Alvin Plantinga’s Evolutionary Argument Against Naturalism has attracted a great deal of serious critical discussion (e.g. Naturalism Defeated?) and has had a substantial impact on ‘popular’ appraisals of naturalism. (For example, William Lane Craig frequently uses it, and it also appears in the dismissal of naturalism in The Experience of God.) Many philosophers have pointed out various problems with the EAAN, and in my judgment the EAAN is not only flawed but fatally flawed. Nevertheless, it’s a really interesting argument and it could be worth exploring a bit. I’ll present the argument here and then we can get into it in comments if you’d like — though I won’t be offended if you’d rather spend your time doing other things!
The EAAN has gone through various iterations, but here’s the latest version, from Where the Conflict Really Lies: Science, Religion, and Naturalism (2011). Intuitively, we regard our cognitive capacities — sense-perception, introspection, memory, reasoning — as reliable, where “reliable” means “capable of giving us true beliefs most of the time” (subject to the usual caveats). Call this claim R (for ‘reliable’). But how probable is R?
Suppose that one accepts evolution (E) but also affirms naturalism, defined here as the belief that there is no God or anything like God (N). What is the probability of R, given N & E? One might think it’s quite high. But Plantinga thinks that, however high the probability of R, nevertheless the probability of R given N&E is low or inscrutable. Why’s that?
Now, here’s the key move (and in my estimation, the fatal flaw): beliefs are invisible to selection. Why? Because selection only works on behavior. If an unreliable cognitive capacity is causally linked to adaptive behavior, then the unreliable capacity will be selected for (i.e. not selected against). Even a radically unreliable capacity — that one never or almost never yields true beliefs — can be selected for. Selection only “cares” about adaptive behaviors, not about true beliefs. (More precisely, we have no reason to believe that the semantic content is not epiphenomenal.)
So, Plantinga thinks, given N&E, the probability of R is very low. But, if the probability of R is low, given N&E, then that should ‘infect’ the likelihood of all of the beliefs produced by those capacities — including N&E themselves. So, given N&E, we should it think it extremely unlikely that N&E is true. And so the initial assumption of N&E defeats itself. (Here I’m being much too quick with the argument, but we can get into the details in the comments if you’d like.)
Anyway, it’s a really cool little argument, and it’s not immediately clear what’s wrong with it — and I thought it might be worth discussing, given how influential it is.