1. BEST EXPLANATIONISM

Before criticizing Conee and Feldman's proposal, it is important to get clear on what their thesis is. Conee and Feldman are well-known as defenders of evidentialism, which says a doxastic attitude is justified if and only if that attitude fits the evidence possessed by the subject.Footnote ¹ In order to evaluate evidentialism, we need to understand what it means for a proposition to “fit” someone's evidence; that is, we need an analysis of evidential support.

Conee and Feldman have recently attempted to provide such an analysis. In their 2008 paper “Evidence” they argue that evidential fit is a matter of explanatory coherence with a body of evidence. The rough idea is that one's evidence supports a proposition if and only if that proposition is part of the best available explanation of the evidence.Footnote ² That notion of evidential fit, plus evidentialism, suggests this thesis:

Presumably, for an explanation or a consequence to be available is for it to be actually or potentially within the subject's awareness. That understanding of awareness allows for the possibility of propositions being supported by a subject's evidence when the subject has not thought of the propositions. It also prevents the support relation from extending beyond what the subject can understand. Consider complex logically necessary propositions. They are entailed by every body of evidence, but many of them are too complicated for many to understand. So, in order to be justified in believing any proposition on the basis of a body of evidence, the subject must at least be capable of becoming aware of the entailment between the body of evidence and the proposition which the evidence justifies.

Three more notes about the wording of BE: first, BE is a principle of propositional justification, not doxastic justification. BE says how a proposition is supported by evidence. Whether the subject has a justified belief on the basis of some evidence requires the satisfaction of further conditions, such as a basing condition. Second, it may be that a proposition is well-supported by some of a subject's evidence, but the subject is also in possession of other evidence which defeats the supporting evidence. So the subject's whole body of evidence does not justify the proposition. BE allows us to make that distinction.Footnote ⁴ Third, a proposition may be supported to some degree by a whole body of evidence, but not enough to justify believing the proposition. This happens in cases of underdetermination. Suppose that my entire body of evidence e supports p more than it supports competing available explanations q, r, s, or t. So p is the best available explanation of e. But p is not flat-out justified because the disjunction of q or r or s or t is supported even better than p.Footnote ⁵ So, that a proposition is part of the best available explanation does not entail that the proposition is justified full-stop.

Now that there is a clear proposal on the table, let us evaluate it.

2. BELIEFS ABOUT THE FUTURE

T. Ryan Byerly has recently proposed a counterexample to BE (Reference Byerly2013). Kevin McCain (Reference McCain2014a) has responded to Byerly, as well as developed explanationism about evidential support in more detail (Reference McCain2013; Reference McCain2014b). Byerly and Martin (forthcoming) reply to McCain (Reference McCain2014a). The main issues in the debate appear in Byerly (Reference Byerly2013) and McCain (Reference McCain2014a), so I will focus on those papers. The Byerly–McCain exchange concerns the prospects of BE justification of beliefs about the future. BE claims that a necessary condition of any proposition being justified is its being a part of the best explanation of the subject's evidence or a logical entailment of that explanation. Byerly thinks some intuitively justified beliefs about the future do not satisfy BE (Reference Byerly2013). He describes a scenario in which he has been golfing well and in good conditions for an entire day. He starts a putt that looks like it will go in. Byerly claims that the proposition the ball will go into the cup is justified for him.Footnote ⁶ But its justification does not depend on it being a part of the best explanation of his current evidence, he argues, because “the ball's rolling into the cup at some later time doesn't explain why right now [he has] the evidence that [he has]” (Reference Byerly2013: 235). He argues that the causes of his current evidence are past and present facts, not future ones; therefore past and present facts explain his evidence, not future ones.Footnote ⁷ So, the best explanation of his evidence, and therefore what he has justification for believing, consists entirely of propositions about the conditions under which he is putting. No claims about the future enter into this explanation. Therefore, he concludes, BE delivers the intuitively implausible verdict that he does not have justification for the ball will go into the cup.

