Symposium Introduction

A Theory of Evidence and Well-Founded Belief: Kevin McCain’s Evidentialism and Epistemic Justification

Evidentialism, at its core, is the view that the epistemic rationality of a belief is determined only by the evidence relevant to it. There are two domains wherein evidentialism occasions controversy. The first is whether a belief’s practical and spiritual valences may yield some counterexamples. And so the pragmatist may hold that it is better for some practical reason to believe beyond one’s evidence. Or the fideist may say that faith is a theological virtue. Call these the ethics of belief controversies. The second domain of controversy for evidentialism is comprised by the questions of what exactly the theory is. The evidentialist holds, as a theory of justification, that beliefs are rational only if they fit the evidence, and as a theory of belief-management, that one’s beliefs are well-founded when they are based on the evidence they fit. But the conceptual questions loom: What exactly is evidence? What is it to have evidence? What is it for a belief to fit the evidence? And how is it that beliefs are based on evidence? Call this second domain the meta-epistemic controversies for evidentialism.

Kevin McCain’s Evidentialism and Epistemic Justification (2014) is an essay in the meta-epistemic domain of controversy over evidentialism. (For recent work in the ethics of belief domain of controversy, see Syndicate Philosophy’s recent/upcoming symposium on Miriam McCormick’s Believing Against the Evidence.) Within this domain of meta-epistemic controversy, there is a wide variety of views. With the question of what evidence is and what sorts of things can be evidence, there are views, on the one hand, that evidence must be propositional, because evidential support must be in some recognizable inferential form. On the other hand, there are views that evidence may be simply psychological states. This is the propositionalism-psychologism debate. Further, there is the question of whether evidence must be factive or not. In many ways, this debate depends on whether one thinks there is a kind of contradiction in terms to say something of the form, S has evidence for p, but p is false. If it is a contradiction, one must think that evidence is factive. If one does not take that statement to be contradictory, then one must think that evidence is non-factive. This is the factivity of evidence debate.

With the question of what it is to possess evidence, one may be highly inclusive of what can be possessed by a subject, so one’s available evidence could be all of one’s total possible evidence, even the memories from one’s childhood one cannot recall. Alternately, one could be highly exclusive, so one’s available evidence is only what one is currently thinking about. Finally, one could be moderate with the possession requirement and hold that one’s available evidence is either what one is currently thinking or what one can easily remember or bring to mind when thinking about the matter. Views about how beliefs may properly fit the evidence range from the thought that they must be entailed by, made probably by, or are the best explanation for the evidence.

Finally, there is the question of what it is to believe on the basis of one’s evidence. On the one hand, there are doxastic theories, which run that the connection is made by a second-order belief about the evidence and one’s belief. On the other hand, there are causal accounts, which run that subjects believe what they do as the result of their reasons. Call this domain of controversy the basing question.

With the publication of Earl Conee and Richard Feldman’s 1985 essay, “Evidentialism,” and their further developmental work on the subject, questions about the nature of evidence, the fit between beliefs and evidence, and the basing relation between evidence and belief have been central meta-epistemological challenges for evidentialists. Kevin McCain’s book is designed to be a complete theory of epistemic justification in this vein. McCain defends a version of evidentialism that is mentalist and internalist, as he holds that theories of epistemic justification should be bounded by the thoughts driving the New Evil Demon Problem. McCain holds that evidence is either a mental state or a propositional content of one’s psychological state (27). Further, McCain holds that evidence is non-factive, since it seems intuitive that two mentally identical subjects can have the same evidence but one can be fooled by an illusion or have some bad epistemic luck and the other not. McCain defends a moderate view of evidential possession, so one’s dispositionally available thoughts or memories are evidence one possesses, but not all one’s memories.

With regard to the question of how one’s belief can fit the evidence, McCain adopts a modified form of explanationism. He holds that one’s belief that p fits one’s evidence (i.e., one’s non-factive mental states or what one is disposed to recall) when p is either the best explanation for that evidence or is available as a logical consequence of that best explanation (79). And so, if one has the visual impression of two birds and two squirrels on the deck, one would be justified in holding that there are four animals on the deck because it is a logical consequence of the best explanation for the visual impression.

Finally, with regard to the basing question, McCain develops a particular form of the causal theory of the basing relation, one that depends on an interventionist account of causation. McCain holds that his version avoids the challenges of overdetermination and deviant causal chains that plague other versions of the causal theory, and so S’s belief is based on S’s evidence when each piece of S’s evidence is both the direct and actual causes of S’s belief (91). S’s evidence directly causes S’s belief when it causes the belief without detouring through any evidentially irrelevant mental states. In other words, when S’s evidence directly causes S’s belief it doesn’t cause the belief by way of causing some other mental state that in turn causes the belief. When it comes to S’s evidence being an actual cause of S’s belief the rough idea is that S’s actually having that evidence makes a difference to S’s having or not having the belief.

McCain’s program is posited on the distinction between propositional and doxastic justification, and evidentialists have regularly thought that both parts are required for an appropriate theory. We want an account of what it is, first, for a belief to be appropriately supported by evidence (so propositional justification) and also for us to be correctly cognitively moved by those justifying reasons (doxastic justification). That is, we want the evidence not only to favor our beliefs, but for us to hold those beliefs because the evidence favors them. So a complete theory of evidentialism requires this two-part account. In its complete form, McCain’s two-part theory is as follows:

Propositional Justification: Explanationist Epistemic Justification (Ex-EJ)

  1. Believing p is epistemically justified for S at t if and only if at t S has considered p and:

1) p is part of the best explanation available to S at t for why S has her occurrent non-factive mental states and the non-factive mental states that she is disposed to bring to mind when reflecting on the question of p’s truth

OR

2) p is available to S as a logical consequence of the best explanation available to S in (1

2. Withholding judgment concerning p is epistemically justified for S at t if and only if at t S has considered p and neither believing p nor believing ~p is epistemically justified for S.

Doxastic Justification: Explanationist Well-Foundedness (Ex-WF)

At t, S’s belief that p is well-founded if and only if:

At t,I. 1. Each

I. 1. Each ei E (the non-factive mental states that constitute S’s evidence) is a direct cause of S’s believing that p

AND

  1. Each eiE is an actual cause of S’s believing that p

AND

  1. It is not the case that intervening to set the values of all direct causes of S’s believing that p, other than the members of E, to 0 will result in S’s not believing that p when every eiE is held fixed at its actual value (i.e. S’s evidence plays a strong enough causal role in S’s believing that p).S’s belief that

II.        S’s belief that p is epistemically justified for S at t by E (i.e. condition (I) of Ex-EJ is satisfied).

III.       At t there is no set of S’s evidence, E* such that:

  1. E is a subset of E*

AND

  1. p is not epistemically justified for S at t by E*

 

McCain then turns to making the case that this explanationist evidentialism accommodates the intuitive results of a number of important cases. It can explain why firsthand experience with whether it is a warm day puts one in a better position than another who has only read the paper about the weather (121–22), and it explains why faint memories provide less justification than vivid memories (124). Further, McCain argues, explanationism yields important anti-skeptical consequences. Explanationism provides us with reasons to hold that the truth of our commonsense beliefs about the world is the best explanation (over various skeptical hypotheses) for the relevant features of our sensory experience (216). The viability of McCain’s explanationist evidentialism demonstrates that evidentialism can be fleshed out into a complete theory of epistemic justification. The upshot, then, is explanationist evidentialism promises a variety of important and appealing consequences for several debates central to contemporary epistemology.

Brian Cutter and Philip Swenson

Response

Does It Matter Which Explanation Is Best?

Explanationists hold that epistemic justification depends fundamentally on explanatory considerations.1 Kevin McCain’s excellent book Evidentialism and Epistemic Justification defends an evidentialist version of Explanationism. We greatly benefited from reading McCain’s book and are sympathetic with much of what he has to say. But we will suggest one modification to his account. McCain requires that the best available explanation provide support for a proposition in order for the proposition to be justified. We will argue that this requirement faces difficulties and suggest an alternative approach. Explanationists should accept that epistemic justification does not just depend on the support it receives from the best explanation of one’s evidence, but on the collective support it receives from all competing explanations of one’s evidence.

McCain appears to be committed to both of the following:

Explanationism (roughly construed): Whether p is justified for S depends fundamentally on explanatory considerations.2

And also:

Best Explanationism: Whether p is justified for S depends fundamentally on p’s relationship to the best (or best available) explanation of S’s total evidence.

In his book, McCain holds that S is justified in believing p if and only if either

(i) p is part of the best explanation of S’s evidence available to S

or

(ii) p is available to S as a logical consequence of that explanation.3

In a more recent paper, he replaces requirement (ii) with the following:

(iii) p is available to S as an explanatory consequence of the best explanation [of S’s evidence] available to S.4

Both of these views are versions of Best Explanationism. They both say that justification depends on some relationship holding between p and the best available explanation. It is not surprising that McCain accepts Best Explanationism. It is common for Explanationists to think that justification should be accounted for in terms of which explanation is best.5 We will argue that Best Explanationism creates unnecessary difficulties for Explanationists and that there is a more plausible form of Explanationism available, one which does not assign the best explanation a position of overriding privilege.

1: Worries for Best Explanationism

Consider the following situation. There are three (mutually exclusive, jointly exhaustive) potential explanations of your total evidence available, and they have the following epistemic probabilities for you:

E1: 34% probability

E2: 33% probability

E3: 33% probability

E1 contains p as a part, while E2 and E3 both contain ~p.

There are two potential worries that arise for McCain (and Best Explanationism in general) as a result of this scenario.

Worry 1: Can McCain (and Best Explanationism in general) account for the fact that you should not believe p?

Worry 2: Can McCain (and Best Explanationism in general) account for the fact that you are justified in believing ~p? (If you do not think that a 66% epistemic probability is sufficient for justified belief, then imagine a version of the case where there are 100 available explanations and all but one, which is very slightly more likely than the rest, contain ~p.)

McCain has two plausible responses to Worry 1. First, McCain could require that in order for you to be justified in believing p, E1 must not only be the best explanation, “it must also be a sufficiently good explanation” (McCain 2015, 339). Perhaps E1 does not qualify as “sufficiently good,” since it is only 34% likely on your evidence. Second, McCain has suggested in correspondence that there may be a more general explanation available which is better than E1. Perhaps:

E4: Some explanation other than E1 is true.

is a better explanation of your evidence than E1.

Both of these approaches seem like initially promising responses to Worry 1. Worry 2, however, is more serious. In order to secure the result that you are justified in believing ~p, McCain must establish that ~p is either part of the best available explanation or entailed by the best available explanation (or alternatively, is an “explanatory consequence” of the best available explanation). McCain could attempt to appeal to explanations along the following lines:

E5: Either E2 or E3 is true.

Or, more generally:

E6: Some explanation that includes ~p is true.

E5 and E6 both entail the truth of ~p. And one could make a case that they are better explanations than E1. However, it is not plausible that either of them are the best explanation of your evidence. Consider:

E7: Either E1 or E2 is true.

Once we allow general or disjunctive explanations into play, E7 looks pretty good. Given your evidence, E7 is 67% probable, while explanations like E5 and E6 are only 66% probable. Thus E7 looks like a better explanation than either E5 or E6.

E7 does not entail ~p. In fact, given E7, ~p is more likely to be false then true. (This is because E1 contains p and has a higher probability than E2.) Thus it is implausible that E7 either contains ~p in the relevant sense or has ~p as an explanatory consequence.

So it looks like Worry 2 creates significant trouble for Best Explanationism. In order to respond to it, the Best Explanationist would have to find an explanation that is better than E1 and E7, and is related to ~p in the appropriate way. This looks like a difficult task.

2: An Alternative Approach

Reflection on the foregoing leads us to think that, even by Explanationist lights, Best Explanationism is misguided. Intuitively, you can be justified in believing ~p if the available explanations of your evidence collectively support ~p to a sufficiently high degree, even if the best explanation does not. E2 and E3 both provide some support for ~p. The combined strength of these different sources of support suffice to justify you in believing ~p. Instead of Best Explanationism, Explanationists should therefore accept:

Multiple Source Explanationism: S’s justification for believing p depends on the collective support that p receives from all competing explanations of S’s evidence.

We conclude by proposing a (somewhat idealized) formal framework for Multiple Source Explanationism to illustrate how our justification for believing a proposition is sensitive to its relationship to all the candidate explanations for our evidence.

