20th WCP: What Anti-Individualists Cannot Know A Priori

Compatibilism, or the attempt to hold both Burgean anti-individualism and common intuitions about privileged self-knowledge, appears to entail that a person (say, Oscar) could come to know a priori the premises of arguments of this sort:

B 1. If I am now thinking that water is wet, then water exists.

2. I am now thinking that water is wet.

Therefore,

3. Water exists.

Naturally, from such premises Oscar could come to know a priori that water exists (i.e., by simple deduction) — yet nobody can know that a priori! Some take this to be a reductio of compatibilism (call them "incompatibilists"). But I shall suggest how anti-individualists could retain common intuitions about self-knowledge, without fearing that their views would entail unacceptable claims about the epistemic status of empirical propositions. First I shall ask why incompatibilists hold that, given both anti-individualism and privileged self-knowledge, each premise of inferences such as (B) are knowable a priori. I shall then argue that there is no plausible construal of "a priori" that could be used to generate the alleged reductio.

First, consider (B)'s second premise: could Oscar come to know a priori the proposition that he is thinking that water is wet, according to the criteria for "a priori" knowledge in the main philosophical traditions that invoke knowledge of this sort? He would certainly have privileged access to this proposition, since it involves a belief about the contents of his own mind. (Call the process of forming such beliefs "introspection.") But could knowledge of this sort count as a priori, for instance, in the rationalist tradition, where the criterion for a priori knowledge is being acquired by reason alone? Certainly not. And similarly for Kantians — who take "a priori" knowledge to be independent of all experience, and most propositions about the contents of one's own mind to be empirical propositions of "inner sense."

On the other hand, the empiricist tradition is quite restrictive in their criteria of apriority: those who do not follow Mill in rejecting this type of knowledge altogether, tend to assimilate it to knowledge of necessary truths. The logical positivist, for instance, takes a priori propositions to be those that cannot be denied without contradiction — viz., the propositions of logic and pure mathematics, and those that are true by virtue of the concepts alone. Accordingly, for the logical positivist a priori knowledge is knowledge of analytic propositions; and so, knowledge of any proposition about the contents of one's own mind (such as that expressed by (B)'s second premise) must count as a posteriori.

In none of these traditions, then, does knowledge of the contents of one's own mind counts as a priori. Therefore, the claim that Oscar can know a priori that he is thinking that water is wet would be clearly false if such knowledge were construed in accordance with any of the traditional criteria. I suggest, then, that a charitable view of the incompatibilists' claim is that they are using "a priori" with some stipulative meaning. But which one?

III

One of the criteria of apriority suggested by incompatibilists is

a priori₁ =_df the property of being knowable just by thinking.

And I take it that those who employ the metaphor "armchair knowledge" also have in mind a property of this sort. But what is the cash value of such a criterion? One may be tempted to answer that

(1) Knowable just by thinking = knowable on the basis of reason alone.

If this is what the incompatibilists' a priori is intended to mean, then any position that entails that someone could know in this way that water exists is unacceptable. Assume that the Burgean is in such a position and thus, that he uses "a priori" to mean "knowable just by thinking" — cashing this out as in (1). Then, if it could be shown that the Burgean is also committed to the apriority of the premises of Oscar's argument, this would entail the absurd conclusion that Oscar could know on the basis of reason alone that water exists (for if he could come to know a priori₁ the premises of his argument, he could infer the conclusion that water exists, and that process would, of course, also count as a priori₁).

But I have argued that the criterion of "a priori" as "knowledge attainable by reason alone," commonly held by rationalists, is not available to incompatibilists, since it would make premise (2) of their argument blatantly false: no one (including the Burgean) is committed to saying that one can know the contents of one's own mental states on the basis of reason alone. And, given that this premise would not meet the criterion of a priori knowledge thus construed, nothing would follow about the epistemic status of the conclusion that water exists. Therefore, the attempted reductio would fail.

