Following a tradition of presuppositionalist apologetics, Anderson takes up the position that God stands among the necessary preconditions for knowledge itself, and that we must assume the existence of a divine author of the universe if we are to free ourselves from a paralyzing epistemological skepticism. He bases his outlined case, a kind of transcendental argument for theism, on the observation that one of same notorious difficulties of metaethics also frustrates inquiries into epistemology, namely the question of how normative standards can arise out of the impersonal properties of the universe; for the very concept of rationality presupposes epistemological norms to which we ought to conform. Similar to certain moral arguments for the existence of God, he suggests that we cannot derive the prescriptive standards we need for epistemology from a Godless description of reality.
As we might expect, then, Anderson's argument features some of the very same strengths and weaknesses as those moral arguments he emulates. For example, when addressing the Platonic approach, he rightfully observes how difficult it is for us to imagine how abstracta governing personal relationships could just happen to exist in a universe without a personal author. On the other hand, though such an argument from incredulity may have some inductive force, its controversial form leaves it potentially vulnerable to a variety of serious criticisms, chief among them the dubiousness of taking our personal experience with finding explanations for certain phenomena as evidence concerning which explanations may ever be found and which may not. So, a Platonist could respond that our present inability to conceive an impersonal accounting for epistemically prescriptive abstracta hardly constitutes evidence that no account can ever be found. Anderson in turn might point out that it does indeed serve as an inductive argument against Platonism, and that clinging to the hope of some day discovering how to explain the existence of personal normativity in an impersonal universe only ignores real difficulties through ad-hoc assumptions. In this way, a dispute over epistemic normativity may closely resemble more familiar moral points and counterpoints.
Anderson devotes the bulk of his essay to attacking what he calls (metaphysical) "naturalism," which he takes to be "the thesis that only 'natural' entities exist, i.e., entities which can be described (at least in principle) in terms of the methods and inventories of the natural sciences (e.g., physics and chemistry)." For the most part, though, he frames his criticisms of naturalism in sufficiently general terms such as they may apply to certain non-naturalistic positions nearly or even equally as well. So, we should not misunderstand him to have constructed merely a critique of one metaphysical view. In fact, his principal argument seems to turn on his position that any ontology whose foundation limits itself to descriptive terms can never inform us "how things ought to be," a required component for epistemic normativity. Since science is a "purely descriptive" enterprise, then naturalism, which is founded exclusively on scientific principles, leaves no space for epistemic normativity and thereby rational warrant. Yet, if beliefs can never have rational warrant, then knowledge itself is impossible. Organized into a deductive form, this reading of his argument consists of the following:
(1) Naturalism is a purely descriptive ontology;
(2) A purely descriptive ontology excludes epistemic normativity;
(3) Knowledge is impossible on any ontology which excludes epistemic normativity;
(4) Therefore knowledge is impossible on naturalism.
Since clearly we do have knowledge, it follows further that "naturalism must be false."
We may wonder, though, what is stopping us from applying this argument to any ontology at all, given that ontologies are always by their very nature descriptive? Put another way, since descriptions can only ever describe, then what advantage has a non-scientific description of reality over a scientific description? Perhaps, then, Anderson means to suggest that scientific descriptions in particular, i.e. not merely by virtue of being descriptive, are somehow incompatible with normative standards. For instance, one might argue that standards, and possibly also the very relationship between standards and human beings, are abstract objects, the reality of which the scientific method disallows or ignores. However, Anderson explicitly rejects this view later in his essay, where he denies that "impersonal abstracta could give rise to epistemic duties." Alternatively, he might try taking the position that standards can only influence conscious minds, but that consciousness cannot in principle have any explanation on naturalism; so he writes that, on naturalism, "mental phenomena...must either be reduced to the physical (i.e., explained in terms of more fundamental physical phenomena) or eliminated (i.e., explained away altogether)." This position would require significant defense, though, which he declines to offer in this particular essay. It seems most likely, then, that he would prefer to argue along the lines that, on naturalism, we must treat epistemic standards as a kind of human convention, and that this view carries with it certain serious, even insurmountable problems.
Now, for clarity, we needn't treat human convention as mutually exclusive with respect to biological evolution and development. In other words, it may not be biologically possible, even in principle, for a human society to hold to radically different epistemic norms than they do in actuality. It could be, for instance, that we are developmentally or even genetically predisposed to adopt certain kinds of standards. Perhaps we are wholly unable, again by virtue of our biology, to reject them. So, convention ought not be confused with whimsy. Understood in this way, that is, as rooted in biology, human convention seems an attractive characterization of epistemic standards, whether or not one is a naturalist. Anderson, however, argues that human convention, though it might do for founding, say, moral standards, can never provide a suitable foundation for epistemic standards. Using an analogy similar to that introduced a few years ago by Laurence Bonjour (In Defense of Pure Reason (1998), §7.4, p199), he invites us to consider a population which takes by convention wishful thinking to constitute a form of rational warrant, and furthermore which rejects beliefs formed from "reliable perceptual faculties" as non-rational. If epistemic norms are just human conventions, then this hypothetical community of wishful thinkers really could come about, at least in principle, and they would have no less powerful a claim to rationality than the actual population of non-wishful thinkers. "But clearly this can't be right," he protests, thereby concluding that epistemic norms must be more than simply human conventions. Indeed, he adds, human conventions are always "subject" to such norms. He appears to be arguing, then, that no description of reality can be correct which has as a consequence that normative standards develop as human conventions.
I agree that something has gone wrong in his thought experiment, but it is not the assumption that epistemic norms can develop through human convention; for Anderson has quietly taken on the additional assumption that there can be a population which rejects its perceptive faculties and functions as a community. Yet what does it mean for a community of people to reject their perception? It seems to me that any person who failed to yield to his own sensations and perceptions could never interact successfully with his environment, even to the smallest degree. In order to, say, communicate with other people, he must be willing and able to interpret his perception such that he even recognizes that there exist such other people with which to communicate! If we amend the analogy such that the hypothetical community recognizes perceptive faculties as important for acting in the world, then we are left with the suggestion that in addition to ordinary rationality, it also embraces wishful thinking. In that case, however, the problem with the community is that their standards are self-defeating insofar as they are inconsistent even when evaluated with respect to themselves---a criticism (hopefully) inapplicable to the epistemic standards we all actually use. It seems to me, then, that he needs other grounds on which to argue that epistemic standards cannot be taken as human conventions. As long as the naturalist---or the non-naturalist, for that matter---appeals to convention for the origin of this kind of normativity, Anderson shall require more than just the criticisms given in his essay to make his case for the impossibility of knowledge on such a view. In particular, he requires further, independent defense of his position that any account of epistemic standards "must posit a fundamental ontological distinction between that which grounds or originates epistemic norms and that which is subject to epistemic norms," a key component in his stated case for theism.
In sum, though many of Anderson's observations deserve (and, to the extent that others have preceded him, have received already) serious attention, his overarching argument for theism requires significant development. This does not in itself constitute a deficiency in his essay, which he clearly intends only as a "thumbnail sketch;" we should therefore refrain from complaining that it falls short of a robust defense of theism. However, Anderson's criticism of naturalism as purely descriptive seems fundamentally unable to function as a defeater for naturalism, regardless of development; furthermore, his resistance to understanding epistemic norms as human conventions seems to me entirely unsupported. Inasmuch as his case for theism depends on these two points, we cannot properly take it as persuasive.