Kevin McCain attempts to defend BE from Byerly's argument by offering a reconstruction of Byerly's justification which cites only presently available evidence (Reference McCain2014a). That evidence includes these claims:

According to McCain, the best explanation of (E1) and (E2), respectively, includes:

These form the premises of a justificatory argument for the conclusion:

McCain's reconstruction, if sound, successfully avoids Byerly's objection that beliefs about the future justified by BE illicitly require future causes of present evidence. But is McCain's reply successful?

I think there are two problems with McCain's reply. The first problem concerns how subjects could possess justification for (BE1). The second concerns the nature of the difference between BE and another view about evidential support Conee and Feldman reject. I will present both problems before considering solutions.

First. Clearly, if the subject is justified in believing (BE1) and (BE2), from them competently deduces (BE3), and lacks defeating evidence, then the subject possesses justification for (BE3). The premises relevantly entail the conclusion and it requires little intellectual sophistication to recognize that fact. (BE1) states that all golf balls in C “go” into the cup. In order for justification to transmit from the premises to (BE3), which is a claim about the future, (BE1) must apply to the behavior of future golf balls in C. The universal generalization (BE1) makes no temporal distinctions. What is the justification for (BE1)? According to our reconstruction, it is (E1): all observed golf balls rolling toward a cup in circumstances C have gone into the cup.

Hume taught us that just because things have always gone a certain way doesn't mean they always will. Something besides (E1) is needed to justify (BE1). Clearly, if it is for the subject coincidentally the case that all observed golf balls in C have gone into the cup, that will not justify the inference to (BE1). There has to be some reason for the subject to think that the observed regularity will continue. This is not to require that in order to have justification for beliefs about the unobserved, a subject must have evidence that guarantees the observed regularity fits a general pattern. Indeed, the subject's justification may only make probable that the observed regularity is part of a general pattern. This justification may be defeated by future evidence, including future observations in similar circumstances where the previously observed regularity fails to obtain. The fact that the regularity might fail to hold in some future circumstances is insufficient to defeat the subject's justification at an earlier time for believing that the regularity would hold in the future, provided that the subject had, at that earlier time, good reason to think the regularity would hold in the future.Footnote ⁹ Regardless of how strong one's justification must be to justifiably believe a proposition describing a lawful connection between an observed regularity and a general regularity that would make probable the observed regularity, some such law-proposition must be justified in order for one to believe with justification that there is a general regularity. Thus, the universal generalization (BE1) needs to be justified by an inference from (E1) to a claim connecting being a golf ball in C with going in the cup as a matter of law, and from those premises to (BE1). Otherwise, there is no assurance that the behavior of future golf balls in C will resemble the behavior of past golf balls in C, and so (BE1) would be unjustified. And if (BE1) is unjustified, it can't justify (BE3).

Byerly agrees with this point. He argues that for his present evidence to justify (BE3) would require subjects to possess justification for believing there are laws of nature which entail (BE3), but such justification would require a degree of intellectual sophistication ordinary subjects lack (Reference Byerly2013: 239). Consequently, he thinks ordinary subjects are not justified in using descriptions of laws in their inferences. McCain agrees with Byerly that subjects are often not justified in believing there are laws of nature which would make claims like (BE3) probable, but McCain claims such justification is not necessary (Reference McCain2014a: 6).Footnote ¹⁰ The problem is that the universal generalization fails to justify propositions about the future unless the subject possesses justification for believing the observed regularity will continue as a matter of law.

In an instructive note, Laurence BonJour concludes his abductive argument for external world realism by saying that even if his argument works, and even if (as he thinks) the reasoning he requires on the part of ordinary subjects is not too complicated, it is unlikely people even implicitly work through such reasoning (Reference BonJour, BonJour and Sosa2003: 96). He concludes with skepticism about doxastic justification. BonJour's idea is that epistemologists ought to first identify the standards for justification, and then see what is required to meet those standards.

