Introduction
Berker (Reference Berker2013a, Reference Berker2013b, Reference Berker2015) – along with others (Jenkins Reference Jenkins2007; Littlejohn Reference Littlejohn2012; Greaves Reference Greaves2013) – recently revives the interest in the problems concerning epistemic consequentialism originally posed by Firth (Reference Firth1998). He argues that epistemic consequentialism is structurally flawed. This is supposed to be deeply worrying because Berker observes that consequentialism is pervasive in epistemology, as he comments:
One naturally expects that, just as in the ethics literature where a wide variety of teleological theories are defended and a wide variety of nonteleological theories are defended, so too the epistemology literature should be populated by a wide variety of both teleological and nonteleological theories. But what one finds when one turns to the epistemology literature as it exists today is quite surprising, for a teleological approach to normative epistemology is overwhelmingly the dominant view. (Berker Reference Berker2013a: 350–1)
I agree with Berker that epistemic consequentialism is structurally flawed. And if Berker is right that epistemic consequentialism is pervasive, then epistemology is in deep trouble. Fortunately, he is wrong about the last point. I will argue that most of the “card-carrying consequentialists” picked out by Berker are better to be considered non-consequentialists. There are indeed self-claimed epistemic consequentialists. But if I am right, I suspect that they are probably mistaken about their own views. They are better to be interpreted as what I call epistemic instrumentalists.
In section 1 and 2 respectively, I explain what I mean by epistemic consequentialism and epistemic instrumentalism. Both views maintain that epistemic justification is a means for achieving the epistemic goal(s). But one crucial difference separates them. Supposing that truth is the sole epistemic goal, epistemic consequentialism maintains that truth-conduciveness is the necessary and sufficient condition for epistemic justification, whereas epistemic instrumentalism maintains that truth-conduciveness is merely a necessary condition. This clarifies the structural problem of epistemic consequentialism Berker tries to identify: that is, the problem of epistemic consequentialism is caused by the claim that truth-conduciveness alone is sufficient for justification. Accordingly, epistemic instrumentalism can avoid the consequentialist problem. In section 3, I argue that most “consequentialists” Berker identifies are better to be considered instrumentalists. Thus, consequentialism is not pervasive in epistemology.
To be clear, I do not intend to defend epistemic instrumentalism as a viable theory of epistemic justification, though I do argue that it can avoid the structural problem of consequentialism.Footnote 1 Rather, I want to discredit the claim that epistemic consequentialism is pervasive. If few of the “consequentialists” are genuinely consequentialists, then the fact that epistemic consequentialism is structurally flawed is not worrying.
Recently, some commentators (Piller Reference Piller, Schmechtig and Grajner2016; Driver Reference Driver, Ahlstrom-Vij and Dunn2018; Singer Reference Singer2018) argue that the analogy between ethical and epistemic consequentialism is misplaced and therefore Berker cannot pose similar counterexamples from ethics to epistemic “consequentialism”. While our views seem similar, they focus more on the disanglogy and do not explain why several epistemologists misunderstand themselves as consequentialists because epistemic consequentialism and instrumentalism are easily confused. I offer a diagnosis of their confusion, which would be helpful for future discussion about epistemic teleology.
1. Epistemic consequentialism and its structural problem
As Berker (Reference Berker2013a, Reference Berker2013b) analyses, all forms of consequentialism – with the exception of scalar consequentialism which discards deontic properties – consists of three components: (i) a theory of final value that specifies what final values are; (ii) a theory of overall value that explains how to assess and rank the overall values of certain entities by measuring how they conduce toward final values; and (iii) a deontic theory that determines the deontic properties of objects – being right, justified, or permissible – based on the theory of overall value. The theory of final value is thus the most fundamental one. One of the most popular theories of epistemic final value is veritism, which maintains that truth is the sole epistemic final value. In the following discussion, I will focus exclusively on veritist epistemic consequentialism, though nothing of what I will argue is affected by this assumption.
Under the consequentialist framework, deontic properties have only instrumental value because their values are determined solely by the overall final values to which they conduce. As Berker explains, “nonfundamental [nonfinal] value is explicable in terms of conduciveness toward fundamental [final] value,” and “deontic notions such as obligation, permission, and rightness must obtain in virtue of facts about value” (Reference Berker2013a: 344). Based on Berker's account, therefore, epistemic consequentialism is committed to the following thesis:
Instrumental-Value. Epistemic justification is only instrumentally valuable for achieving the truth goal.
