what we mean in normal speech by “random,” then the consistent application that principle would apply to all kinds of events, such as, for example coin flipping. So without any knowledge of evolution we would have to say that all kinds of physical processes have random elements, and therefore no unfolding of the universe could be completely predictable. But of course this means that the same argument against divine planning would apply without evolution, and therefore the MES is not what would make belief in a divine plan unreasonable.

The case of non-deterministic QM follows a parallel line of logic, with a twist. Imagine having our current science, absent any good theory of evolution or genetics. From the time of the Big Bang through today there would have to have been an enormous number of events that were “truly random.” How can there be any kind of plan, never mind and kind of divine plan that privileges humans, unfolding in this case? How could something like a point-mass at the time of the Big Bang unfold into some pre-determined or desired structure that we have now, billions of years later. Note that we’ve said nothing about evolution here. That is, one could make the argument that QM makes belief in a divine plan untenable, but the MES does not add any incremental uncertainty. This is a specific point that I was very careful to make in the original post. If you think that there could be some kind inherent structure built into physical laws that sat hierarchically above the “truly random” events and caused them to cancel one another out around some underlying direction of in principle predictable change . . . well, since we’ve established that the process of evolution doesn’t impart additional in principle unpredictability, then what is evolution but precisely an example of this?

So in sum, there are many arguments that point out contradictions between our view of logic plus observed reality on one hand, and any conception of a divine plan that privileges humans designed by a God that just about anybody would want to worship on the other. The argument from change and the problem of evil pre-date the MES, and would continue to exist if the theory of evolution is someday overturned by new scientific discoveries. The problem of how any kind of divine plan could operate in a non-deterministic universe also exists independently of any unique characteristics of the MES. We can’t credit the MES with any of these arguments. The processes that define evolution don’t provide any relevant incremental knowledge that makes such a belief in a divine plan less logically warranted.

I’ll draw a bright line between everything up to this point in the post, which is a clarification of a fact claim that I am prepared to defend, and what follows, which is more speculative.

What I find interesting is that if one were to ask how could a divine plan ever be consistent with non-deterministic QM, the obvious way you could do this would be to have some kind inherent structure built into physical laws that sounds a lot like evolution. If one were to posit non-deterministic QM, then, the irony is that all else equal, the incremental effect on our beliefs of the discovery of the MES should actually be to make the idea of a divine plan more plausible rather than to make it less plausible. Note that I am not arguing that this is compelling evidence of design, or that we can read this into the evolutionary algorithm, or that this observation is in any way relevant to the practical execution of science, or that there is anything in this specific observation that has anything to do with privileging humans.

I do think that it is quite fair to say that this progress of science has demonstrated that numerous religious beliefs and just-so stories held to be true by many people are wrong. That is, in my view, the element of truth to Coyne’s perspective. Again, ironically (and again speculatively) science has pushed careful consideration of the purported nature of God to describe something a lot like what Augustine and Aquinas put forward. They described God as acting through laws or processes. In about the year 400, Augustine described a view of Creation in which “seeds of potentiality” were established by God, which then unfolded through time in an incomprehensibly complicated set of processes. In the 13th century, Aquinas — working with the thought of Aristotle and Augustine — identified God with ultimate causes, while accepting naturalistic interpretations of secondary causes. Neither Augustine nor Aquinas was a proto-Darwinist: Augustine, for example, thought species were immutable. What is striking about both of them, however, is their insistence on understanding and incorporating the best available non-theological thinking into our religious views. It is hard for me to imagine either feeling theologically threatened by the MES. The Catholic Church has formally accepted it as consistent with that faith.

Jerry Coyne has argued that the MES makes “real religion” as actually practiced untenable. But if so, he’s reading, for example, Augustine, Aquinas, and Pope Benedict out of the Christian religion. That’s a pretty idiosyncratic definition of Christianity. I guess Jerry Coyne wants religion to be, literally, more Catholic than the pope. But a theologian might be tempted to say that this progress of science hasn’t threatened religious belief, but rather helped to perfect it.



Text