There is an important moral to be drawn from the Byerly–McCain exchange on this point. Their debate is an instance of the epistemologist's concern with discovering the conditions under which subjects could possess justification, while hoping those conditions can be met by non-epistemologists. Those who seem to require too much intellectual work for people to possess justification are charged with “over-intellectualization” while those who don't are accused of making justification too easy. Part of Byerly's criticism seems to be that the standards that explanationism requires are too high for ordinary subjects. In reply, McCain appears to lower the standards, instead of identifying a way to meet them.Footnote ¹¹ McCain seems to worry that requiring too much cognitive work will yield a result similar to BonJour's: it is possible, though unlikely, that anyone possesses the relevant sort of justification. I think it is clear that justification for believing a regularity is lawful is required to justifiably believe the regularity will hold in some future case. Explanationism has to accommodate this while avoiding skepticism about justification for beliefs about the future.

The second problem is that McCain's justification of (BE3) is accomplished solely by the use of enumerative induction plus syllogistic inference. No distinctly explanatory reasoning was involved. I don't wish to argue here that induction is logically independent of explanatory inference.Footnote ¹² That is a separate issue. The problem here is that, according to Conee and Feldman, BE is superior to “evidential proportionalism”: the thesis that evidential support is a matter of logical and probabilistic relations holding between propositions (Reference Conee, Feldman and Smith2008: 94). But if BE ‘explanations’ consist only in non-explanatory forms of logical and probabilistic inference, such as induction and syllogistic inference, then BE and proportionalism are not competitors. Propositions supported by explanatory relations to one's evidence will simply be propositions supported by logical or probabilistic relations to one's evidence. If that is so, BE has all the problems Conee and Feldman claim for evidential proportionalism, and evidential proportionalism has all the virtues Conee and Feldman claim for BE.Footnote ¹³

One might reply that if enumerative induction is a kind of explanatory reasoning, then the justification a subject gets through induction is justification the subject gets using explanatory reasoning. Consequently, Byerly could use induction while BE remains unscathed. While analyzing enumerative induction as a kind of explanatory inference could be used as part of an argument that BE and proportionalism are analytically distinct, the problem remains: BE and evidential proportionalism are not competing accounts of evidential support because they justify all and only the same propositions to the same degree. So, BE cannot solve problems raised for proportionalism. To block the objection, explanationists must identify non-inductive justification-conferring explanatory inferences at work in our reasoning. McCain at least seems to allow that inductive inference is sufficient to justify (BE1) independently of any non-inductive inference.Footnote ¹⁴ His argument for (BE1) does not use any independent explanatory reasoning. While ‘explanatory’ reasoning justifies (BE1), so does induction. That threatens the claimed superiority of BE. There is a dilemma here for BE. Either induction is a species of explanatory inference, and evidential proportionalism is no longer BE's competitor; or induction is not a species of explanatory inference, and BE is false.

Now for some replies on behalf of explanationism. To the first objection: explanationists should accept that beliefs about accidental regularities are insufficient to provide justification for universal generalizations. Therefore, subjects with justification for beliefs about the future have justification for beliefs about laws guaranteeing that regularities observed in the past will continue in the future. Fortunately, that justification is not too difficult to acquire.Footnote ¹⁵