Let E1, . . . , En be the complete menu of competing explanations for our total evidence. For simplicity, we’ll assume that E1, . . . , En form a partition on logical space, i.e., that they are pairwise inconsistent and their disjunction is logically necessary. From here, we make two theoretical assumptions. First, we take for granted that explanations can be better or worse and, more specifically, that there is gradable aspect or dimension of explanatory goodness that corresponds to (perhaps determines) the degree of confidence we rationally ought to have in the truth of the explanation, given our total evidence.6 We make the common idealizing assumption that levels of confidence (“credences”) can be represented by real numbers in the interval [0, 1]. And since the relevant dimension of explanatory goodness for an explanation E (symbolized as “G(E)”) corresponds to the rational credence in E given our total evidence (symbolized as “C(E)”) we’ll assume that G can be measured on a similar scale. That gives us:

Assumption 1: C(E) = G(E)

It would be convenient if we could assume that every candidate explanation either entails p or entails ~p (as with E1–E3 above). For in that case, the question “how much justification do we have to believe p?”—or, better, “how confident should we be in the truth of p?”—would have a straightforward answer: our confidence in p should be equal to the sum of the G-values over each Ei that entails p. But there is no reason to expect this convenient assumption to hold in general. In addition to explanations that entail p and explanations that entail ~p, there will typically be a range of intermediate cases, in which an explanation makes p probable to some positive degree short of certainty. (McCain [2015] appears to grant something along these lines with his appeal to non-entailed “explanatory consequences.”) Thus, our second assumption: an explanation can (as we’ll say, for lack of a better word) support a proposition p to a greater or lesser extent. More specifically, we’ll assume that the degree to which p is supported by an explanation E (symbolized as “S(p, E)”) corresponds to (perhaps determines) the level of confidence one rationally ought to have in p given E. That gives us:

Assumption 2: C(p|E) = S(p, E)

We now ask: are we justified in believing p? Or, better: how confident should we be in the truth of p? Answer: our confidence in p should be equal to the weighted sum of the support given to p by all the candidate explanations of our total evidence, with weights provided by the G-value of each explanation. That is:

Total Support: C(p) = S(p, E1)G(E1) + . . . + S(p, En)G(En)

The reader may notice that Total Support takes the form of the theorem of total probability, according to which

Total Probability: Pr(B) = Pr(B|A1)Pr(A1) + . . . + Pr(B|An)Pr(An),

where B is any proposition, and A1, . . . , An form a partition on logical space. This is no coincidence. Given the assumption that rational credences conform to the axioms of the probability calculus, Total Support is a straightforward consequence of the theorem of total probability, given Assumptions 1 and 2.

From the perspective of Multiple Source Explanationism, we can see that there cannot be a general recipe for determining our degree of justification for p that relies exclusively on information about p’s relationship to the best explanation. Formally, this is because there is no way of working out the value of C(p) in the Total Support equation if we’re only given information about the values of S(p, Ei) and G(Ei) for a single explanation Ei (except in the special case where G(Ei) = 1). As a simple illustration of this point, let’s consider two cases. In each, there are only two explanations for our total evidence, “Best Explanation” and “Worst Explanation” (as before, we assume they partition logical space), with epistemic probabilities and support relations as indicated below:

Case 1:

Best Explanation: 60% likely; supports p to degree .4

Worst explanation: 40% likely; supports p to degree 0 (entails ~p)

Case 2:

Best Explanation: 60% likely; supports p to degree .4

Worst Explanation: 40% likely; supports p to degree 1 (entails p)

In Case 1, our credence in p should be .24. In Case 2, our credence in p should be .64. But notice that the facts about p’s relationship to Best Explanation (including facts about the goodness of Best Explanation) are the same in each case. Any theory that only takes account of the best explanation will therefore lack the resources to generate the intuitively correct result that we ought to have different attitudes toward p in these cases. The general lesson is that the attitude we are rationally justified in taking toward a proposition is not a function of its relationship to the best explanation alone. Given the Explanationist assumption that epistemic justification is fully determined by explanatory considerations, we must allow that epistemic justification is a function of a proposition’s relationship to all the available explanations of our evidence.

References

McCain, Kevin. “Explanationism: Defended on All Sides.” Logos & Episteme 6:3 (2015) 333–49.

McCain, Kevin. Evidentialism and Epistemic Justification. New York: Routledge, 2014.

Poston, Ted. Reason and Explanation: A Defense of Explanatory Coherentism. Palgrave Macmillan, 2014.


  1. Thanks to Simon Goldstein, Kevin McCain and Andrew Moon for helpful comments.

  2. Ted Poston offers a somewhat different definition. “Explanationism is the view that one’s normative standing in the space of reasons is constituted by one’s explanatory position” (2014, p. 69).

  3. Note that this is our statement of McCain’s view. See McCain (2014, p. 117) for the official statement.

  4. By “p is an explanatory consequence of the best explanation available to S” McCain means that “if p were true, the best available explanation of S’s evidence would better explain its truth than it would the truth of ~p, if ~p were true” (2015, 339).

  5. Poston (2014) also endorses something along the lines of Best Explanationism.

  6. We emphasize that this is only an aspect or dimension of explanatory goodness because the overall extent to which E is a good explanation for some phenomenon X is not only a matter of how likely E is given X, but also (inter alia) how likely X is given E. This is why Jones received something nice in the mail may in some contexts be a better explanation of the fact that Jones is happy than Jones received something in the mail, though the latter is more likely given the explanandum.

  • Kevin McCain

    Kevin McCain

    Reply

    It Matters Which Explanation Is Best: A Reply to Cutter and Swenson

    Brian Cutter and Philip Swenson (hereafter “C&S”) challenge Explanationist Evidentialism on the grounds that it makes justification depend upon the best available explanation. The worry that C&S raise, if successful, threatens not just Explanationist Evidentialism, but any theory of epistemic justification that restricts justification to the best explanation (I will follow C&S in referring to such views as “Best Explanationism”). Although this sort of objection to explanationism is one that is often brought up in conversations, it is seldom made explicit in print.1 I am grateful to C&S for clearly presenting the challenge and explaining its purported force, though I will be arguing that it ultimately fails as an objection to Explanationist Evidentialism, and Best Explanationism in general.

    Before diving into the details of C&S’s case against Best Explanationism it is worth pausing to note that in addition to raising an objection they also offer an alternative way of developing explanationism. C&S suggest that “Multiple Source Explanationism” can avoid the problem that they maintain arises for Best Explanationism. According to Multiple Source Explanationism, justification isn’t a matter of which explanation is the best. On the contrary, C&S’s theory construes justification as “a function of a proposition’s relationship to all the available explanations of our evidence.” C&S’s proposal is interesting and worthy of careful consideration. I should mention at the outset that it strikes me as a theory with some initial plausibility. Additionally, it is should be noted that one could accept C&S’s Multiple Source Explanationism without abandoning the central explanationist tenant that justification ultimately depends on one’s overall explanatory position. So, although this view is a rival of my own, it’s still in the explanationist family. Of course, whether one should accept Multiple Source Explanationism or a form of Best Explanationism, such as my Explanationist Evidentialism, comes down to how they stack up as theories of epistemic justification. In other words, the decision between these theories comes down to which one provides us with the best explanations of the relevant data. Unfortunately, I do not have the space to thoroughly evaluate the comparative merits of these competing explanationist approaches here. Instead, I will focus on showing that C&S’s attack on Best Explanationism doesn’t motivate accepting their theory because it fails to pose a genuine problem for Best Explanationism.

    In order to illustrate their purported problem for Best Explanationism C&S ask us to consider a situation where “there are three (mutually exclusive, jointly exhaustive) potential explanations of your total evidence available” to you: E1, E2, and E3. The epistemic probabilities for you for these explanations is as follows: E1 = .34, E2 = .33, and E3 = .33. C&S add that “E1 contains p as a part, while E2 and E3 both contain ~p.” What’s thought to make this case problematic for Best Explanationism is that it seems that p is part of the best explanation of your evidence (E1 is more epistemically probable than either of the other alternatives), but as C&S point out the probability of p in this case is only .34, but the probability of ~p is .66. Hence, it seems that you should believe that ~p is true even though p is part of the best explanation of your evidence.2 C&S claim that this fact is contrary to the verdict that Best Explanationism yields in this case.3 So, it may appear that Best Explanationism is in serious trouble.

    Initial appearances notwithstanding, Best Explanationism is not impugned by the sort of case that C&S describe. As a matter of fact, Best Explanationism yields the intuitively correct result that you should believe ~p in this case. Recall, Best Explanationism says (roughly) that you should believe that a particular proposition is true when it is part of the best explanation of your total evidence. In this case the best explanation of your evidence includes that the probability of p = .34 and the probability of ~p = .66. The reason for this is that your total evidence includes the epistemic probabilities that C&S mention for E1, E2, and E3. Consequently, the best explanation of your total evidence is that ~p has probability .66. Assuming that probabilities link up with belief in the way that C&S assume (high enough probability means that one should believe the proposition in question), your total evidence justifies you in believing ~p.

    It is worth unpacking this response on behalf of Best Explanationism a bit more. This response rests on the fact that the explanation of which p is a part, E1, is not the best explanation of all your evidence. Your evidence includes that it is almost twice as likely that E1 is false than it is that E1 is true. As a result, even though E1 is the best explanation of a portion of your evidence, the best explanation of your total evidence includes what C&S call “E4,” “Some explanation other than E1 is true.” Of course, as C&S point out the truth of E4 by itself isn’t enough for Best Explanationism to deliver the result that you should believe that ~p. Fortunately for Best Explanationism, the best explanation of your total evidence also includes the proposition <If some explanation other than E1 is true, then ~p is true>. Why is this? It is because as C&S have set things up your evidence includes the information that the only alternatives to E1 include ~p (after all, they stipulate that E1, E2, and E3 are “jointly exhaustive”). Now, plausible forms of Best Explanationism, such as my Explanationist Evidentialism, allow that a proposition is justified for you when it is either part of the best explanation of your total evidence or a logical consequence of the best explanation of your total evidence.4 It is clear that <Some explanation other than E1 is true> and <If some explanation other then E1 is true, then ~p is true> together entail that ~p is true. Thus, given that these two propositions are part of the best explanation of your total evidence, Best Explanationism yields that believing ~p is justified for you. And, as we’ve seen there’s good reason to think that these propositions are part of the best explanation of your total evidence. Therefore, Best Explanationism provides the intuitively correct result in C&S’s case. As a result, while C&S do a very nice job articulating an objection that often arises when one is discussing explanationism, they don’t provide a reason to abandon Best Explanationism.5


    1. A notable exception is Richard Fumerton, Metaepistemology and Skepticism (Lanham, MD: Rowman & Littlefield, 1995).

    2. C&S note that if one doesn’t think that .66 probability is sufficient for justifying belief, their case can be modified so that the probability of ~p is very high even though the explanation containing p is still better than any of the individual ~p explanations. In light of this, I’m willing to grant C&S’s assumption that .66 probability is sufficiently high to justify belief.

    3. C&S’s challenge is a version of what Ted Poston and I (unpublished manuscript) have termed the “Disjunction Objection” to Best Explanationism.

    4. As C&S note, I actually think that the best way to construe Explanationist Evidentialism is in terms of explanatory consequence rather than logical consequence. However, for the present purpose this distinction will not make a difference because both ways of understanding Explanationist Evidentialism produce the same result in this case. Readers interested in seeing why I opt for explanatory consequence are encouraged to see Kevin McCain, “Explanationism: Defended on All Sides,” Logos & Episteme 6 (2015) 333–49.

    5. Thanks to Matt Frise for helpful discussion.

    • Philip Swenson

      Philip Swenson

      Reply

      Question for McCain

      Hey Kevin,

      Thanks for replying (and for writing such a great book)!

      You say that

      E4: Some explanation other than E1 is true.

      is the best explanation in the case we give. But I wonder why you think E4 is a better explanation than:

      E7: Either E1 or E2 is true.

      E7 does not entail ~p. And it seems to be slightly more likely than E4.

      -Philip

    • Kevin McCain

      Kevin McCain

      Reply

      Brief reply to Swenson

      Hey Philip,

      Thanks for the great follow-up question! Like the essay you and Brian wrote, it’s helping me work through some important issues. Thinking some more about the example (after sending in the final draft of my above response) led me to believe there’s a more straightforward way for the “best explanationist” to handle the case you all described.

      Here’s what I’m thinking:

      In this case part of the best explanation of my total evidence is that p is .34 probable and ~p is .66 probable. After all, the epistemic probabilities work out this way. As a result, best explanationism holds that my evidence supports believing that ~p is .66 probable.

      Should I believe that ~p is true in this case? It depends on how we understand the relationship between epistemic probability and belief. I mention in the book that best explanationists might go a couple ways here—one way is to link epistemic probability directly with what’s reasonable to believe so that a proposition’s having a sufficiently high epistemic probability just is for it to be reasonable to believe that the proposition is true; another way to go is to maintain that these are separate so that when the epistemic probability of ~p is .66 we shouldn’t believe that ~p is true, but instead should only believe that ~p is .66 probable. Either way works fine for the best explanationist.

      I believe that this handles the issue that you and Brian raise in a fairly simple and straightforward manner. What do you think?

    • Brian Cutter

      Brian Cutter

      Reply

      Follow-up question

      Hi Kevin,

      You’ve suggested that the relevant facts about epistemic probabilities (e.g. that p has an epistemic probability of .34 on your evidence) would themselves be parts of the best explanation. I’m not sure how much hangs on the issue, but this isn’t quite how we were thinking about things. I was thinking that the situation was something like this: we start we our total evidence T. Then there are various candidate explanations for our evidence, each of which has a certain epistemic probability for us, given our evidence. In a simple case (like the one we describe in the paper), where the candidate explanations partition logical space and a certain proposition p is a part of one of the explanations (call it E) and ~p is a part of all the others, presumably we’d want to say that the epistemic probability of p is equal to the epistemic probability of E given T. So, if the probability of E (given T) is .9, then the probability of p would, in this case, also be .9. But this fact about epistemic probabilities—that p has an epistemic probability of .9—does not itself belongs to E (what belongs to E is just p itself), nor does it necessarily belong to any other candidate explanation.