In short, if incompatibilists could take "just by thinking" to refer to the rationalist criterion of a priori knowledge, then, assuming that the Burgean were really committed to the conclusion that one could know a priori (in this sense) that water exists, his view would turn out to be absurd. Unfortunately, incompatibilists cannot take "a priori" to refer to the rationalist criterion, and the maneuver of tacitly switching to it (in order to get the absurd conclusion required by their reductio) is not available to them. For such a maneuver would make their argument a fallacy of equivocation (though, of course, the switch may occur inadvertently, given the incompatibilist persistence in using metaphorical expressions such as "knowable just by thinking," "knowable from the armchair" and the like).

But again, charity requires us not to construe the incompatibilists' notion of apriority as implicitly appealing to any of the traditional criteria for this sort of knowledge. Yet if incompatibilists must reject the rationalist criterion as a way of spelling out a priori₁, we must then ask again, what is the cash value of their stipulation? Perhaps we could understand it as follows:

(2) Knowable just by thinking = knowable just by inference.

This candidate, however, fails to generate a reductio against the Burgean, for it is vulnerable to two related objections. First, there is nothing absurd about coming to know that water exists by (deductive and nondeductive) inference, unless, of course, some further assumptions are made — e.g., that all the premises of the inference to the conclusion that water exists are available a priori, or nonempirically.

The incompatibilist, then, must argue that all the premises of the inference are a priori in this sense. And how is he going to do that? This leads us to another objection to those incompatibilists who may wish to cash out a priori₁ as the property of being knowable by inference: viz., that such an epistemic property would fail to apply equally to all the premises of Oscar's argument. Clearly, the proposition that "I am now thinking that water is wet" cannot count as a priori in this sense, for one of the common intuitions about self-knowledge (that the Burgean does hold) is that self-knowledge is non-inferential!

But it may be that none of these interpretations captures what incompatibilists have in mind by a priori. They could insist that the problem they are raising is that of someone being able to know that water exists from premises known entirely by introspection or by philosophical argument — both forms of knowledge that (allegedly) do not require empirical investigation.

Perhaps there is another criterion of apriority that better captures

what incompatibilists have in mind:

a priori₂ =_df the property of being knowable independently of empirical investigation.

This looks like the Kantian criterion of apriority, but it isn't: for, as I have argued, the Kantian takes "a priori knowledge" to be knowledge independent of all experience, yet holds most self-knowledge to be empirical. Naturally, incompatibilists cannot concede this, since it would entail that Oscar cannot know a priori that he is thinking that water is wet. In order to generate their reductio, incompatibilists must construe "experience" very narrowly, in a way that excludes introspection. For instance, if we assume that "experience" stipulatively refers to observation and the testimony of others, then it may be that some philosophical theories (e.g., anti-individualism) and propositions about one's own thought-contents could be knowable without "experience" — that is, that they would be non-empirical or a priori in this stipulative sense. Yet such a construal of "experience" must be supported. For it is being used to raise an epistemic problem for a position that allegedly entails that one could know without experience propositions that can only be known empirically. Moreover, whether thought-contents, and anti-individualist entailments from thought-contents to the world are knowable without experience is a controversial question. We may then ask, what reasons are there to hold that, given both anti-individualism and privileged self-knowledge, each of the premises of Oscar's argument are available to him a priori₂?

Incompatibilists might answer that the first premise is available to him if he knows anti-individualism — which, arguably, he could come to know by thought experiments (philosophical arguments, etc.). Call knowledge so obtained "a priori_te." And the second premise is available to him because he has privileged access to the contents of his thoughts. Call knowledge so obtained "a priori_int." But then in each of these premises "a priori" appears to refers to a different epistemic property, and what at first seemed to be an instance of an obviously valid modus ponens begins look like an equivocation.

Yet incompatibilists may insist that by "a priori" they mean the general property of being knowable without empirical investigation — which equally applies to each premise of the argument and thus avoids equivocation. But why hold that such an epistemic property could be predicated of each of the premises of Oscar's argument? Again, premise (1) would be a priori because a person could come to know it if he knows anti-individualism — and he could know the latter just by running standard twin-earth thought experiments. And premise (2) would be a priori because a person could know the contents of his thoughts directly: they are accessible to him by introspection.

