But is the possession of good reasons necessary, under the given circumstances, for the production of the belief in question? What we have to evaluate is the following pair of counterfactual conditionals:
(a) She would have accepted the belief if she had not seen that it was supported by good reasons.
(b) She would not have accepted the belief if she had not seen that it was supported by good reasons.
I submit that, in the absence of further information, neither of these counterfactuals can be evaluated as true. Following John Pollock, we assume that a counterfactual conditional is true if and only if the consequent is true in all those worlds minimally changed from the actual world in which the antecedent is true. Would a world minimally changed from the actual world in which she doesn't see that her belief is supported by good reasons, be one in which she would not accept that belief? No doubt there are a number of different ways in which the world could be changed just enough to satisfy the antecedent of the conditional; in some of these she accepts the belief while in others she doesn't. And there is no basis for saying that those in which she doesn't accept it are less changed from the actual world than those in which she does -- or vice versa. We conclude, then, that (a) and (b) are both false; what is true is
(c) If she had not seen that the belief was supported by good reasons, she might have accepted the belief, but it's also the case that she might not have accepted it.
Consider, on the other hand:
(d) She would have accepted the belief if the antecedent conditions were not sufficient (as determined by the laws of physics) for her accepting it.
(e) She would not have accepted the belief if the antecedent conditions were not sufficient (as determined by the laws of physics) for her accepting it.
Here we can state unambiguously that (d) is false and (e) is true; the fact that one, and not the other, is in accord with the physical laws means there is no question that worlds in which she does not accept the belief are closer to the actual world than the ones in which she does. All of this merely restates, in the language of counterfactual conditions, what should by now be obvious: In a physicalistic world, principles of sound reasoning have no relevance to determining what actually happens.
Because the matter is so crucial, I am going to risk excess by restating the point once more, this time in terms of possible worlds. In order to identify the possible worlds we want to consider, note again the final clause in the definition of strong supervenience: "necessarily, if any y has [physical property] G, it has [mental property] F." If "necessarily" here is understood as physical necessity, identifying the relevant world is easy: consider a possible world that is physically exactly similar to the present world, but in which the natural laws establishing psychophysical connections do not obtain. In such a world all the physical facts, and with them the entire physical course of events, are exactly as in the actual world: the complete absence of mentality makes no difference whatever. Similarly, we may consider a possible world physically identical with the actual world, but in which mental properties are redistributed in as bizarre a fashion as one might wish: this world is still indistinguishable from our own in all physical respects. Could there be a more dramatic demonstration of the fact that, given the closure of the physical, mental facts are irrelevant to the physical course of events?
Suppose, however, "necessarily" in the definition of supervenience is understood as metaphysical necessity. This embodies the idea (for which I've expressed some sympathy) that the natural laws that obtain are expressions of the essential causal powers of the kinds of objects that exist, so that in no possible world do those very same objects exist governed by different natural laws. This means we can't simply cancel the psychophysical connections while leaving the rest of the actual world unchanged. Instead, we proceed as follows: take a world consisting of objects exactly similar to the objects of our world, except with regard to the psychophysical connections that obtain. For reasons that should be evident, we will designate this as the physically equivalent zombie-world to our own world. In the zombie-world matter will not consist of protons, neutrons, electrons, etc., but rather of zombie-protons, zombie-neutrons, zombie electrons ... The zombie-electrons will not have the properties of mass, charge, and spin but rather of zombie-mass, zombie-charge, and zombie-spin. Such a world will be similar to the "mindless world" described in the previous paragraph in every respect but one: it will not contain the identical objects, organisms, etc. that exist in our world (and in the mindless world) because those objects consist of ordinary matter and not of zombie-matter. But the zombie-world is physically equivalent to both the mindless world and the actual world; all three worlds are identical in all physically observable respects. Once again, we have a dramatic demonstration of the fact that neither in the zombie-world, nor in the mindless world, nor in the actual world given the assumption of the causal closure of the physical, do principles of rational inference play any role whatever in determining what happens. And in the actual world (which is not mindless), the principles of inference play no role, given causal closure and supervenience, in determining what beliefs people come to accept.
One way in which a physicalist might respond here is by questioning the assumption that good reasons and principles of rationality need to be thought of as causally relevant to what happens in the world. Wittgensteinians often adopted the stance that reasons-explanations and causal explanations belong to different language-games and so do not conflict with each other. And Kim recommends as "well worth exploring" the idea that rationalizing explanation is "a fundamentally noncausal mode of understanding actions," so that "a rationalizing explanation is to be viewed as a normative assessment of an action in the context of the agent's relevant intentional states."
Whatever its merits in general, Kim's suggestion is singularly unpromising in its application to the relation between reasons and beliefs. To see this, the reader is asked to reflect on the way she goes about assessing an argument of moderate complexity. I presume she begins by reflecting on the premises of the argument -- are they propositions she believes, or at least considers reasonably plausible? She then considers carefully the logical connections that are alleged to obtain between the premises and the conclusion -- do the premises indeed provide support for the conclusion, and if so does the support amount to deductive validity, or is there some lesser degree of support? Are there ambiguities in the argument which might undermine the soundness of the inference? Sometimes these questions are assessed in the light of specific, explicitly formulated principles of logic and argument; at other times she relies on a more intuitive grasp of the particular argument at hand. If the assessment is favorable, she accepts the conclusion, either tentatively or with considerable firmness, depending on the particulars of the case. If she is skillful in carrying out such assessments, she is said to possess "good logical insight," an intellectual virtue which is prized, in part, for the specific reason that it enables one to reach good, well-justified conclusions about the arguments one encounters. The entire process makes no sense at all, except on the assumption that a person's awareness of reasons and her knowledge and application of principles of rationality make a difference to the conclusions that are accepted.
Kim's suggestion, as applied to the relation between reasons and beliefs, is not only implausible; it is also futile. For surely those who would argue that principles of rationality serve the purpose of a "normative assessment" of our reasoning would allow that these principles can in fact be used in making such an assessment. But of course, such a normative assessment of a piece of reasoning is itself also an example of the kind of reasoning that is being assessed. (Note that the example of reasoning described above involved precisely the examination of an already formulated argument.) Are good reasons, and the principles of sound reasoning, allowed to be causally effective in determining the outcome of the assessment process? Or is some other account to be given of how the process goes? In any case, whatever answer is give here could equally well have been given in the first place; the move to the level of "normative assessment" changes nothing.
The Emergent Self