      Applied to our case: we initially listed three possible explanations of our total evidence: E1, E2, and E3. But none of these explanations includes “p has a probability of .34,” or “~p has a probability of .66,” or any other fact about epistemic probability. Nor does any disjunction of the E’s contain such facts. So I don’t see why we should say that “~p has a probability of .66” belongs to the best explanation of one’s evidence, as opposed to being a fact grounded in ~p’s relation to all candidate explanations of one’s evidence.

      — Brian

    • Brian Cutter

      Brian Cutter

      Reply

      Generalizing our concern

      Hey Kevin,

      First of all, let me echo Philip by thanking you for your generous reply. I want to try to distill our general worry in a way that abstracts away from the specific case we give involving E1-E3. (I think this post will do a better job getting to the heart of my concern than my previous post.)

      Background Question: it is possible for an explanation E to be the *best* explanation of one’s total evidence, even if E is not *certain* (i.e. has an epistemic probability < 1) given one's evidence?

      I'm pretty sure your answer to this question is "yes," so I'll proceed on that assumption. Now I want to argue that, in cases where the best explanation is not *maximally good* (i.e. is not certain, given one's evidence), the correct doxastic attitude toward p cannot be determined by p's relationship to the best explanation alone.

      So, suppose that E is the best explanation of my total evidence, and suppose (as should be possible if your answer to Background Question is "yes") that the epistemic probability of E for me is less than 1. (For definiteness, we can suppose that it is 0.7, but nothing here hinges on the precise number, or even on the assumption that levels of certainty can be adequately represented by precise real numbers.) Now consider two cases:

      Case 1: p is a part of (or is entailed by) E, and moreover p is a part of (or entailed by) all the alternative (suboptimal) explanations as well.
      Case 2: p is a part of (or is entailed by) E, and moreover all the alternative explanations entail ~p (that is, p is inconsistent with all the alternative explanations).

      The two cases are identical with respect to p's relationship to the *best* explanation, since in both cases, p is a part of (or entailed by) the best explanation—namely, E. Therefore, it seems that Best Explanationism entails that we ought to have the *same* doxastic attitude toward p in each case. But intuitively, this is the wrong result. We should be more confident in p in Case 1 than in Case 2. More specifically, our confidence in p should be around 1 in the first case, and around .7 in the second case.

      I suppose I can summarize the point with a question: do you think Best Explanationism has the resources to yield (what I take to be) the intuitively correct verdict here, namely that we ought to have different doxastic attitudes in these two cases?

      — Brian

    • Kevin McCain

      Kevin McCain

      Reply

      Brief reply to Cutter

      Hi Brian,

      This is very helpful! You’ve helped me see that there is an important assumption at play in the case that you and Philip describe. In your case it is assumed that it’s possible that the best explanation (presumably it is a good explanation too) of one’s total evidence isn’t likely to be true even when you have 3 jointly exhaustive explanations. (I missed this earlier because I was thinking that you would allow the additional evidence about epistemic probabilities that I mentioned in my earlier responses to be part of the total evidence. In which case, E1, E2, and E3 would no longer be jointly exhaustive explanations of one’s total evidence.)

      I think that explanationists should simply deny this assumption. After all, accepting it amounts to accepting that explanatory reasoning isn’t justifying. Of course, this is something that explanationists should deny. As a result, explanationists should deny that this case is possible.

      To help motivate this denial a bit more, it may be worth considering that the assumption at play in this case involves a more negative view of explanatory reasoning than even van Fraassen’s classic (though mistaken) objection to inference to the best explanation, the Best of a Bad Lot. van Fraassen attacks explanatory reasoning on the grounds that while E1 might be the best explanation, it may only be the best among a set of bad explanations. However, the Best of a Bad Lot only applies when you don’t have jointly exhaustive explanations, i.e. you can’t be sure that the truth is among the set of competing potential explanations. The assumption that drives the case you and Philip describe applies even when one knows that the true explanation is among those considered and the explanations are explanations of your total evidence.

    • Kevin McCain

      Kevin McCain

      Reply

      Reply to the Generalized Concern

      Hey Brian,

      Thanks for expressing the general concern. This is a nice way to put the broader worry that you and Philip have for best explanationism. The short answer to your question is “yes”, I do think that best explanationism has the resources to yield the intuitively correct result in this sort of situation.

      I think there is an important question about both case 1 and case 2 though: is the information about the various explanations and p’s role in them part of one’s evidence or not?

      If the answer to this is “yes”, then I think that best explanationism can yield the result that you mention because the best explanation of one’s total evidence will include information about p being contained in all of the potential explanations in case 1 and only in the best explanation (and none of its rivals) in case 2.

      If the answer to the question is “no”, then I think that best explanationism will yield that we ought to have the same doxastic attitude toward p in both cases. However, I don’t find this problematic because we have stipulated that the information about these alternative explanations and p’s role in them isn’t part of one’s evidence. I think that it is this information that drives the intuition that p is better supported in case 1 than case 2. If this information isn’t available to the subject, then it doesn’t seem implausible to claim that her doxastic attitude toward p should be the same in both cases. What do you think?

    • Brian Cutter

      Brian Cutter

      Reply

      Response to McCain

      Hi Kevin,

      Thanks, this is helpful. Regarding your question, I suppose I’m happy to say yes, the information about the various candidate explanations and p’s relation to them can be regarded as part of one’s evidence. You write,

      “If the answer to this is “yes”, then I think that best explanationism can yield the result that you mention because the best explanation of one’s total evidence will include information about p being contained in all of the potential explanations in case 1 and only in the best explanation (and none of its rivals) in case 2.”

      This is interesting. I have one small-ish lingering question and one conciliatory remark.

      The question: given that my *evidence* includes information about the various candidate explanations and p’s relationship to them, why should it follow that *the best explanation of my evidence* also includes this information?

      The conciliatory remark: if we take your suggestion in the passage quoted above, then it looks like you want to say that the best explanation will *always* include all the relevant facts about the alternative explanations on offer, together with facts about each proposition’s logical/probabilistic relations to those explanations (at least insofar as these facts are part of one’s evidence). In that case, we might be able to determine the appropriate doxastic stance toward each proposition p just by looking at p’s relationship to the *best* explanation, but only because the best explanation implicitly contains all the relevant information about the alternative explanations. In one way, this would be a victory for Best Explanationism, since it allows us to uphold your view that one’s justification to believe p is a function of p’s relation to the best explanation alone. But in another way, it’s a vindication of Multiple Source Explanationism, since we’re left with the conclusion that one’s justification to believe p is, after all, sensitive to the collective support that p receives from all the competing explanations. (It’s just that these facts about collective support are in some way bundled into the best explanation.)

      Does that seem like a fair characterization of where we’re left?

      –Brian

    • Kevin McCain

      Kevin McCain

      Reply

      Reply to Cutter

      Hi Brian,

      Thanks for the follow-up. This is very helpful!

      Concerning your question, “given that my *evidence* includes information about the various candidate explanations and p’s relationship to them, why should it follow that *the best explanation of my evidence* also includes this information”?
      –I was a bit careless with my wording in the previous post! You’re right to point out that if the information in is your evidence it isn’t clear that the information must also be a part of the best explanation of your evidence. What I meant, and should have said, is that this information (about the various candidates and p’s relationship to them) is relevant to determining which explanation is the best. Since this information is in your total evidence, it is something that is to be explained by the best explanation.

      Concerning your conciliatory remark, I do think this is a fair characterization of where we’re left. The only thing that I’d modify is the claim that “It’s just that these facts about collective support are in some way bundled into the best explanation.” Instead of this, I think the facts about the various alternative explanations and p’s role in them are part of what is explained by the best explanation.

      Thanks again for the great discussion!
      -Kevin

Amy Flowerree

Response

Infection and Directness in the Interventionist Account of the Basing Relation

fIn Evidentialism and Epistemic Justification, Kevin McCain puts forward a defense of an Evidentialist Explanationist theory of justification. In it, he presents a novel account of the basing relation. Drawing from the interventionist account of causation, he proposes a solution to the problem of deviant causation (widely viewed to be The Problem haunting causal accounts of the basing relation). In this paper, I will raise two problems for McCain’s account: an Infection Problem (in a Global and Local form), and a Direct Cause Problem. The Infection Problem questions whether the account can capture features that undermine doxastic justification. The Direct Cause Problem queries the adequacy of McCain’s account of basing by challenging how evidence could be a direct cause of belief.

The Account

McCain proposes the following account of the basing relation.

IB-R: S’s belief that p at t is based on her evidence, E, if and only if at t:

  1. Each e1 included in E is a direct cause of S’s believing that p
  2. Each e1 included in E is an actual cause of S’s believing that p
  3. It is not the case that intervening to set the values of all direct causes of S’s believing that p, other than the members of E, to 0 will result in S’s not believing that p when every e1 included in E is held fixed at its actual value.[1]

When evaluating whether S’s belief that p (Bp) is well based, the interventionist model will examine all relevant causal influences on the formation of Bp. The set of causal factors can include evidence (non-factive mental states that propositionally justify Bp), other mental states, mental mechanisms, and external influences. An entity, X, is a direct cause of an entity, Y, just in case intervening on X results in a change in value to Y. X is an actual cause of Y just in case varying the actual value of X would vary the actual value of Y, when all other causal factors are held fixed.

In sum, IB-R aims to overcome the traditional problems of deviance and overdetermination by providing a nuanced framework of causal relations. In what follows, I will challenge the adequacy of this account. But first, I’d like to praise its inventiveness. IB-R gives us a way to pull apart the complex web of causes that form our beliefs, and pinpoint exactly where belief formation can go wrong. While I think the interventionist model needs to be fleshed out more carefully, it is a highly promising and novel account of causal basing. The problems with it are, I think, problems with McCain’s Evidentialism, not problems inherent to the interventionist framework.

The Infection Problem

The Infection Problem is a challenge to the sufficiency of IB-R. Here, I will present cases that meet IB-R, but intuitively fail to be doxastically justified. The beliefs fail to be doxastically justified because they suffer from some kind of infection. Local infection occurs when doxastically unjustified beliefs are included in E. I pose a dilemma for McCain and argue he has no acceptable way out. Global infection occurs when cognitive biases (that can only be understood by looking at the agent’s belief forming dispositions counterfactually) undermine doxastic justification. Since these global features are counterfactual and holistic, they cannot be captured by looking only at actual and direct causes.

Local Infection

S’s evidence, E, consists of non-factive mental states which propositionally justify S in believing p. But what if the beliefs in E are not themselves doxastically justified? Let p be “there is a lottery ticket at my feet,” and q be “the lottery is rigged in my favor.” Now consider the case of Ann.

ANN[2]

Suppose Ann is struck by lightning and the effect of the lightning is that a lottery ticket appears at her feet and she forms the belief that p and if p then q. She reasons to the belief that q. She is propositionally justified in believing this, since her non-factive mental states imply it. But, intuitively, her belief that q isn’t doxastically justified, and even if true, her belief that q isn’t knowledge.

Since IB-R does not require that every member of E must itself be doxastically justified, it seems that it gets the wrong verdict in ANN.[3] McCain could give three responses: first, McCain could hold that Ann has a defeater, and so her belief is not propositionally justified. But suppose Ann is a bit stunned and not very reflective. She doesn’t have any background beliefs that directly undermine her justification. She is just happy at her good fortune. Second, McCain could insist that Ann is in fact justified. I think this is implausible, for reasons that McCain himself raises against the doxastic view of the basing relation.[4] If Ann bases her belief on beliefs that are not themselves well-based, then it seems that her belief could not be doxastically justified.

The most promising route, then, seems to be the third option: augment IB-R to require that S’s beliefs must be well-based on well-based based beliefs in order to count as doxastically justified. This seems to be McCain’s approach. After the official formulation, he suggests an extra condition: a belief can only be well based if it is well based on non-factive mental states that are themselves well based.[5]

However, Modified IB-R generates two regresses. The first regress is conceptual. IB-R is an account of the basing relation. Modified IB-R ends up explaining doxastic justification in terms of a relation between a belief and other doxastically justified beliefs. We cannot understand what it is to be doxastically justified without making use of the concept of doxastic justification. This regress might not be vicious: we could reduce any particular relation to a causal relation, though we cannot explain the concept of doxastic justification without making use of the concept in both the explanandum and the explanans. If our aim was to give an account of the nature of basing, then this would be a problem. But if our aim is just to judge whether a particular belief is properly based, then modified IB-R may be sufficient.

The conceptual regress illuminates an additional problem: Modified IB-R is not an instance of the Orthodox View.[6] It does not explain doxastic justification purely in terms of propositional justification. It does not give an account of what it is for a belief to be doxastically justified. Instead, it only explains how doxastic justification is transmitted.