But if epistemic consequentialism is committed to Instrumental-Value, it must also accept the following thesis:
Truth-Conduciveness. Truth-conduciveness is necessary and sufficient for epistemic justification.
Under the consequentialist framework, Instrumental-Value and Truth-Conduciveness are two sides of the same coin: since justification is measured only by its instrumental value relative to truth, whatever has enough instrumental value can thus provide justification. Truth-conduciveness, it follows, is sufficient and necessary for epistemic justification. Thus, Berker comments on reliabilism – which he regards as a paradigm of epistemic consequentialism – that a more perspicuous name for reliabilism is “truth-conducivism” (Berker Reference Berker2013a: 350).
Given Instrumental-Value and Truth-Conduciveness, we can identify the structural problem of epistemic consequentialism. Berker notes that initially he thought that epistemic consequentialism is wrong “because they ‘ignore the separateness of propositions’,” which allows problematic trade-offs (Reference Berker2015: 146). But he later recognizes that there is a deeper structural problem, though he says, “I am still struggling to figure out the best way to characterize this more general structural feature” (Reference Berker2015: 146).
I would like to suggest that the structural problem is caused exactly by Instrumental-Value and Truth-Conduciveness. In fact, Berker has already indicated that, for Firth, consequentialism is untenable because it “attempts to analyze intrinsic epistemic merit [of justification] in terms of instrumental epistemic merit,” but Berker believes that “his inference is too quick” (Reference Berker2013b: 369).Footnote 2 I disagree; Firth's view is spot-on. Let's examine my diagnosis on some of Berker's counterexamples for epistemic consequentialism:
Scientist. Suppose I am a scientist seeking to get a grant from a religious organization. … I realize that my only chance of receiving funding from the organization is to believe in the existence of God. … Finally, I know that, were I to receive the grant, I would use it to further my research, which would allow me to form a large number of new true beliefs and to revise a large number of previously held false beliefs about a variety of matters of great intellectual significance. … Would [my belief that God exists] be epistemically rational, or reasonable, or justified? (Berker Reference Berker2013a: 363–4)
Prime Number. Suppose the following is true of me: whenever I contemplate whether a given natural number is prime, I form a belief that it is not. “Is 25 prime? No, it is not.” “Is 604 prime? No, it is not.” “Is 7 prime? No, it is not.” Let us also stipulate that this is the only cognitive process by which I form beliefs about the primeness of natural numbers. … Since the ratio of prime to composite numbers less than n approaches 0 as n approaches infinity, my belief-forming process tends to yield a ratio of true to false beliefs that approaches 1. Therefore process reliabilists are forced to say that, because my belief-forming process is almost perfectly reliable, any belief formed on its basis is justified. But that's crazy! (Berker Reference Berker2013b: 374)
My account fits both examples. For epistemic consequentialism, my belief that God exists in Scientist would be considered justified because it will cause many true beliefs and eliminate many false beliefs. In Prime Number, my belief that 7 is not a prime number is considered justified because it is produced by a highly reliable process. In both examples, truth-conduciveness alone is sufficient for justification. Hence, my account does capture the structural problem in Berker's counterexamples.
Therefore, Instrumental-Value and Truth-Conduciveness are the structural problems of consequentialist. But Berker is wrong that the consequentialist problem is worrisome for epistemologists. True, many epistemologists talk as if they are consequentialists, but they do not really accept Instrumental-Value and Truth-Conduciveness. They are what I call epistemic instrumentalists. In the next section, I introduce epistemic instrumentalism.
2. Epistemic instrumentalism
Epistemic instrumentalism is a non-consequentialist twin of epistemic consequentialism. Likewise, epistemic instrumentalism holds the following theses (“+” is added to express the idea that instrumentalism maintains that epistemic justification has more than instrumental value and truth-conduciveness):
Instrumental-Value + . Epistemic justification is a means to the truth goal, which has epistemic final value (or value beyond merely instrumental value).
Truth-Conduciveness + . Truth-Conduciveness is necessary, but not sufficient, for epistemic justification.
Epistemic instrumentalism is easily confused with consequentialism because both regard justification as a means to truth. Unlike consequentialism, however, epistemic instrumentalism holds that justification is epistemically finally valuable, which makes it worth pursuing for its own sake beyond the goal of truth. Since epistemic instrumentalism maintains that justification has value beyond its instrumental value relative to truth, it doesn't need to maintain that truth-conduciveness is sufficient for justification.