If this is why a priori₂ could be predicated of Oscar's premises, then that term would refer to the property of being knowable either by thought experiments or by introspection — a general disjunctive property, where each disjunct is a type of knowledge attainable without an empirical investigation, and where each premise comes out true because a different disjunct obtains.

But, thus construed, the incompatibilists' argument is again invalid, since there is no plausible closure principle allowing transmission of epistemic status from the premises to the conclusion. For consider an analoguous case. Suppose that (like the incompatibilist) I wish to predicate some disjunctive epistemic property of each premise in a certain argument (and each premise of that argument would be true because a different disjunct obtains). I stipulatively define this property as follows:

a priori+ =_df the property of being knowable either conceptually or directly.

I then claim to know a priori+ the premises of the following argument:

C 1. If I am seeing only a tree, then I am not seeing a horse.

2. I am seeing only a tree.

Therefore

3. I am not seeing a horse.

The first premise of my argument turns out to be a priori+ because one of the disjuncts of my stipulated property obtains — i.e., premise (1) is available to me conceptually. Furthermore, being a direct realist, I can claim to know a priori+ the other premise as well — that is, premise (2) is available to me directly. (And this premise comes out true because the other disjunct of my stipulated property obtains.) Naturally, I cannot claim to know a priori+ that I am not seeing a horse, since this conclusion is not available to me either conceptually or directly, but inferentially. Thus closure fails here. But, in exactly the same way, it fails in Oscar's argument, when the property of being knowable without empirical investigation is cashed out as the disjunctive property of being knowable either by thought experiments or by introspection.

To avoid a failure of closure, incompatibilists might now attempt to construe "a priori" as the property of being knowable either by thought experiments, introspection, or inference. Yet this construal is unavailable to them, and it would generate no problem for anti-individualism. For note, first, what is entailed by the incompatibilists' claim that Oscar could come to know anti-individualism by running standard thought experiments: If this were so, then Oscar could know anti-individualism only empirically, since to run his experiments some empirical beliefs, as well as nondeductive inference, must be available to him. He must have at least some empirical beliefs, because he needs to specify the relevant background conditions against which he is to test whether content supervenes locally. And, naturally, providing adequate descriptions of the relevant states of affairs in the actual and possible worlds requires empirical beliefs concerning molecule-per-molecule replicas, planets, natural kinds, etc.

On the other hand, suppose that Oscar has adequately set out a twin-earth case. Would he then be in a position to conclude, by straightforward deduction, that content does not supervene locally? Surely not, since he would then be confronted by contradictory intuitions, neither of which is entailed by the data the imagined state of affairs makes available. For from those data, Oscar could conclude, like the individualist, that when his twin sincerely utters "Water is wet" he expresses the same thought Oscar would (that is, that content supervenes locally) or, like the anti-individualist, that he does not.

How, then, does the anti-individualist use the thought experiments to reach his conclusion? He sets out a case, reflects upon actual (observed) ascriptions of meaning and content, compares competing explanations of these ordinary practices, and finally suggests that the hypothesis that best explains the imagined state of affairs is that content does not supervene locally. But then if Oscar is to learn anti-individualism by running twin-earth cases, nondeductive inference must be available to him.

Once we acknowledge that standard anti-individualist thought experiments require both background empirical beliefs and inference, then what incompatibilists imagine to be a reductio is in fact only that a person could come to know empirical propositions by deductive and nondeductive inference from the contents of his mind: his perceptual, sensory and doxastic states, to which he has privileged access.

Yet this seems very plausible and consistent with many empiricist attempts to explain knowledge of empirical propositions. Don't sense-datum theorists (both indirect realists and phenomenalists) explain it by invoking knowledge of our mental states and inference? And aren't those theorists empiricists? Note that, if we cash out the incompatibilists' "a priori" as the property of being knowable either by thought experiments, introspection, or inference, then sense-datum theories would have to be construed as claiming that we have "a priori" knowledge of empirical propositions. But now, surely, something has gone wrong with the stipulation!

It appears, then, that the anti-individualist is in the clear: for of the various possible ways of construing "a priori" in the incompatibilist argument, it is now plain that each generates a problem for the proposed reductio.