The second regress is a causal one, and it does seem vicious. Modified IB-R requires that for any belief to be well-based, it must stand in a causal relation to E, which includes other beliefs that are themselves well-based. S could not have a doxastically justified belief without having an infinite chain of doxastically justified beliefs. A belief stands in a particular causal relation to a set of evidence, which in turn stands in a particular causal relation to a set of evidence, so on ad infinatum. Modified IB-R seems to rule out foundationalism, since to be well-based, a belief must be based on evidence that is itself well-based. A solution to the causal regress might be coherentism (the preferred view of old school evidentialists). But coherentism does not seem obviously compatible with the interventionist model of basing.

Global Infection

Local infection is not the only way that doxastic justification can be undermined. Doxastic Justification can also be undermined by features of the agent that are not present in E. Turri’s argument against the Orthodox View highlights this point. He presents cases in which the agent is propositionally justified, the evidence causes the belief, but the mechanism that brings about the belief is compromised. The mechanism can be compromised in two ways: (a) the mechanism does not embody formally valid reasoning (PONENS and LACY), or (b) it treats something as evidence when it should not (PROPER and IMPROPER). Turri’s cases rely on this assumption: there is a single mechanism at work in both cases (namely, the mechanism that forms beliefs) and it can function well or function poorly.

McCain avoids Turri’s counterexamples by reifying our belief forming mechanisms into Good Belief Forming Mechanisms and Bad Belief Forming Mechanisms.[7] He posits a rational belief formation mechanism (the one that uses modus ponens, and believing on the evidence) and an irrational belief formation mechanism (the one that engages in wishful thinking, or uses inference rule X). Once we reify belief forming mechanisms in this way, we achieve the intuitive results. We can distinguish PONENS and LACY because they are using different belief forming mechanisms. Believing on the evidence is one belief formation mechanism; wishful thinking is another.

But this is implausible. For psychological and philosophical reasons, wishful thinking and bad inferences are best understood as malfunctions of the same mechanism. It is psychologically implausible to posit that we operate with distinct, non-overlapping belief forming mechanisms, one set of which is “good” and the other set of which is “bad.” The belief forming mechanisms imagined here are causal mechanisms. Even if a mechanism is highly reliable, it is empirically unlikely that it will function perfectly all the time. So even if we can reify the “good” processes and the “bad” processes, it is still implausible to think the good processes will function perfectly. In those cases where the good processes malfunction, Turri’s counterexamples will reemerge. Secondly, we have a disposition to form beliefs. Sometimes that disposition embodies good reasoning. Sometimes it embodies bad reasoning. If we accepted the idea that there are “wishful thinking” mechanisms and “rational” mechanisms, it would be hard to see the output of both mechanisms as beliefs. Consider wishful thinking. We don’t straightforwardly, clear-eyedly, believe something because we want it to be true.[8] Wishful thinking is subtle. It can involve a manipulation of S’s assessment of the exact evidential support that E lends to p. And if this is the case, then it is not clear how IB-R can distinguish wishful basing from well-basing.

To illustrate, consider the Racist Employer, Chet.[9]

CHET

Chet is reviewing applicants for a job, and he forms the belief that Trayvon has a poor file. In fact, Trayvon does have a poor file. Trayvon did not attend a prestigious school, and he has minimal experience. However, had the applicant’s name been Caucasian, Chet would have evaluated the resume differently.[10] Chet is also deeply unaware of his racism. He has no beliefs that he could call to mind that are overtly racist. He is just slogging through a stack of a hundred applicants, trying to find the best candidate.

In this scenario, Chet correctly believes that the Trayvon’s resume is subpar. Considerations of race are playing a role in his belief formation, but not as a premise. His racism is not in the set of beliefs that he has. Rather, his racism is a function of how he interprets his evidence, and which conclusions he draws from that evidence. Even though the racist employer believes p because of evidence that in fact supports his conclusions, he is not reasoning well. It is only when we look at the way his belief forming mechanisms behave in total that we can see the pernicious racism that drives his belief formation. It is only when we examine (counterfactually) how he would reason with Caucasian applicants that we can see his poor reasoning. Chet has a bias that makes him go wrong, and he makes a mistake that corrects the problem. But multiplying mishaps shouldn’t give Chet doxastic justification.

The Direct Problem

One final concern about IB-R. Evidentialism and the interventionist model are not intuitive partners. In what follows, I argue that E could not be a direct cause of S’s belief that p. If IB-R is true and the argument is sound, this argument shows that no belief is or could be well-based. Let E be the set of non-factive mental states that propositionally supports p, and Bp be the belief that p. Let t be the time at which S forms Bp. Let M be the seeming to S that p, and let D be the disposition to form the belief that p when S has M. In order to be a direct cause, a change in the first variable must result in a change in the final variable, when other variables are fixed.[11] Now I will show that fixing D results in E having no causal influence on Bp.

First, E does not immediately cause Bp. If it does cause Bp it does so mediately, through what McCain calls a directed path. This is obvious because one could have evidence that propositionally justifies Bp and yet fail to believe p. This is true for beliefs I haven’t considered. It is true for cases where I am deliberating about whether to form the belief (but haven’t yet formed it). It is true for all inferential beliefs. E is not sufficient to cause Bp.[12] Something else must also be at play. This could be many different things. It could be my attention, it could be my judgment that E supports p, it could be some kind of psychological compulsion within me. But something else is required in order for the belief to come about.

On the interventionist model, this need not be a problem. E could directly cause Bp by directly causing something that directly causes Bp. The result would be a directed path. But E doesn’t directly cause something that directly causes Bp. We can see this because E can support an infinite number of beliefs, but rather than continuously forming infinite sets of beliefs, I am able to deliberate about one matter, and then set it aside. E does not cause me to open deliberation; something else is responsible. I will use the variable X to denote that thing, whatever it is, that causes me to open deliberation. Then, X and E together give rise to M, it seeming to me that p. And M activates D, which causes Bp.

Now that we have this framework in mind, let’s perform interventions. Let V = {E, X, S, D}. Fix X, M and D at 0. Given what we said above, intervening on E will have no effect on Bp. So E is not a direct cause of Bp. Now suppose we fix D and M at 1. Then it doesn’t matter how we intervene on E, Bp will obtain. Why? Because D and M are necessary and sufficient for causing Bp. It is directly analogous to McCain’s spark example:

[The faulty wiring] is not a direct cause of [the fire] because we cannot hold [the other variables] fixed at some values and manipulate the value of [the fire] by changing the value of [the faulty wiring]. . . . The only way that the short-circuiting of the electrical wiring can affect whether or not there is a fire in this case is by affecting whether there is a spark in the room. Once the presence or absence of the spark is held fixed, manipulating whether the electrical wiring short circuits cannot change whether the fire occurs. So the short circuiting of the electrical wiring is not a direct cause of the fire.[13]

Similarly, once we hold fixed that it seems to S that p, and that S has the disposition to form Bp on the basis of this seeming, the result (Bp) is determined. So E is not a direct cause of Bp.

I can see a way around the Direct Problem, but not if we hold on to McCain’s evidentialism. A reliabilist could say that a belief is well-based if it is caused by a reliable belief forming mechanism. It would effectively remove the extra step of what supplied by X above. Once we remove X from our intervention, we can show that the belief forming mechanisms are the direct cause of the belief. The interventionist model seems to be a very promising strategy to tackle deviant causation. But I’m not sure that IB-R is the best version of it.

[Conclusion

McCain presents a thorough, interesting, and rich discussion of the evidentialism and the basing relation. Despite his interesting proposal, the basing relation is a stubborn problem. Infection and Directness problems prevent it from being a satisfying account.

Works Cited

Bennett, Jonathan. (1990). “Why Is Belief Involuntary?” Analysis 50 (2).

Evans, Ian. (2013). “The Problem of the Basing Relation.” Synthese 190: 2943–47.

Davidson, Donald. (1963). “Actions, Reasons, and Causes.” Journal of Philosophy 60 (23): 685–700.

Flowerree, A. K. (2016). “Agency of Belief and Intention.” Synthese: Special Issue on Doxastic Agency, Heinrich Wansing and Andrea Kruse, ed.

Hieronymi, Pamela. (2009). “Believing at Will.” Canadian Journal of Philosophy 39: 149–87.

Lehrer, Keith. (1971). “How Reasons Give Us Knowledge, or the Case of the Gypsy Lawyer.” Journal of Philosophy 68 (10): 311–13.

McCain, Kevin. (2014). Evidentialism and Epistemic Justification. Routledge University Press.

Street, Sharon. (2011). “Evolution and the Normativity of Epistemic Reasons.” Canadian Journal of Philosophy 35, on Belief and Agency, ed. David Hunter, 2011, 213–48.

Shah, Nishi. (2003). “How Truth Governs Belief.” Philosophical Review 112, no. 4: 447–82.

Shah, Nishi, and J. David Velleman. (2005). “Doxastic Deliberation.” Philosophical Review 114, no. 4: 497–534.

Turri, John. (2011). “Believing for a Reason.” Erkenntnis 74 (3): 383–97.

Williams, Bernard. (1970). “Deciding to Believe.” In Problems of the Self. Cambridge, Cambridge University Press.

Wedgwood, Ralph. (2002). “The Aim Of Belief.” Nous 36, no. s16: 267–97.

Velleman, J. David. (2009). The Possibility of Practical Reason. Scholarly Publishing Office, University of Michigan Library.

[1] McCain, 2014.

[2] This example is modified from Flowerree (2016).

[3] I draw this conclusion from the official formulation of IB-R given on McCain, 90. There is a wrinkle to this interpretation which I will discuss in a moment.

[4] Indeed, the problem McCain raises for the doxastic view of basing seems to be a problem that will arise for any theory of doxastic justification (except perhaps coherentism). Justification is intuitively understood as chainlike. Either it will terminate in an unjustified belief, or it will result in an infinite regress.

[5] McCain, 96. “Shirley’s belief that r can be well-founded only if it is based on her evidence for r, which includes her beliefs that q and that q entails r, and those beliefs are themselves well-founded” (emphasis mine).

[6] See Turri (2011) for an explanation of the Orthodox view and an argument against it.

[7] I discuss belief formation in this section, but the same could be said of sustaining belief.

[8] Many philosophers accept that no mental state meeting this profile could count as a belief. See Bennett (1990), Hieronymi (2009), Street (2011), Shah (2003), Shah and Velleman (2005), Velleman (2009), Williams (1970), and Wedgwood (2002).

[9] This example is modified from Flowerree (2016). There, I argue against a doxastic conception of the basing relation. I have modified the example to draw out problems with IB-R. I consider implicit bias to be a cousin to wishful thinking.

[10] Numerous psychology studies have shown that this happens routinely. E.g., Steinpreis, Anders and Ritzke (1999).

[11] “There must be a possible intervention on the first variable that will result in a change in the value of the second when all other variables in the set are held fixed at some value.” McCain, 90.

[12] And it’s a good thing, otherwise we would have infinite causal chains manufacturing trivial beliefs slowing down our cognitive processing.

[13] McCain, 90.

  • Kevin McCain

    Kevin McCain

    Reply

    A Clean Bill of Health for the Interventionist Account of the Basing Relation

    In Evidentialism and Epistemic Justification I put forward an account of the basing relation (the relation that is necessary for moving from being justified in believing that p to justifiedly believing that p). Making use of an interventionist understanding of various causal relations I attempt to construct a causal account of the basing relation that yields the intuitively correct results while avoiding the problems that plague other causal accounts. A. K. Flowerree doesn’t think my interventionist account of the basing relation (hereafter “IB-R”) works. She raises two interesting objections to IB-R in her careful discussion. Surprisingly, Flowerree insists that the problems she sees for IB-R aren’t problems with the interventionist mechanics that I make use of, but rather with my commitment to evidentialism. Fortunately (from my perspective as an evidentialist and proponent of IB-R, anyway), there are subtle flaws in Flowerree’s arguments against IB-R. Let’s begin by seeing where the “Infection Problem” goes astray.

    Flowerree’s first objection to IB-R is what she calls the “Infection Problem.” It comes in two varieties: local and global. It’s worth taking a brief look at both of these beginning with the Local Infection Problem. After considering an “extra” condition I append to the formal presentation of IB-R, Flowerree claims that IB-R leads to regress problems. Both of the regresses associated with the Local Infection Problem arise because Flowerree takes my extra condition on IB-R to be that “a belief can only be well based if it is well based on non-factive mental states that are themselves well based.” Adding such a condition to IB-R would indeed have the effect that Flowerree claims because it would require every appropriately based belief to be based on another appropriately based belief, which would have to be appropriately based on another belief, and so on.

    Flowerree is correct that the condition she describes would straightforwardly lead to regresses, but this isn’t the condition that I include in IB-R. While Flowerree correctly notes that I mention the extra condition for IB-R when considering a case where the subject’s belief is based upon other beliefs, in several places I make it clear that a belief doesn’t have to be based on other beliefs in order to be appropriately based and justified. A belief can be appropriately based on experiences as well. But, experiences aren’t the sort of thing that are appropriately (or inappropriately) based on other mental states. So, the condition that I add to IB-R should be understood as the idea that a belief can only be appropriately based on the evidence when any belief that it is based on is itself appropriately based on supporting evidence. Notice this doesn’t require that a belief be based on other beliefs, i.e., it leaves open the possibility that a belief may be appropriately based on justifying experiences. Once this point is made clear both regresses Flowerree describes disappear. Hence, IB-R can be quickly cleared of Local Infection.