One may question how justification, being a means to truth, can have value beyond instrumental value. Indeed, it appears that the value of epistemic justification lies solely in truth-conduciveness and is thus derivative from truth. That could explain why Instrumental-Value looks so attractive. Since accepting Instrumental-Value leads one to commit to Truth-Conduciveness, let me explain how Instrumental-Value could be resisted.
Sylvan (Reference Sylvan2018) offers such an account. Sylvan, borrowing an insight from Hurka (Reference Hurka2001), points out that, if X is finally valuable and Y is a proper way of valuing X, the value of Y is derivative from X, but is nevertheless finally valuable.Footnote 3 For example, the love of truth derives its value from truth and is also finally valuable. Sylvan argues that, since epistemic justification manifests a way of respecting truth, its value is final and derivative from truth. Therefore, even if one considers epistemic justification as a means to truth, it does not follow that one must accept Instrumental-Value.Footnote 4
As I've discussed, under the consequentialist framework deontic properties have only instrumental value. Deontic properties are excluded from the theory of final value and are determined by their conduciveness to final values alone. Since epistemic instrumentalism holds that epistemic justification has final value, it rejects the whole consequentialist framework. For instrumentalism, the right is something worth pursuing for its own sake. By rejecting Instrumental-Value and the consequentialist framework, epistemic instrumentalism needn't commit itself to Truth-Conduciveness. Instead, it can opt for Truth-Conduciveness+, the thesis that truth-conduciveness is necessary, but not sufficient for, epistemic justification. To be sure, those who accept Instrumental-Value+ can still maintain that truth-conduciveness is sufficient for justification. The point, however, is that they do not have to, unlike epistemic consequentialists.
By rejecting Instrumental-Value and Truth-Conduciveness, epistemic instrumentalism can avoid the structural problem of consequentialism caused by the thesis that truth-conduciveness is sufficient for epistemic justification, because epistemic instrumentalism regards truth-conduciveness merely as a necessary condition.Footnote 5
Let me give an example of epistemic instrumentalism: Plantinga, who Berker (Reference Berker2013a: 355 n.27; Reference Berker2013b: 368) wrongly considers a consequentialist. Here is Plantinga's account of epistemic justification (though he uses the term “warrant” instead):
A belief has warrant for me only if (1) it has been produced in me by cognitive faculties that are working properly (functioning as they ought to, subject to no cognitive dysfunction) in a cognitive environment that is appropriate for my kinds of cognitive faculties, (2) the segment of the design plan governing the production of that belief is aimed at the production of true beliefs, and (3) there is a high statistical probability that a belief produced under those conditions will be true. (Plantinga Reference Plantinga1993: 46–7)
Despite the condition (3) concerning truth-conduciveness, Plantinga is not a consequentialist because truth-conduciveness is only a necessary condition. The first two conditions offer non-consequentialist elements: a belief is justified only if it is produced by a suitable cognitive faculty, which must be truth-conducive in ways proper to its function or design. These non-consequentialist conditions allow Plantinga to deal with Berker's counterexamples. Plantinga could argue that the religious belief in Scientist and the belief-forming process in Prime Number do not satisfy the first two conditions and thus they cannot provide justification.
One may worry that the first two conditions in Plantinga's theory might be consequentialist because what makes them conditions for justification is Truth-Conduciveness. To make this objection work, for sure, it should be the case that Truth-Conduciveness alone is sufficient for a faculty or a function to provide justification. If so, Plantinga could still be considered a consequentialist.
While I agree that Truth-Conduciveness remains an essential part of the first two conditions, it is not sufficient for the idea of proper function. Compare an example of proper function from Plantinga with Prime Number: “when the coolant temperature of my car gets up to 200°F, the thermostat should open” (Reference Plantinga1993: 22). There are many ways to explain their differences. But the idea of sensitivity is useful here. The thermostat functions properly because its function is sensitive to the temperature of the coolant; the thermostat would not open if the coolant is under 200°F. However, the belief-forming process in Prime Number is not sensitive because, even when a number is a not prime number, the process will still form a belief that it is a prime number.
More importantly, while sensitivity does contribute to the Truth-Conduciveness of the thermostat, it is not determined by Truth-Conduciveness alone. One factor in sensitivity, I think, is causality. The mechanism of the thermostat is causally linked to the temperature of the coolant, which explains why its responsiveness to the coolant's temperature is reliable. However, what explains the reliability of the belief-forming process in Prime Number is merely the fact that the overwhelming majority of numbers are not prime numbers. In other words, while the process in Prime Number could be much more reliable than the thermostat, the former does not, but the latter does, form their target beliefs in a proper way. Therefore, not any sort of Truth-Conduciveness is sufficient for proper function.