There are two ways the defender of BE could try to accomplish this. What they have in common is the claim that ordinary subjects are able to entertain a belief that is simple enough to be within their ken, but which counts toward their justification that some observed regularity has a lawful ground. Regarding the first, what McCain explicitly rejects is that ordinary subjects must have justification for beliefs about particular laws of nature.Footnote ¹⁶ Perhaps McCain did not mean to reject any and all requirements concerning justification for beliefs about laws of nature, but only those concerning particular laws. As an initial pass, we might understand a belief about a particular law of nature to be a belief whose content fully describes some fundamental relations and entities (or properties, events, states of affairs, or whatever) and says they are necessarily connected. Such laws might be completely accurate descriptions of occurrences of interactions at the level of quantum physics. Maybe understanding a particular law of nature requires the possession of concepts few people have. One option for the explanationist is to argue that ordinary subjects have beliefs the content of which closely enough approximates beliefs about particular laws of nature. Those simpler beliefs function as proxy for more detailed beliefs about laws. The notion of approximation has to be worked out carefully. The important idea is that ordinary subjects can form the simpler beliefs, and if those beliefs are justified, they are sufficient to justify at least some of the same conclusions that the more complex beliefs would be, if the latter were justified. Applied to BE, this possible solution requires that sometimes subjects have justified beliefs about particular laws of nature, but they have them in an approximate form.

For example, our golfer may justifiably believe the putting green lacks dips and bumps, the swing just executed will send the ball straight, and so on. This belief closely enough approximates, for the purposes at hand, a much more detailed group of propositions describing the behavior of the fundamental entities and forces whose cooperation is necessary and sufficient for the putt to go in. The more detailed proposition has probably never been thought by anyone, and possibly never will. The golfer's belief approximates the complex proposition well enough for the simple belief to count toward the justification of the golfer's belief that the putt will go in. This notion of approximate belief allows the explanationist to require justified belief in laws of nature, but makes those beliefs easier to come by.Footnote ¹⁷

The second way is to forego any requirement of a belief in a law of nature, and allow that subjects instead reach justified conclusions about the future through beliefs about lawful regularities. The content of a belief about a lawful regularity simply asserts a lawful connection between events, properties, or states of affairs. A thought of the form, “As and Bs always go together” is a thought about a lawful regularity. Beliefs about regularities are simpler than beliefs approximating propositions describing laws of nature. Importantly, beliefs about regularities do not necessarily approximate more complex propositions describing the fundamental workings of nature. A subject can form a belief about a lawful regularity of As and Bs by inferring from the ubiquitous co-occurrence of As and Bs to some causal grounding or other underlying that regularity.Footnote ¹⁸

Which of the two views is preferable need not be resolved here because the reasoning a person must perform to meet plausible explanationist standards is the same for both. In the golf ball case, the subject must have a sufficiently great number of experiences of golf balls in C going into the cup. Bracketing worries about inductive skepticism, the subject may then reason as follows: “(1) All golf balls in C have gone into the cup. (2) The best explanation of (1) seems to be that the behavior of golf balls in C is not accidental; rather, there is some causal grounding underlying their behavior. (3) Therefore, all golf balls in C will go into the cup.” Explanationists should deal with putative counterexamples this way: allow that subjects reason from observed regularities to a lawful grounding of that regularity, and from that ground to specific instances. Explanationists will have to argue (on pain of inconsistency) that the law-belief is justified by inference to the best explanation. If the explanationist requires justified beliefs about particular laws of nature for the justification of beliefs about the future, then the reasoning the subject performs will have to sufficiently approximate a fully articulated belief about a law of nature; if the explanationist does not require that, then justified belief about a lawful regularity will have to be sufficient. In neither case is the required intellectual sophistication problematic. Whichever way one goes, one of those solutions should be acceptable to explanationists like McCain, and will allow them to defend BE without conceding to skepticism about the future.

While the responses just outlined adequately handle the first problem, I think the second problem remains unsolved. Even if BE and proportionalism are analytically distinct, BE is not a genuine rival to evidential proportionalism if BE “explanations” are equivalent to disguised uses of other forms of logical and probabilistic inference. I am not sure what such forms of irreducibly explanatory inference there might be, and that is no argument against the possibility of their existence: but that no plausible candidates suggest themselves leads me to doubt there are any. It is up to explanationists interested in securing the independence of BE from proportionalism to argue that explanatory reasoning is independent of these other forms of reasoning. For without being told what distinctly explanatory reasoning is, BE is not a new theory of evidential support, but a different way of describing an old theory of evidential support.Footnote ¹⁹