    Now let’s explore the Global Infection Problem. Recall, Flowerree illustrates this purported problem with CHET:

    Chet is reviewing applicants for a job, and he forms the belief that Trayvon has a poor file. In fact, Trayvon does have a poor file. Trayvon did not attend a prestigious school, and he has minimal experience. However, had the applicant’s name been Caucasian, Chet would have evaluated the resume differently. Chet is also deeply unaware of his racism. He has no beliefs that he could call to mind that are overtly racist. He is just slogging through a stack of a hundred applicants, trying to find the best candidate.

    Flowerree rightly notes that Chet’s belief in this case isn’t appropriately based on his evidence, and so isn’t justified. As she explains, “It is only when we examine (counterfactually) how he would reason with Caucasian applicants that we can see his poor reasoning. Chet has a bias that makes him go wrong, and he makes a mistake that corrects the problem. But multiplying mishaps shouldn’t give Chet doxastic justification.” Flowerree thinks there is a problem for IB-R here though. The reason she thinks IB-R has a problem is that she believes it yields the result that Chet’s belief is appropriately based on the evidence. Does IB-R yield this result though? No, it does not.

    Flowerree informs us that had the applicant’s name been Caucasian, then Chet wouldn’t have the belief that the applicant has a poor file. This is so given the exact same evidence that Chet has with respect to Trayvon’s file. The problem she notes is that Chet has a racist bias that is causing his belief. IB-R is perfectly able to handle this sort of case. Recall, that IB-R requires that in order for Chet’s belief to be appropriately based on the evidence it must not be the case that intervening to set the values of all direct causes of his believing that Trayvon has a poor file, other than his evidence, to 0 (in this case removing them) will result in Chet’s failing to believe that Trayvon has a poor file when Chet’s evidence is held fixed. This condition is clearly not met in Flowerree’s example. If Chet’s racist bias were removed while his evidence is held fixed, he wouldn’t believe that Trayvon’s file is a poor one. As she says, it is “the pernicious racism that drives [Chet’s] belief formation.” Additionally, as we noted above, Flowerree tells us that Chet wouldn’t form his belief about the applicant’s file on the basis of the evidence he has if the applicant had a Caucasian name. Consequently, IB-R yields the result that Chet’s belief that Trayvon’s file is poor isn’t appropriately based on his evidence. Global Infection is easily treated.

    The final problem that Flowerree presents is one she dubs the “Direct Problem.” The gist of this problem is that one’s beliefs are never based on one’s evidence, but instead they are always based on seemings that arise from the evidence and whatever causes one to deliberate. The upshot is that IB-R can never be satisfied by any of our beliefs, so accepting it would entail that none of our beliefs are appropriately based on our evidence. And so, IB-R would entail that none of our beliefs are justified. This would be a very troubling consequence, one that would be grounds for abandoning the account of the basing relation that entails it. Happily, IB-R doesn’t have this consequence.

    To begin with why should we accept Flowerree’s picture of the role of evidence and seemings? Admittedly, fans of phenomenal conservatism are apt to think that she is correct in thinking all of our beliefs are directly based on how things seem to us. Nonetheless, many epistemologists disagree with this claim concerning the role of seemings.1 In fact, some are skeptical that there even are seemings.2 While I agree with Flowerree that seemings do exist, and I agree that they can play a role in justification, I don’t see why we should accept her claims about all beliefs being directly based on seemings and not other evidence. Setting this aside though, say that we grant Flowerree’s claim that all beliefs are directly based on seemings. Does this pose a problem for IB-R? No, it doesn’t. The reason for this is that it is plausible to think that seemings are themselves evidence. After all, supporters of phenomenal conservatism (some of whom self-identify as evidentialists) hold that seemings provide propositional justification, and some explicitly put this in terms of seemings being evidence.3 Thus, at most accepting Flowerree’s claim would commit us to allowing that seemings can be evidence. With respect to my own theory, Explanationist Evidentialism, this would amount to embracing an element of phenomenal conservatism. I don’t think this is much of a cost. In fact, I’ve begun to explore the pairing of Explanationist Evidentialism and phenomenal conservatism for other reasons.4

    I appreciate Flowerree’s clear and helpful discussion of my interventionist account of the basing relation. Nevertheless, I think her objections to IB-R miss their mark. The interventionist account of the basing relation is healthy and thriving.5


    1. A cursory examination of the literature surrounding self-defeat arguments in support of phenomenal conservatism reveals this.

    2. Ted Poston has expressed his skepticism concerning the existence of seemings as a genuine kind of mental state in personal correspondences.

    3. See Blake McAllister, “Common Sense Epistemology: A Defense of Seemings as Evidence,” PhD diss., Baylor University, 2016.

    4. Kevin McCain, “Explanationist Aid for Phenomenal Conservatism,” Synthese, forthcoming.

    5. Thanks to Matt Frise for helpful discussion.

    • Amy Flowerree

      Amy Flowerree

      Reply

      Reply to the reply to the reply :)

      Many thanks to Kevin for his reply to my reply. I am not persuaded by all of his points. Here I will say why I am not persuaded, and raise a number of questions that I think bring out our disagreement.

      Local Infection:
      The structure of Local Infection was this: I gave a belief that satisfies IB-R, but that is intuitively not doxastically justified (Ann). Kevin doesn’t directly address the case, but I take it that he agrees with me that Ann is not doxastically justified. I listed one possible response, to amend IB-R. Kevin seems to agree with this response, but I am unclear exactly what his proposal amounts to. My first question for Kevin is this:
      (1) Is he adding a condition to IB-R? and if so, what is it?

      He accepts that sometimes beliefs must be well-based on beliefs that are themselves well-based, and he makes use of this feature to get around a counterexample. But, he adds, not all beliefs must stand in the relation (due to the lurking vicious regress). And that some beliefs are well-based on experiences, which are not the sort of thing that could be well-based on beliefs. This seems to amount to a version of foundationalism. I have many thoughts in response to this, but before I trace out all the possible ways Kevin could develop IB-R, I will just ask another question:
      (2) Are there things other than experiences that serve as a foundation for our evidence? And are all experiences included in E?

      I can see how appealing to foundationalism will get around the second regress I propose. But I do not see how it avoids the conceptual regress. Recall, the first regress was this: IB-R is supposed to be an account of what it is for a belief to be well-based. But if we include a condition in IB-R that says that every element in E must be either well-based (or requires no further basing), then we have not analyzed the basing relationship. We are analyzing well-basing in terms of well-basing. We’ve included the term in the explanandum and the explanans. This means that, at best, IB-R is an analysis of the transmission of well-foundedness, not well-foundedness itself. I don’t see how Kevin has addressed this issue.
      And finally, I do not see how Kevin’s response answers the case of Ann. This ties back into question (1), and it will depend exactly how Kevin wants to reformulate IB-R. Ann’s beliefs are causally based on the experience of being struck by lightning. Does that experience, e, belong in E? if so, then the modification still predicts that Ann is doxastically justified (contra our intuitions). And if the e does not belong in E, then we need some further account of which experiences belong in E and which do not. I would like to hear more about how Kevin would deal with this case.

      Global Infection
      Kevin’s reply is perplexing to me. The issue I raise is whether IB-R can capture the subtly required for implicit bias cases. In this case, E causes Chet’s belief that p, and E does propositionally justify the belief that p. And the structure of the case, similar to the cases Turri raises, is one where the causal mechanism from E to the belief that p is consistent with IB-R. But – and this is my central contention in the case – that causal mechanism is sometimes wishful and sometimes not. Its not two different mechanisms. Its one mechanism that sometimes works well and sometimes doesn’t. And in this case, it did not work well, but Chet got lucky, so his belief is still propositionally justified. But it shouldn’t count as doxastically justified.

      The central contention between Kevin and I is whether wishful thinking is its own causal mechanism. I think it is not. I think we have belief forming mechanisms, and sometimes they are wishful and sometimes they are not. I gave some reasons for this in my first reply. But Kevin’s response is to say “If Chet’s racist bias were removed while his evidence is held fixed, he wouldn’t believe that Trayvon’s file is a poor one.” But this is exactly what is in contention.

      In order to get the case right, here is the counterfactual that needs to be evaluated: Chet evaluating a resume identical to Trayvon’s, but with a Caucasian name. But how is IB-R going to make that counterfactual judgment? The only counterfactuals IB-R examines is (a) the world is the same, except E is set to 0; and (b) E is present, but all other causal factors are set to 0. What are we imagining here? That Chet has Trayvon’s resume, but with no name? that Chet has a sheet of paper with only Trayvon’s name, and no resume? It just doesn’t seem like IR-B is going to capture the right counterfactuals. The right counterfactuals are only captured by a global assessment of the agent. Hence, I called it global infection.
      So that leads to this question:
      (3) How does IR-B exclude wishful thinking from the case of Chet?

      Directness
      It seems to me that Kevin’s reply misses the mark by attributing to me a position that I do not hold. I use the term ‘seeming’ as a placeholder for whatever it is in the causal chain that activates the disposition to believe p. I am not committed to this seeming having phenomenal content or epistemic significance. Instead, it is an important part of the causal story. I gave several arguments for why I think that E cannot directly cause beliefs, and why there must be something playing the causal role of activating the disposition to believe p. It doesn’t really matter what we call it, the important thing is that (a) a plausible account of belief formation needs it; and (b) it is not evidence (again, I think of it as causal, rather than epistemic in nature).

      Kevin’s second reply is to deny (b), and say that such seemings could be evidence. To me, that seems like a reductio of the view (if the view commits you to phenomenal conservatism, then you should reject it). But I wont go into that here, since I think there is an even more pressing problem for Kevin: seemings will swamp all other evidence. Since the seemings are necessary and sufficient for activating the disposition to believe p (recall, a seeming is picked out by its functional role of activating the disposition to believe p; so this is true by definition), then the only thing that will count as an actual and direct cause of the belief that p is the seeming. The result would be that if seemings are evidence, then they are the only sort of evidence that we can base our beliefs on, according to IB-R. The rest of E is irrelevant for the basing relation.

      I thank Kevin for his thoughtful replies, but I’m afraid I have only more questions!

    • Kevin McCain

      Kevin McCain

      Reply

      Response to Flowerree

      Hi Amy,

      Thanks for the reply to the reply to the reply! I appreciate you following up on these issues. In this brief response I’ll try to answer your helpful questions.

      (1) Is he adding a condition to IB-R? and if so, what is it?
      Short answer:
      No. 🙂

      More complete answer:
      Due to the length constraints on my initial replies I couldn’t go into a discussion of your ANN case. So, I appreciate the opportunity to discuss it here. I agree with you that Ann’s belief that q isn’t doxastically justified. However, I don’t think that anything needs to be added to IB-R in order to accommodate this fact. As the case is described, Ann’s belief that q isn’t propositionally justified (we disagree on this point as you mention that you think it is). In ANN the lightning strike causes a lottery ticket to appear at Ann’s feet and for her to believe: p (“there is a lottery ticket at my feet”) and if p, then q (if “there is a lottery ticket at my feet,” then “the lottery is rigged in my favor”). Since there actually is a lottery ticket at Ann’s feet, it’s plausible that she can see it and has propositional justification for believing that p. But, when it comes to if p, then q it doesn’t seem that she has any reason to accept this. A common view of inferential justification, which I accept, is that in order for beliefs to provide justification they must be justified (or, more carefully, they must have positive epistemic status). As a result, Ann’s belief that q isn’t propositionally justified because her belief that if p, then q isn’t justified. Since q isn’t propositionally justified for Ann, her belief that q can’t be doxastically justified.

      (2) Are there things other than experiences that serve as a foundation for our evidence? And are all experiences included in E?

      I’m not quite sure that I understand the first question here—I don’t typically think of something serving as a foundation for our evidence. That being said, I think that the sort of foundational evidence needed to end avoid regresses will have to be something that has at least some presumption (in Chisholm’s sense) in its favor without gaining that from other beliefs. Experiences can have this. At times I go back and forth on the issue of whether beliefs themselves can sometimes have this feature (that is, I’m not ready to dismiss epistemic conservatism just yet), but this can be set aside here. Let’s assume that only experiences can serve this function.

      Concerning the second question, the answer is “no”. On pg. 11, I point out that a requirement for non-factive mental states to be evidence is that they have to have content with a “mind-to-world direction of fit” (in Anscombe’s terms); (in Pryor’s terms) they must have “phenomenal force”—that is, the experiences that are included in E “represent content to us in such a way that it “feels as if” the content is true and we can tell that it is true.”

      With respect to your question about ANN, the lightning strike doesn’t get included in E. (More carefully, it doesn’t count as part of E with respect to p, if p, then q, or q. It might count as evidence for believing that Ann was struck by lightning.)

      With respect to your claim that “We are analyzing well-basing in terms of well-basing.” I disagree. IB-R says that a belief is well-founded (appropriately based on the evidence that propositionally justifies the belief) when the evidence that propositionally justifies the belief plays certain causal roles in one’s having that belief. That seems to be an analysis in terms of causal relations. Now, one might say, “well, when the evidence includes other beliefs well-foundedness (well-basing) gets included in the picture.” While that’s true, it doesn’t change the fact that the central analysis doesn’t use well-foundedness/well-basing to analyze well-foundedness/well-basing.