To be clear, my aim is not to offer a defense of epistemic instrumentalism, but merely to correct a grave self-misunderstanding among epistemologists. It is enough for my purpose to show that epistemic instrumentalism can avoid the structural problem of consequentialism and most epistemic “consequentialists” are better to be interpreted as instrumentalists, which I argue in the next section.
3. Consequentialists or instrumentalists?
Given the similarity between epistemic consequentialism and instrumentalism, it is unsurprising that they could be easily confused. In a footnote (Reference Berker2013a: n.27), Berker names 35 epistemologists as consequentialists. They are “consequentialists,” according to Berker, mainly because they consider Truth-Conduciveness as the hallmark of epistemic justification, or they consider justification a means to truth (Reference Berker2013a: 350–7). For example, he considers Richard Foley a consequentialist just because Foley insists that “epistemic rationality in particular is structured around the fundamental goal of ‘now believing those propositions that are true and now not believing those propositions that are false’” (Berker Reference Berker2013a: 353–4).
We now know that instrumentalism also has these features. To show that they are consequentialists, he needs to demonstrate that they do accept Instrumental-Value and Truth-Conduciveness. For example, Berker regards John Greco as a consequentialist, probably because Greco thinks that justified belief is “adequate to her goal of believing the truth” (Greco Reference Greco1999: 290). But Greco requires a belief's doxastic justification is grounded in one's cognitive disposition one manifests when one is thinking conscientiously (Reference Greco1999: 289). Since Greco does not accept Truth-Conduciveness, he is not a consequentialist. The same problem applies to other epistemologists in Berker's list. So, I think that there is no need to go through every one of them.
I have argued that Plantinga is not a consequentialist. I will discuss three prominent epistemologists whom Berker identifies as consequentialists – William Alston, Laurence Bonjour, and Alvin Goldman. I argue that Berker fails to demonstrate that they are consequentialists because either there is no evidence that they do accept Instrumental-Value and Truth-Conduciveness, or they do reject them. I also discuss the defense of process reliabilism by Ahlstrom-Vij and Dunn (Reference Ahlstrom-Vij and Dunn2014; Dunn and Ahlstrom-Vij Reference Dunn and Ahlstrom-Vij2017) and argue that their defense in fact supports that reliabilism is instrumentalist.
3.1. William Alston
Alston (Reference Alston2005) urges us to dispense the monistic conception of epistemic justification with the pluralist conception of epistemic desiderata. Apparently, Alston rejects Instrumental-Value because all epistemic desiderata are epistemically desirable for their own sake. However, Berker maintains that Alston is a consequentialist because “for Alston, all epistemic desiderata are ultimately defined in terms of how well they help us further ‘the goals of cognition’” (Berker Reference Berker2013a: 354). It's true that Alston takes truth as the most fundamental epistemic goal, so he laboriously explains how each desideratum is connected to the truth goal. But notice that the goals of cognition are plural. Each epistemic desideratum is truly finally valuable from the epistemic point of view. So, in the discussion of whether “goals of cognition that are partly independent of any connection with the goal of truth” should count as epistemic desiderata, Alston comments,
The question is whether we want to loosen up the requirements of epistemic desirability to include items the intrinsic desirability of which is over and above that of the true-false balance but which presupposes such a balance as a necessary condition of that desirability. In the absence of any sufficient reason for being hardnosed on this issue, I will allow the realization of these cognitive goals to count as epistemic desiderata. (Alston Reference Alston2005: 46–7)
When Alston tries to relate other desiderata to truth, he is not claiming that all those desiderata are only instrumentally valuable in relation to truth. Instead, he is trying to define why those desiderata are epistemic rather than moral or prudential (that's why he italicized “epistemic” in the above quotation). For Alston, the final value of an epistemic desideratum is not exhausted by truth or truth-conducive alone; it does have epistemic final value over and above truth. Obviously, Alston rejects Instrumental-Value, so he is not a consequentialist.Footnote 6
3.2. Laurence Bonjour
In an oft-quoted passage, Bonjour apparently espouses epistemic consequentialism:
What makes us cognitive beings at all is our capacity for belief, and the goal of our distinctively cognitive endeavors is truth. … The basic role of justification is that of a means to truth. … If epistemic justification were not conducive to truth in this way, if finding epistemically justified beliefs did not substantially increase the likelihood of finding true ones, then epistemic justification would be irrelevant to our main cognitive goal and of dubious worth. … Epistemic justification is therefore in the final analysis only an instrumental value [my italics], not an intrinsic one. (Bonjour Reference Bonjour1985: 7–8)
Bonjour accepts Instrumental-Value, which suggests that he is a consequentialist. However, he might wrongly believe that something's being a means implies its having only instrumental value. And if he does not adopt the consequentialist framework that the right is determined by the good, the acceptance of Instrumental-Value does not necessarily lead him to Truth-Conduciveness. To show that Bonjour is a consequentialist, therefore, it is better to demonstrate that he also accepts Truth-Conduciveness.
