Plantinga's critique of naturalism

PZ Myers and I certainly agree that Alvin Plantinga gives philosophy a bad name. Unfortunately, like lawyers, 99% of philosophers give the other 1% a bad name, and Plantinga doesn't really stand out in the crowd.

Myers, like Stephen Law, misses the crux of Plantinga's argument: On Plantinga's account the truth or falsity of a belief has nothing to do with its adaptive functionality. Truth and falsity are essentially uncorrelated with adaptation, selection, learning or any other "reality-based" cognition.

Plantinga cannot take the sole blame for this view of truth and falsity: he relies on an underdetermination argument that more staid philosophers, including the venerable Quine, have advanced. Plantinga's argument essentially follows Carnap's critique of Popper's probabilism, with Carnap's conformance to experiment mapping directly to Plantinga's adaptation.

Essentially, Plantinga argues that while the truth might be necessarily adaptive, many false beliefs and belief-producing mechanism can also be adaptive. (The arguments for beliefs can be applied directly to belief-formation mechanisms and vice-versa without loss of generality.) There are infinitely many beliefs that might be adaptive, but ex hypothesi exactly one is actually true. The probability is infinitesimal that any given belief could be actually true even if it were adaptive. Plantinga is being charitable, perhaps unreasonably so, by giving any given belief a numerical, non-infinitesimal probability, much less an enormously large numerical probability of 0.5. Furthermore, under Plantinga's critique, the probability of any belief actually being true is indeed independent of other beliefs being true.

If there really are infinitely many false beliefs that will produce adaptive behavior, we have said nothing specific about the probability of some specific false belief being adaptive; we could say empirically that there is a finite measure relating the infinite number of false adaptive beliefs to the infinite number of all false beliefs. Even a very small finite measure permits evolution of a large number of adaptive beliefs over time.

It is not enough, as Myers and Law note, to say that beliefs under evolution are dependent: They are only adaptively dependent. If false beliefs could be adaptive, adaptive dependency does not entail truth dependency.

Plantinga goes wrong (as does Law in his critique) in assuming on some unspecified intuitive basis that our beliefs really are truth-reliable. We simply don't know that: we know only that our beliefs are adaptive-reliable, an unsurprising conclusion from evolution. We cannot, therefore, conclude that a god exists to make our beliefs truth-reliable. Likewise Law's critique fails on the same grounds: we don't know that conventional logic is actually true or truth-preserving; we know only that logic is adaptive.

Plantinga — and indeed philosophers in general — also go wrong in assuming a priori that there is exactly one and only one statement set that can be true, and all deviations from that statement set are equally and absolutely false. But why should we take that view? Carnap's critique of probabilism and Plantinga's critique of evolution must still assume that we can make some distinction between "good" and "bad" beliefs: i.e. those do or do not correspond to experiment or are or are not adaptive. Furthermore, the distinction between true and false beliefs cannot be made on this basis; we cannot make the distinction between true and false beliefs at all! Even if we were to stumble by the purest chance on the one true belief among all the infinite and equally explanatory or adaptive false beliefs, we would have no way of knowing that we had stumbled upon the truth.

An alternative view is simply to define truth and falsity in terms of the distinction we can make: explanatory accuracy and adaptation: A true belief is an explanatorily accurate or adaptive belief. Truth thus becomes at least a scalar property: a belief is more or less true based on how much and how precisely it predicts experimental results, or on how widely and how effective it is at producing adaptive behavior.