      (3) How does IR-B exclude wishful thinking from the case of Chet?

      You helpfully explain your thoughts on the CHET case: “But – and this is my central contention in the case – that causal mechanism is sometimes wishful and sometimes not. Its not two different mechanisms. Its one mechanism that sometimes works well and sometimes doesn’t. And in this case, it did not work well, but Chet got lucky, so his belief is still propositionally justified. But it shouldn’t count as doxastically justified… The only counterfactuals IB-R examines is (a) the world is the same, except E is set to 0; and (b) E is present, but all other causal factors are set to 0. What are we imagining here? That Chet has Trayvon’s resume, but with no name? that Chet has a sheet of paper with only Trayvon’s name, and no resume? It just doesn’t seem like IR-B is going to capture the right counterfactuals.”

      I fail to see a problem for IB-R here. I don’t see why we should accept that wishful thinking, or in this case Chet’s bias, shouldn’t count as a separate mechanism. However, even if we grant that there is only one mechanism as you claim, there doesn’t appear to be a problem for IB-R. As I explain on pg. 92, “Although for simplicity we will tend to follow a binary valuation schema, we should not feel limited to such a coarse-grained schema. One way in which we could adopt a more fine-grained valuation schema is to adopt a schema for these variables that assigns values based on the strength of the mental state/mechanism or influence in question. So, for example, we might want to assign values to perceptual states in terms of their degree of vividness; the more vivid the perceptual state the higher the value. We could develop similar fine-grained schemas for any other variables we need to include in a given variable set.” It is for the sake of simplicity I put IB-R in terms of just setting causal factor variables to 0. So, it is possible to capture many more counterfactuals than those the basic picture of IB-R suggests. In particular we can evaluate the key counterfactual of what would happen if we keep everything else as it is in the case, but change the name on the resume from Trayvon to a Caucasian name. When we intervene in this way (which is, admittedly more complicated than simply setting various influences to 0 when we insist that Chet’s racial bias isn’t a mechanism of its own) we find that it affects Chet’s belief. Since, as you say, “However, had the applicant’s name been Caucasian, Chet would have evaluated the resume differently”, IB-R yields the correct result here.

      Concerning directness, I appreciate you helping me better understand what you meant by “seeming”. Since you had originally said “Let M be the seeming to S that p, and let D be the disposition to form the belief that p when S has M”, I took it that seemings on your view had content (in the case described I took it to have the content p because it is the seeming “that p”).

      I am, however, still a bit confused by the purported problem here. You say, “I gave several arguments for why I think that E cannot directly cause beliefs, and why there must be something playing the causal role of activating the disposition to believe p.” You also say, “I use the term ‘seeming’ as a placeholder for whatever it is in the causal chain that activates the disposition to believe p.” Finally, you say, “Since the seemings are necessary and sufficient for activating the disposition to believe p (recall, a seeming is picked out by its functional role of activating the disposition to believe p; so this is true by definition), then the only thing that will count as an actual and direct cause of the belief that p is the seeming. The result would be that if seemings are evidence, then they are the only sort of evidence that we can base our beliefs on, according to IB-R.”

      If I have understood you correctly, on your view seemings are just whatever activates the disposition to believe that p. As a result, you maintain that seemings are the only actual and direct causes of the belief that p. Consequently, according to your worry; evidence won’t really count as causes of beliefs. Hence, there’s a problem for an evidentialist view (like mine) because evidence won’t count as direct, actual causes of beliefs, so no beliefs will be doxastically justified.

      I see no problem here because the troublesome conclusion doesn’t follow from the premises. If seemings are defined functionally as simply “whatever it is in the causal chain that activates the disposition to believe p”, why can’t non-factive mental states that are part of one’s evidence be seemings? For instance, when I have a visual experience as of seeing a tree, it seems that that mental state is evidence. It also seems that it could be “whatever it is in the causal chain that activates the disposition to believe” there’s a tree. It seems like the evidence is a seeming (in your sense), and it is a direct, actual cause of the belief. As a result, it is plausible that on IB-R the belief is doxastically justified.

      I don’t see a problem of swamping that you are concerned about in connection to the directness issue either. Once we accept that seemings are just whatever activates the disposition to believe that p, it straightforwardly follows on IB-R that the parts of E that constitute the seeming (i.e. are the direct and actual causes of the belief that p) are the only evidence that matter for basing. This is something that is already acknowledged on pg. 96 when I discuss chains of reasoning.

      Thanks again for the great discussion both in your initial essay and your reply. I hope this helps address your questions.

      -Kevin

Andrew Moon and Pamela Robinson

Response

Why Explanationists Shouldn’t Make Evidential Fit Dispositional

1. Introduction

Kevin McCain’s Evidentialism and Epistemic Justification is the most thorough defense of evidentialism to date. In this work, McCain proposes insightful new theses to fill in underdeveloped parts of evidentialism. One of these new theses is an explanationist account of evidential fit that appeals to dispositional properties. We argue that this explanationist account faces counterexamples, and that, more generally, explanationists should not understand evidential fit in terms of dispositional properties.

2. Explanationist Accounts of Evidential Fit

Suppose you have the sort of visual experience one typically has when seeing a red object. Plausibly, this visual experience evidentially supports your believing that something is red. But what makes this proposition fit that evidence?

McCain provides the following theory:

Explanationist Fit (EF)p fits S’s evidence, e, at t IFF either p is part of the best explanation available to S at t for why S has e or p is available to S as a logical consequence of the best explanation available to S at t for why S has e. (65)

The first disjunct of the analysans explains why the visual experience supports believing that something is red. The proposition <something is red> is ‘part of the best explanation available to you’ for why you have that experience. The second disjunct explains why the visual experience supports believing any logical consequence of the best explanation available, such as <something is there>.

Crucial to understanding EF is ‘p is part of the best explanation available to S for why S has e’. According to McCain,

Available as part of the best explanation: S has p available as part of the best explanation for why S has e (at t) IFF: (at t) S has the concepts required to understand p and S is disposed to have a seeming that p is part of the best answer to the question “why does S have e?” on the basis of reflection alone. (67)

Note that according to this definition, ‘S has p available as part of the best explanation for why S has e’ does not entail that p is part of the best, or is even part of any, explanation for why S has e. This is because S’s being disposed to have a seeming that p is part of the best answer to some question does not entail that p really is part of the best, or is part of any, explanation of that question.1

So, a more transparent way of writing EF references dispositions rather than explanations, for the dispositions are what matter:

Explanationist Fit (EF)p fits S’s evidence, e, at t IFF S has the concepts required to understand p and, either: (i) S is disposed to have a seeming that p is part of the best answer to the question “why does S have e?” on the basis of reflection alone, or (ii) S is disposed to have a seeming that BA is the best answer to the question “why does S have e?” on the basis of reflection alone, and p is a logical consequence of BA.2

Suppose you have the concepts required to understand the proposition <something is red>. Presumably, if you were to reflect on the question “Why do I have this visual experience?”, it would seem to you that <something is red> is part of the best answer to that question. So, McCain’s theory correctly predicts that your belief that <something is red> fits your perceptual evidence.

It is easy to miss that EF is a highly subjectivist theory. In this context, roughly, a theory is subjectivist to the degree that evidential fit depends on the subject’s inclination to take p as part of the best explanation for why S has e. In the rest of this section, we will compare EF to three related explanationist theories of evidential fit, each increasing in its degree of subjectivity.

Let ‘BE’ stand for ‘the best explanation at t for why S has e.’ The least subjectivist version of explanationist fit is:

Strong Objectivist EFp fits S’s evidence, e, at t IFF either (i) p is part of BE, or (ii) p is a logical consequence of BE.

On this view, a proposition fits your evidence simply when it’s part of (or entailed by) the best explanation for why you have the evidence. Note that Strong Objectivist EF does not require that the explanation is available to you. How inclined you are to take p to be part of the best explanation for why you have e makes no difference to evidential fit.

The following is a more subjectivist view:

Moderate Objectivist EF: p fits S’s evidence, e, at t IFF S has the concepts required to understand p and, either: (i) p is part of BE and S is disposed to have a seeming that p is part of the best answer to the question of why S has e (on the basis of reflection alone), or (ii) p is a logical consequence of BE and S is disposed to have a seeming that BE is the best answer to the question of why S has e (on the basis of reflection alone).

On this version of evidential fit, a proposition fits your evidence when it’s part of (or entailed by) the best explanation for why you have the evidence, and you’re disposed to take it to be part of this best explanation, where this just means that you’re disposed to have a seeming that it’s part of the best answer to the question of why you have that evidence. Moderate Objectivist EF is more subjectivist than Strong Objectivist EF since it requires that the explanation be available to you. On the other hand, Moderate Objectivist EF is less subjectivist than EF, which doesn’t actually require that p be part of the best explanation.

The next account is even more subjectivist and requires some set up. There is some set of explanations, {E1, …, En}, such that, for every member of {E1, …, En}, S is disposed, on the basis of reflection alone, to have a seeming that it is an answer to the question of why S has e. This is the set of explanations that are ‘available’ to S in McCain’s sense. Note that {E1, …, En} might not include all the explanations for why S has e; there might be explanations that S is not disposed to consider. Now let ‘BE*’ pick out the best of those explanations {E1, …, En}. The next account is:

Moderate Subjectivist EF: p fits S’s evidence, e, at t IFF S has the concepts required to understand p and, either: (i) p is part of BE* or (ii) p is a logical consequence of BE*.3

This view is more subjectivist than both Strong Objectivist EF and Moderate Objectivist EF since it doesn’t require that p be part of the actual best explanation for why S has e. However, it’s not as subjectivist as EF since it requires that p be part of (or a logical consequence of) the actual best of {E1, . . . , En}. EF doesn’t require that p actually be part of an explanation, of the restricted set of explanations {E1, . . . , En}, or of any other interesting set; all that’s required is that S be disposed to have a seeming that p is part of the best answer to the question of why S has e.

By canvassing these different degrees of subjectivity, we can see just how strongly subjectivist EF is. A more appropriate label for EF might be ‘Strong Subjectivist EF,’ but we will stick with ‘EF.’4

3. Problems with McCain’s Account of Evidential Fit

In 3.1, we give counterexamples that exploit the fact that, according to EF, p isn’t actually required to be a part of the best, or even an, explanation of e. In 3.2, we give a counterexample that exploits the fact that, according to subjectivist explanationist accounts, the explanation (or apparent explanation) must be dispositionally available to S. In 3.3, we show how explanationist views that appeal to dispositional accounts of available explanations face a new evil demon problem.

3.1 Weird Dispositions Counterexamples

Paranoid: Unbeknownst to Bucky, a demon has tampered with him and has given him a certain complex disposition. Now, when he has the visual experience of his friend Steve smiling, he is disposed to have a seeming that <Steve wants to kill me> is part of the best answer to the question of why he has this visual experience. Bucky has no memory or past experience of Steve ever posing a threat. They are—or were—friendly acquaintances.

According to EF, <Steve wants to kill me> fits Bucky’s visual experience of Steve smiling. But intuitively, this is false. This visual experience might evidentially fit <Steve is smiling> or <Steve is happy>, but it does not fit <Steve wants to kill me>.

Consider another version of Paranoid, call it ‘Nonsensical’, in which the demon had instead caused Bucky to have the disposition to have a seeming that <grass is green> is part of the best answer to the question of why he has the visual experience of Steve smiling. EF has the counterintuitive result that <grass is green> fits Bucky’s evidence. These cases show that EF is too permissive. It allows anything that you’re disposed to take as part of the best explanation for why you have your evidence to count, even if it’s not a good explanation (as in Paranoid) or an explanation at all (as in Nonsensical).

3.2 Impoverished Dispositions Counterexample

McCain might be tempted to give up EF and retreat to Moderate Subjectivist EF or Moderate Objectivist EF, which require that p is part of an, or the best, explanation. However, the following case poses a problem even for these less subjectivist views.

Blue Sky: Suppose Carla lies resting in a field, closing her eyes and occasionally looking up at the sky. Nothing interesting is going on with the sky: it’s clear and blue whenever Carla happens to open her eyes. At t, Carla opens her eyes and has the mental state of being appeared to blue-skyly. At t Carla is disposed to have the seeming that <I opened my eyes> is the best answer to the question of why she is appeared to blue-skyly. At t, Carla is not disposed to have any other seemings about other answers, or parts of answers, to the question of why she is appeared to blue-skyly.

According to EF, <Carla opened her eyes> fits her evidence, that is, her mental state of being appeared to blue-skyly. This seems correct. But it also seems that propositions like <Carla looked at the sky> or <the sky is blue> fit her evidence. And according to EF they do not: they are not part of (or entailed by) the best explanation available to Carla.