Berker (Reference Berker2013a: 353) cites several of Bonjour's objections to foundationalism and coherentism as evidence for his being a consequentialist. First, in his early objection to foundationalism, Bonjour argues that foundationalists have difficulty in explaining why basic beliefs are justified because “for a belief to be epistemically justified requires that there be a reason why it is likely to be true” (Bonjour Reference Bonjour1985: 32). Second, Bonjour includes the “doxastic presumption” – that my representation of my belief system is approximately correct – as part of his early coherentism (Bonjour Reference Bonjour1985: Ch. 5.4). Finally, in his discussion about the problems of coherentism (Bonjour Reference Bonjour1985: Ch. 5.5; Bonjour and Sosa Reference Bonjour and Sosa2003: Ch. 3), all of the problems of coherentism – the input objection, the alternative systems objection, the meta-justification objection – are about how an internally coherent system of beliefs is not truth-conducive. However, his objections amount to the claims that foundationalism has difficulty in explaining how basic beliefs are likely to be true and that being coherent is genuinely truth-conducive. These objections require only that truth-conduciveness be a necessary condition for justification. They cannot demonstrate that Bonjour is a consequentialist.
That said, I could not find where Bonjour clearly endorses Truth-Conduciveness + . Perhaps, he is indeed a consequentialist, but Berker has yet proven it. Certainly, the fact concerning Bonjour alone cannot demonstrate that consequentialism is pervasive in epistemology.
3.3. Alvin Goldman
Alvin Goldman is “the father of process reliabilism” (Berker Reference Berker2015: 146) and a self-proclaimed epistemic consequentialist (Goldman Reference Goldman1986: 97; Reference Goldman2002: Ch. 3). So there should be no doubt that Goldman is a consequentialist. Or is he?
In his reply to Berker's Prime Number example, Goldman argues that the counterexample does not work because the belief-forming process fails to satisfy the content-neutrality constraint (Goldman Reference Goldman2015: 141–2). Goldman first proposes the content-neutrality constraint in his seminal paper on reliabilism, “What is Justified Belief?” – though he appears to maintain that reliability is sufficient for justification (Goldman Reference Goldman and Pappas1979: 13) – in order to address potential counterexamples like Prime Number:
It is clear that our ordinary thought about process-types slices them broadly, but I cannot at present give a precise explication of our intuitive principles. One plausible suggestion, though, is that the relevant processes are content-neutral. It might be argued, for example, that the process of inferring p whenever the Pope asserts p could pose problems for our theory. If the Pope is infallible, this process will be perfectly reliable; yet we would not regard the belief-outputs of this process as justified. The content-neutral restriction would avert this difficulty. If relevant processes are required to admit as input beliefs (or other states) with any content, the aforementioned process will not count, for its input beliefs have a restricted propositional content, viz., “the Pope asserts p”. (Goldman Reference Goldman and Pappas1979: 12)
Berker responds that the content-neutrality constraint is either implausible or ineffective (Reference Berker2015: 150–1). It is implausible because many reliable processes that can justify beliefs do not admit the inputs and outputs with any contents (for example, visual perception would normally justify only beliefs about visual features of objects). If the constraint requires only that the process admit the inputs or outputs with many different contents (Goldman Reference Goldman2015: 141), Berker contends that it is ineffective because Prime Number can easily be modified to satisfy the constraint. Hence, Berker concludes that the content-neutrality constraint does not save process reliabilism.