This is also a problem for both Moderate Objectivist EF and Moderate Subjectivist EF. Plausibly, <Carla opened her eyes> is part of the best explanation for why Carla has her visual experience. So, since this explanation is available to Carla, <Carla opened her eyes> fits her evidence according to Moderate Objectivist EF and Moderate Subjectivist EF. However, since this is the only explanation (or only part of an explanation) available to Carla, no other proposition fits Carla’s evidence according to Moderate Subjectivist EF or Moderate Objectivist EF.5

Interestingly, Strong Objectivist EF seems to get the right result in all these cases. Plausibly, neither <Steve wants to kill me> nor <grass is green> are part of the best explanation for why Bucky has his evidence, so they do not fit his evidence according to Strong Objectivist EF. But <Steve is smiling> and <Steve is happy> are part of the best explanation for why Bucky has his evidence, so they fit it according to Strong Objectivist EF. Also, it’s plausible that <Carla opened her eyes>, <Carla looked at the sky>, and <the sky is blue> are all part of the best explanation for why Carla has her evidence, so they all fit her evidence according to Strong Objectivist EF.

3.3 The New Evil Demon Problem

The New Evil Demon Problem (NEDP) purports to show that certain properties are irrelevant to whether a belief is justified. For example, consider the view that a belief’s being reliably formed is relevant to whether it is justified. Now suppose that Smith’s belief seems justified and is formed by a reliable process. We can imagine a mental duplicate of Smith, Smith*, whose belief is not formed by a reliable process due to the machinations of a demon. Intuitively, Smith and Smith* are equally justified in their belief. It seems that any two individuals who are identical except for the mere difference of reliable belief production will be justificationally identical. This seems to show that reliable belief formation is not relevant to justification.6

We argue that a version of the NEDP shows that the dispositions McCain appeals to are irrelevant to justification. Consider Paranoid. Imagine a Bucky* who has the same memories, beliefs, and experiences as Bucky but who never had a demon tamper with him. He thereby does not have the paranoid disposition. When Bucky and Bucky* have qualitatively identical visual experiences at t of Steve smiling, it follows from EF that Bucky has evidence that fits <Steve wants to kill me> while Bucky* does not.7 Given evidentialism and EF, Bucky has more justification for believing <Steve wants to kill me> at t than Bucky* does. But this is counterintuitive. It seems that the two Buckys are justificationally alike.

Now consider the dispositional requirements of Moderate Subjective EF and Moderate Objective EF. Suppose Bucky** is just like Bucky*, but a demon has erased his normal human disposition to have a seeming that <Steve is smiling> is part of the best explanation for why he has the visual experience of Steve smiling. On either of the two theories, it’s plausible that Bucky* has evidence that fits <Steve is smiling>. However, Bucky** does not because he lacks the relevant disposition. This is counterintuitive. It seems that they are justificationally alike.

These instances of the NEDP show that the dispositions McCain appeals to are irrelevant to justification, just as more standard versions of the NEDP show that reliable belief production is irrelevant. It seems that any two individuals who are identical except for the mere difference of dispositional availability of explanations will be justificationally identical. Strong Objectivist EF is not subject to this version of the NEDP since it does not appeal to dispositions.

We have argued that there are counterexamples to EF. One might think that EF merely requires tweaking. But Moderate Subjectivist EF and Moderate Objectivist EF also face counterexamples. Furthermore, our discussion of the NEDP reveals that the underlying problem is the appeal to dispositions. Strong Objectivist EF, which does not appeal to them, faces neither the counterexamples nor the NEDP. We conclude that explanationists should not analyze evidential fit in terms of the dispositional availability of explanations.8

Works Cited

McCain, Kevin. Evidentialism and Epistemic Justification. New York: Routledge, 2014.

Moon, Andrew. “Three Forms of Internalism and the New Evil Demon Problem.” Episteme 9 (2012) 345–60.

Poston, Ted. Reason and Explanation: A Defense of Explanatory Coherentism. New York: Palgrave Macmillan, 2014.


  1. Assuming that answers to questions are propositions, our thought is that not every proposition can be an explanation for anything. For example, <Jupiter is the largest planet in the solar system> is not a poor explanation for the fact that 2 + 2 = 4; it’s not an explanation at all.

  2. It’s possible that the account is actually:

    p fits S’s evidence, e, at t IFF S has the concepts required to understand p and, either: S is disposed to have a seeming that p is part of the best answer to the question “why does S have e?” on the basis of reflection alone, or S is disposed to have a seeming that p is a logical consequence of the best answer to the question “why does S have e?” on the basis of reflection alone.

    The choice between the two versions depends on how McCain understands ‘p is available to S as a logical consequence of the best explanation available to S.’ We think this version is less charitable and will assume the more objectivist understanding in what follows, but our argument does not depend on this choice.

  3. Thanks to Liz Jackson, Dustin Crummett, and especially Philip Swenson for help formulating Moderate Subjectivist EF.

  4. Note that a theory of evidential fit need not fall under McCain’s general framework in order to be explanationist. For example, on Ted Poston’s (2014, 87–95) explanationist view ‘Ex-J’, “S has justification for believing p if and only if p is a member of a sufficiently virtuous explanatory system, E, and E is more virtuous than any competing system E” (87). According to Poston, evidence “resides in the virtues of the entirety of a system of beliefs” (93) and “[a] proposition is a member of an explanatory system by being a part of an explanans or part of an explanandum” (87). Now, it is not obvious how Ex-J can be converted into a theory of evidential fit, but note that nothing in his theory implies that p must explain why S has her evidence, or that S must be disposed to have a seeming that p is the best answer to why she has her evidence. So, Poston’s theory of evidential fit, when developed, may be very different from McCain’s.

  5. One option for the subjectivist about explanationist evidential fit would be to describe the evidence being explained in such a way that one would be more likely to be disposed have the ‘right kind’ of seeming about what best explains it. For example, a description of Carla’s evidence with an appropriate contrast—like the mental state of being appeared to blue-skyly rather than green-grassly—could not be best explained by the proposition that Carla opened her eyes. This fix would not save EF from weird disposition cases, but it would make versions like Moderate Subjectivist EF and Moderate Objectivist EF more plausible.

  6. For a definition and further explanation of the NEDP, see Moon, 348.

  7. To ensure that Bucky and Bucky* are identical with respect to memories, beliefs, and experiences at t, suppose that Bucky’s disposition is never manifest because he never reflects: even though he has the paranoid disposition at t, it doesn’t seem to him at t that Steve wants to kill him or that <Steve wants to kill me> is the best answer to the question of why he has his visual experience.

  8. Thanks to Kevin McCain, Ted Poston, and Philip Swenson for helpful discussion.

  • Kevin McCain

    Kevin McCain

    Reply

    The Role of Dispositions in Evidential Fit

    Dispositions play an important role in Explanationist Evidentialism. In particular, S must have certain dispositions in order for a given proposition to fit her evidence (be supported/justified by that evidence). Andrew Moon and Pamela Robinson (hereafter “M&R”) challenge this understanding of the role of dispositions in evidential fit. M&R begin by providing an excellent overview of several ways accounts of evidential fit similar to mine might be construed. After discussing these various conceptions of evidential fit M&R detail three purported counterexamples to any account that analyzes evidential fit in terms of explanations being dispositionally available to a subject. Despite their careful exploration of explanationist accounts of evidential fit and creative examples, M&R fail to produce a genuine counterexample to the account of evidential fit utilized in Explanationist Evidentialism.

    The place to start in responding to the challenges that M&R pose is exactly where they begin—getting clearer about how I construe evidential fit. M&R canvas several ways that one might understand evidential fit from an explanationist perspective. I am particularly grateful for their careful articulation of these alternative views of evidential fit and their close examination of my presentation of an account of evidential fit. M&R’s discussion has made me realize that I wasn’t sufficiently clear in my presentation of Explanationist Fit. My actual view of evidential fit is closest to the view that M&R call “Moderate Subjectivist EF”:

    Moderate Subjectivist EF: p fits S’s evidence, e, at t IFF S has the concepts required to understand p and, either: (i) p is part of BE* or (ii) p is a logical consequence of BE*.

    Aside from time indexing the right side of the biconditional like the left side, I would only add to Moderate Subjectivist EF the additional condition that p be available as part of (or a logical consequence of) BE*1 where this is understood in terms of S having the disposition to have a seeming that p is part of (or a logical consequence of) the best explanation of e.2 Essentially, I believe that in order for p to fit S’s evidence it must be part of an explanation that is available to S, that explanation must in fact be better than the other rival explanations available to S, and S must be disposed to have a seeming that p is part of, or entailed by, this best explanation. This is how Explanationist Fit (EF) should be understood.

    Now that my view of evidential fit has been clarified it’s time to turn toward M&R’s purported counterexamples. The importance of clarifying my account of evidential fit is readily apparent when we recall what M&R say about their first purported counterexample, it “exploit[s] the fact that, according to EF, p isn’t actually required to be part of the best, or even an, explanation of e.” As the previous paragraph makes clear, EF shouldn’t be understood in this way, so the first example that M&R describe doesn’t apply to it. In light of this I’ll focus the remainder of the discussion on responding to M&R’s other examples.

    Here’s M&R’s “Blue Sky”:

    Suppose Carla lies resting in a field, closing her eyes and occasionally looking up at the sky. Nothing interesting is going on with the sky: it’s clear and blue whenever Carla happens to open her eyes. At t, Carla opens her eyes and has the mental state of being appeared to blue-skyly. At t Carla is disposed to have the seeming that <I opened my eyes> is the best answer to the question of why she is appeared to blue-skyly. At t, Carla is not disposed to have any other seemings about other answers, or parts of answers, to the question of why she is appeared to blue-skyly.

    The problem here is supposed to be that “propositions like <Carla looked at the sky> or <the sky is blue> fit her evidence. And according to EF they do not: they are not part of (or entailed by) the best explanation available to Carla.” Is there a problem here though? It’s far from clear that we should think that these propositions actually fit Carla’s evidence. It seems that she is missing something very important—she can’t (at least at t) connect these propositions to her evidence. In a sense, Carla is like a student who has been presented with all the premises for a proof for p, but doesn’t really understand the proof, and so can’t see that p is the conclusion that should be drawn. P doesn’t fit the student’s evidence—after all, it seems that if the student were to suddenly believe that p without gaining a proper appreciation of the proof, her belief wouldn’t be justified. Similarly, if Carla were to suddenly believe <Carla looked at the sky> or <the sky is blue> without being in a position to appreciate the relation that these propositions bear to her evidence, i.e., without being disposed to recognize them as answers to why she has that evidence, her beliefs wouldn’t be justified. Hence, contra M&R, it seems that <Carla looked at the sky> and <the sky is blue> do not fit Carla’s evidence in this case.

    For the final purported counterexample M&R appeal to the New Evil Demon Problem (NEDP). They consider a couple different cases in order to make the point that the sort of dispositions that I maintain are required for an explanation to be available in the relevant sense are not necessary for justification. Recall, M&R’s Bucky*. Bucky* is a normal human who has visual experiences of Steve smiling and the disposition to believe that the best explanation of his experiences is that <Steve is smiling>. M&R offer the case of Bucky*’s demonworld counterpart as a counterexample:

    Suppose Bucky** is just like Bucky*, but a demon has erased his normal human disposition to have a seeming that <Steve is smiling> is part of the best explanation for why he has the visual experience of Steve smiling . . . it’s plausible that Bucky* has evidence that fits <Steve is smiling>. However, [according to EF] Bucky** does not because he lacks the relevant disposition. This is counterintuitive. It seems that they are justificationally alike.

    They go on to say, NEDP cases “show that the dispositions McCain appeals to are irrelevant to justification, just as more standard versions of the NEDP show that reliable belief production is irrelevant. It seems that any two individuals who are identical except for the mere difference of dispositional availability of explanations will be justificationally identical.”

    While M&R are correct that the standard NEDP cases expose a serious problem for reliabilism, they are mistaken when they claim that these cases show that dispositions don’t matter for justification. Consider a situation where an expert is teaching a novice how to identify a maple tree by its appearance. They both have the same visual experiences, however, when they see a maple only the expert has justification for believing that it is a maple. Why? Plausibly, it is because the expert has a disposition to recognize that <that’s a maple> is part of the best explanation of her visual experiences. As I explain in Evidentialism and Epistemic Justification (67), the expert has this disposition in virtue of having relevant information as background evidence. Similarly, a logic teacher and a student may be looking at the exact same premises, but only the former has justification for believing that p on the basis of these premises. Why? Again, it is plausible that the logic teacher is disposed to see how p follows from the premises, but the student isn’t. What cases like this help demonstrate is that M&R are mistaken in thinking that a demon could remove the dispositions that Bucky** has without affecting what fits his evidence. Such dispositions provide a necessary link between one’s evidence and propositions supported by that evidence. Thus, removing these dispositions changes what fits a subject’s evidence. As a result, NEDP cases of the sort that M&R present fail to pose a problem for EF as well.3


    1. Actually, since the publication Evidentialism and Epistemic Justification I have come to think that explanatory consequence is the relevant relation rather than logical consequence. However, to keep things more clearly in line with the present discussion I will stick to logical consequence; this won’t make a difference in what follows. For discussion of why explanatory consequence is the relevant notion see Kevin McCain, “Explanationism: Defended on All Sides,” Logos & Episteme 6 (2015) 333–49.