While Berker's response is plausible, I think that he misses the big picture, a sign that both sides fail to grasp the nature of epistemic consequentialism. For a consequentialist to use the content-neutrality constraint, the constraint must not contain any non-consequentialist element. That is, the reason for a consequentialist to adopt the content-neutrality constraint must solely be truth-conducive, or the fact that content-specificity would make a process less reliable. But that reason doesn't seem correct. In Goldman's original example, the process of “inferring p whenever the Pope asserts p,” Goldman admits, is perfectly reliable. Why would a consequentialist deny that it can deliver justification? Goldman offers no explanation. As long as the Pope remains infallible, Goldman could not reject content-specificity on consequentialist grounds. The content-neutrality constraint is more likely to be non-consequentialist. Thus, Goldman should be regarded as an instrumentalist.
3.4. Kristoffer Ahlstrom-Vij and Jeffrey Dunn
It's interesting to compare the response of Goldman with that of Kristoffer Ahlstrom-Vij and Jeffrey Dunn. Unlike Goldman, Ahlstrom-Vij and Dunn (Reference Ahlstrom-Vij and Dunn2014) maintain that the process in Prime Number can indeed provide justification. Whether their defense is correct is not the concern here. The crucial point is that this reply is genuinely consequentialist. In their more recent paper (Dunn and Ahlstrom-Vij Reference Dunn and Ahlstrom-Vij2017), they argue that some versions of consequentialism can avoid the trade-off problems and process reliabilism is one of them.
So, if Ahlstrom-Vij and Dunn's response is acceptable, it seems that process reliabilism could ditch the content-neutrality constraint and remain true to consequentialism. However, this is not true. In their responses to other counterexamples, they show that process reliabilism essentially contains non-consequentialist elements.
To see that, let's examine their responses to counterexamples like Scientist. Ahlstrom-Vij and Dunn argue that Scientist poses no problem to process reliabilism because my belief in God is considered justified in the forward-looking way, but process reliabilism determines the justification of belief in the backward-looking way, namely, in terms of whether the belief is produced by a truth-conducive process. Thus, they conclude that counterexamples like Scientist cannot threaten reliabilism (Reference Ahlstrom-Vij and Dunn2014: sec. 3).
Berker does comment that epistemic justification is essentially backward-looking, whereas consequentialism is forward-looking (Reference Berker2013b: 377). So, it appears that process reliabilism could avoid Berker's criticism. However, Berker anticipates this response (Reference Berker2013b: 544). Consider a modified version of Scientist. Now I form the belief that God exists through the process of forming the belief x whenever forming it will promote true beliefs over false beliefs. Given that the process is truth-conducive, process reliabilism would have to consider my belief justified.
To respond to this objection, Ahlstrom-Vij and Dunn argue that process reliabilism requires that the reliability of a process must be evaluated only in terms of the truth ratio of belief it directly produces (Reference Ahlstrom-Vij and Dunn2014: 544). Hence, the true beliefs that my belief in God promotes must be excluded when the reliability of the process is evaluated. In that case, the process is not reliable, so they think that Berker's objection still fails.
Ahlstrom-Vij and Dunn deny that their defense is ad hoc. But the issue is whether it is consequentialist. In their later paper, Ahlstrom-Vij and Dunn argue that consequentialism doesn't need to consider all consequences: for example, act consequentialism considers only the consequences of the act evaluated, and rule consequentialism only consequences of the rule when people comply. So, they conclude that it is fine for reliabilism to consider only the consequences directly produced by the process (Dunn and Ahlstrom-Vij Reference Dunn and Ahlstrom-Vij2017: sec. 4).
However, the analogies break down. Rule consequentialism considers only consequences of the rule when people comply because they are what complying to the rule will or will be likely to cause. But it does consider all consequences – directly and indirectly – caused by complying to the rule, at least those that could be reasonably anticipated. True, one may worry that some consequences are so remote or deviated that it's difficult or even wrong to take them into consideration. But any indirectly-caused consequences that can be reasonably foreseen should be considered. This is what happens in Scientist. The scientific discoveries that my belief in God will promote are reasonably foreseeable. It's puzzling why any consequentialism would disregard them. So, their positing the ad hoc condition is more likely to be driven by non-consequentialist concerns. What is wanting in their defense is an account of why the exclusion of indirectly-caused consequences can enhance Truth-Conduciveness. Without such an account (I can't think of any), we should maintain that process reliabilism is instrumentalist.
In conclusion, epistemic consequentialism is not pervasive. It is true that many epistemologists consider themselves consequentialists. When examining their views more closely, I show that they are epistemic instrumentalists, who maintain that Truth-Conduciveness is merely a necessary condition of justification. By distinguishing between instrumentalism and consequentialism, I also identify the structural problem of consequentialism, that is, the thesis that Truth-Conduciveness is sufficient for justification.Footnote 7