    2. My discussion of instances where multiple explanations are available to S helps to illustrate why my view is best understood in this way rather than the way M&R construe it. See, in particular, Kevin McCain, Evidentialism and Epistemic Justification (New York: Routledge, 2014), 66. This is no criticism of M&R’s exegesis of my position; rather it is an acknowledgement of the fact that I should have made this portion of the discussion clearer.

    3. Thanks to Matt Frise for helpful discussion.

Joshua Smart

Response

Coherence Concerns for McCain’s Explanationist Evidentialism

Introduction

Kevin McCain’s Evidentialism and Epistemic Justification contains careful, well-organized argumentation in the service of an important and clearly defined positive project: to rectify the fact that “Evidentialism remains more of a schema for a theory of epistemic justification than a full-fledged theory” (4). McCain calls the full-fledged theory that he develops Explanationist Evidentialism (EE), the core idea of which is that S justifiedly believes that p when p is part of the best explanation for S’s non-factive mental states.1

While I find much to agree with in McCain’s book, I think that EE ultimately fails to capture the importance of coherence to epistemic justification.2 Below, I present a (schematic) case that I argue is a counterexample to EE when combined with some plausible epistemological principles. 

Two Features of EE

The most relevant feature of EE to this discussion is McCain’s moderate view of evidence possession, MVP*.

MVP*

S has p available as evidence relevant to q at t iff at t S is currently aware of p or S is disposed to bring p to mind when reflecting on the question of q’s truth.” (51)

There is a second principle that I will take to be a feature of EE despite its not having been explicitly endorsed by McCain. However, I take it to be an implicit part of McCain’s overall view, given that he takes it to be “very important” that his theory of well-foundedness includes a no-defeater condition (119).3 McCain does not discuss the nature of defeat, but he does at least seem to have it that if q functions as a defeater for S’s belief that p, then q is part of S’s total evidence for p.4 Since it would be ad hoc to include S’s beliefs about q or ~q in S’s evidence for p only in cases in which q actually defeats S’s justification for p, I will below treat the following corollary as part of the EE package.[/footnote]I follow McCain in talking of “evidence for p,” which he does in an attempt to remain neutral on the question of whether one’s evidence consists in one’s non-factive, assertive mental attitudes or their propositional objects.[/footnote]

Defeater Corollary

If S’s justification for p would be defeated were S to believe that q, then S has available as evidence relevant to p S’s doxastic attitudes with respect to q and ~q of which S is aware or is disposed to bring to mind when reflecting on the question of p’s truth.5

The Coherence Concern

Consider the following, schematic case.

Incoherence Case

  • S believes that p, which is part of the best explanation for S’s total evidence (as defined by MVP*).
  • S believes that q, r, and s.
  • ~q would defeat S’s justification for p.
  • S is not disposed to bring either r or s to mind when reflecting on the question of p’s truth.
  • The propositions q, r, and s are mutually inconsistent.

(B) and (C) tell us, in combination with the defeater corollary and MVP* respectively, that q, but neither r nor s, are part of S’s evidence for p. But (D) tells us that S’s belief that q is a member of an incoherent belief-set. This, I argue, is a spanner in the works given the following plausible principles.

Justified Basis

S’s belief that p is justified only if the beliefs on which it is based are themselves justified.

Unjustified Incoherence

S’s belief that p is unjustified if p is a member of a set of S’s beliefs that is incoherent.6

Respecting this pair of principles in the Incoherence Case yields the verdict that S’s belief that p is unjustified. This is because S’s justification for p is (partly) based on q, which is unjustified for S because it fails to cohere with r and s. But (A) tells us that S’s belief that p is justified on EE—r and s are not part of S’s evidence for p, so they do not come into play on the view, and, while ~q would act as a defeater, S in fact believes that q. So, EE yields a false positive for any set of beliefs that fits the Incoherence Case schema, if Unjustified Incoherence and Justified Basis are true.

Are they true? Justified Basis is the intuitive first step in the traditional regress problem (solutions to which distinguish foundationalism, coherentism, and infinitism7. Although he considers the question of where in the traditional taxonomy EE belongs, McCain does not discuss the traditional regress motivation for the distinctions, and its like is held by almost all internalists.8 That leaves Unjustified Incoherence as the most likely target for a counter-attack.

Indeed, as it stands, few will endorse the principle. If nothing else, it is implausible because (as Frank Ramsey noted long ago) we none of us have a globally coherent set of beliefs (Ramsey, 1990).9 Therefore, taken strictly, Unjustified Incoherence leads directly to skepticism—a consequence McCain would likely see as beyond the pale (despite admirably rejecting turning anti-skepticism into an assumption). However, it is also clear that incoherence undermines justification in at least some cases, so it seems that what we need to restrict, rather than abandon, the principle. Unfortunately, this is no simple matter. There is a tension between the restriction’s tightness and its plausibility that is not unlike the tension that leads McCain to adopt a moderate view of evidence possession over inclusive and restrictive variants.

Considering MVP* puts a fine point on the problem. Salvaging EE via a restriction on Unjustified Incoherence requires finding a restriction such that q is unjustified for S, in virtue of its being inconsistent with S’s beliefs that r and s, only when r and s are part of S’s evidence for p according to MVP*.10 But, given the quirks of human psychology, I am skeptical, to say the least, of finding any non–ad hoc restriction of this sort.

Options

There are certainly other options. McCain could restrict MVP* instead. Though, here again, I see no non–ad hoc helpful way to do so (at least short of abandoning it for its accessibilist counterpart, RVP11). Similar dangers lie the way of amending the Defeater Corollary.

Finally, there is the possibility of rejecting Justified Basis. This might even be McCain’s initial inclination, given that his statement of EE places no justification condition on evidence. But without going full accessibilist, it seems a bridge too far that, for example, p could be justified for me despite my evidence for it consisting entirely in the products of wishful thinking or laughably sloppy reasoning.

At this point, I will leave it to McCain to attempt to make one of these unsavory options more palatable, or perhaps to concoct a dish of his own.

References

Littlejohn, Clayton. “Stop Making Sense? On a Puzzle about Rationality.” Philosophy and Phenomenological Research, 2015.

McCain, Kevin. Evidentialism and Epistemic Justification. New York: Routledge, 2014.

Ramsey, Frank. “Truth and Probability.” In Philosophical Papers, edited by D. H. Mellor, 52–109. Cambridge: Cambridge University Press, 1990.

Worsnip, Alex. “The Conflict of Evidence and Coherence.” Philosophy and Phenomenological Research, 2015.


  1. This is a brutally rough statement of the nature of propositional justification on EE. For the full statement, and that of well-foundedness (what is often called “doxastic justification”), see 117–18.

  2. Recently, Clayton Littlejohn (2015)—as well as Alex Worsnip (2015) and others—has criticized evidentialism by appealing specifically to considerations coherence between first and higher order attitudes. While I think that EE has problems along these lines as well, I do not have anything to add to Littlejohn’s argument. I will therefore focus on a problem having to do with first order coherence.

  3. Ex-WF (IV), the details of which are not important for our current purpose, but are found on p. 118.

  4. E.g., “It is assumed that in the examples S’s total evidence does not include defeaters relevant to the proposition being discussed” (65).

  5. I have elided temporal references, but the principle should be read as synchronic.

  6. Many epistemologists would want to include a rider to the effect that S would become aware of the incoherence upon reflection on the belief-set. We can stipulate that S’s beliefs that q, r, and s meet any such condition, so I leave it elided.

  7. And the newcomer—foundherentism—in which camp McCain seems to think EE is best seen as falling (121)

  8. At least, it is if we restrict it to non-basic beliefs—but of course p is a non-basic belief in the incoherence case.

  9. There are also looming concerns about preface and lottery paradoxes, as Andrew Moon has pointed out to me.

  10. This would lead, presumably, to a rejection of (A) on the basis that there is no best explanation for an incoherent set of mental states.

  11. McCain’s label for Richard Feldman’s restrictive view of evidence possession, against which McCain argues against at length (35–49)

  • Kevin McCain

    Kevin McCain

    Reply

    The Coherence of Explanationist Evidentialism

    J. A. Smart’s discussion of my theory of Explanationist Evidentialism begins by noting that coherence is an important component of epistemic justification. I completely agree with Smart on this point, and I appreciate his careful evaluation of my view. However, he and I disagree over his claim that Explanationist Evidentialism “ultimately fails to capture the importance of coherence to epistemic justification.” In this brief reply I will show where Smart’s argument for this conclusion goes wrong. As a result, it will become clear that there is no reason to think that Explanationist Evidentialism fails in the way that Smart suggests.

    Let’s begin by recalling Smart’s “Incoherence Case”:

    • S believes that p, which is part of the best explanation for S’s total evidence (as defined by MVP*).
    • S believes that q, r, and s.
    • ~q would defeat S’s justification for p.
    • S is not disposed to bring either r or s to mind when reflecting on the question of p’s truth.
    • The propositions q, r, and s are mutually inconsistent.

    According to Smart, the coherence problem for Explanationist Evidentialism arises when we couple Incoherence Case with two principles:

    Justified Basis

    S’s belief that p is justified only if the beliefs on which it is based are themselves justified.

    Unjustified Incoherence

    S’s belief that p is unjustified if p is a member of a set of S’s beliefs that is incoherent.

    Essentially, the purported problem is that these two principles yield that S’s belief that p is not justified in Incoherence Case, but Explanationist Evidentialism seems to commit one to claiming that this belief is justified. Hence, Smart concludes that Explanationist Evidentialism doesn’t respect coherence as it should because it gets this case involving incoherence wrong.

    Admittedly, if Smart’s argument is sound, it seems that Explanationist Evidentialism faces a serious problem. Fortunately, there is good reason to think that the argument is unsound. The flaw in this argument is that it relies on a false principle, Justified Basis.1 It is clear that not every component of S’s basis for believing that p has to be justifying in order for S’s belief that p to be justified. As I discuss in chapter 5, it is possible that S’s belief that p is justified even though she believes it partially on the basis of something like wishful thinking or unjustified beliefs. (97–99) There is not space to recount that entire discussion here, but the upshot is that S’s belief that p can be justified so long as the reasons that propositionally justify her in believing that p play a strong enough causal role in her having the belief that p. It is fortunate that this is so because if it weren’t the case, many of the beliefs we take to be justified wouldn’t be so. After all, it is plausible that the bases for many of our beliefs have at least some components that are not justifying whether this is due to bias, wishful thinking, unjustified beliefs, or something else. In light of the fact that not all of S’s basis for believing that p has to be justified in order for her belief that p to be justified it is clear that Justified Basis is false.

    Although Justified Basis is false, it is easy to see why one might be initially misled into thinking otherwise. This principle is similar to a true principle—all that is needed is a restriction in the scope of Justified Basis. If we modify this principle so that it applies to just those beliefs that are necessary for S’s propositional justification for p, we arrive at the following true epistemic principle:

    Justified Basis*

    S’s belief that p is justified only if all the beliefs necessary for S’s propositional justification for p on which S’s belief that p is based are themselves justified.2

    Justified Basis* is correct, however, it can’t do the work that Justified Basis did in Smart’s argument. Consider, once we replace Justified Basis with Justified Basis* we face a question concerning Incoherence Case: Is S’s belief that q necessary for S’s propositional justification for p? If the answer is “no,” then Smart’s coherence worry doesn’t arise. S’s belief that q isn’t among the beliefs that are necessary for S’s propositional justification for p. As a result, Justified Basis* cannot be used to draw the conclusion that S’s belief that p isn’t justified in Incoherence Case. So, Explanationist Evidentialism doesn’t face a problem. Alternatively, if the answer is “yes,” (A) from Incoherence Case is false. After all, if S’s belief that q is a necessary component of her propositional justification for believing p, and q isn’t justified, there’s no reason to accept that p is part of the best explanation of S’s total evidence. In order for beliefs to be part of S’s evidence they must be justified, or at a minimum have an overall positive epistemic status.3 So, S’s belief that q isn’t part of her total evidence, but it would have to be in order for p to be justified for S. Consequently, (A) from Incoherence Case is false, which means that Incoherence Case is depicting an incoherent scenario. Again, Explanationist Evidentialism doesn’t face a problem.

    In sum, Smart poses an interesting challenge for Explanationist Evidentialism that is worth thinking through carefully. Close inspection of this challenge reveals that it relies on a false principle. Once a corrected form of the principle in question is utilized the challenge for Explanationist Evidentialism disappears. In light of this it is safe to conclude that Explanationist Evidentialism doesn’t have the sort of coherence problems that Smart envisions.4


    1. Actually, I think that both of the principles Smart appeals to are false. Nonetheless, I won’t focus on his Unjustified Incoherence here. For the interested reader, I discuss reasons for thinking this sort of principle is false when I discuss problems with early forms of coherentism in chapter 5 of Kevin McCain, The Nature of Scientific Knowledge: An Explanatory Approach (Switzerland: Springer, 2016).

    2. “Necessary” here should be understood in a fairly weak sense. It simply means that the belief in question is part of S’s actual reasons which propositionally justify her in believing that p and removing that belief would make it so that S is no longer propositionally justified in believing that p if all of her other mental states are held fixed.

    3. I discuss this point on p. 96 of Evidentialism and Epistemic Justification.

    4. Thanks to Matt Frise and Ted Poston for helpful discussion.

Shares