Veritism and ways of deriving epistemic value

Abstract

Veritists hold that only truth has fundamental epistemic value. They are committed to explaining all other instances of epistemic goodness as somehow deriving their value through a relation to truth, and in order to do so they arguably need a non-instrumental relation of epistemic value derivation. As is currently common in epistemology, many veritists assume that the epistemic is an insulated evaluative domain: claims about what has epistemic value are independent of claims about what has value simpliciter. This paper argues that the insulation approach to epistemic value is incompatible with non-instrumental epistemic value derivation. Veritsts who want to avail themselves of this important explanatory resource should therefore abandon the insulation approach.

1 Introduction

Veritistic monism – veritism for short – is the view that only truth (or accuracy – I will use these two terms interchangeably) is fundamentally epistemically good. An x has fundamental epistemic value just in case its epistemic value cannot be fully explained in terms of the epistemic value of some distinct y(s). It is uncontroversial, and compatible with veritism, that many things – and different kinds of things – have derivative epistemic value. An x has derivative epistemic value just in case its epistemic value is at least partly explained in terms of the epistemic value of some distinct y(s). Veritism is a popular view and apart from having several explicit defenders (see e.g. Ahlström-Vij 2013; David, 2001; Engel, 2017; Goldman, 1986, 2002; Olsson, 2007; Pritchard, 2014; Sylvan, 2018), it is often implicitly assumed by epistemologists who do not otherwise discuss questions of epistemic value.

Many veritists explicitly endorse what I call the insulation approach (IA) to epistemic value. The core idea behind this approach is that claims about epistemic value have little or no implication for, and do not depend on any particular claims about, what has value simpliciter. In this paper, I will argue that in order to obtain a much-needed explanatory resource for their view, veritists must reject the IA. For any alleged instance of epistemic goodness, veritists are committed to either explaining it in terms of the goodness of accuracy, or denying that it is an instance of epistemic goodness. Denying (too many) instances of seeming epistemic goodness is arguably theoretically costly, so the more instances they can explain the better. So, the more ways there are of deriving epistemic value from a fundamental epistemic good – in this case, the more value-transferring ways there are of relating to accuracy – the greater the variety of cases that a veritist can plausibly explain. In particular, it would be good for veritists if there was a relation of non-instrumental value derivation, since there are many instances of seeming epistemic goodness that fail to be truth-conducive. But, I will argue, non-instrumental value derivation is only available to veritists if they reject the IA.

2 The Insulation Approach to Epistemic Value

At issue with veritism is the question of fundamental epistemic value. But what does it mean to say that some x has epistemic value (whether fundamental or derivative)? As Ridge (2013) was relatively early to point out, the debate over epistemic value has at times been confused with respect to this question. To help sort this out, he reminds us of the way Geach (1956) highlights the relevance for philosophers to distinguish between predicative and attributive uses of modifying adjectives. In the locution ‘is a grey elephant’, the modifier ‘grey’ is predicative since the locution entails both being an elephant and being grey. Compare the locution ‘is a small elephant’, where the modifier ‘small’ is attributive as the locution does not entail being small. Ridge proposes to extend Geach’s distinction to locutions of the form ‘is F-ly G’, which count as predicative just in case they entail ‘is G’, and attributive if not. For instance, ‘is brightly coloured’ entails ‘is coloured’, but ‘is superficially reasonable’ does not entail ‘is reasonable’. Now the question is whether a claim of the form.

x is epistemically valuable

should be read predicatively or attributively. In the former case, being epistemically valuable entails being valuable, in the latter case it does not.

Adherence to the attributive reading of claims about epistemic value has become increasingly popular in the relevant literature. I will refer to endorsement of this reading as the insulation approach (IA) to epistemic value. The core idea behind the IA is that we can theorise about epistemic axiology and its internal organisation without regard for what is or isn’t valuable simpliciter, or to use a more colloquial term, genuinely valuable. Duncan Pritchard is one proponent of IA. He writes that the veritist’s claim that truth has fundamental epistemic value is “entirely consistent with arguing that the general value of truth is merely instrumental (or indeed, that it has no value at all)” (2014, p. 113).¹ Ridge also appears to recommend the IA², as does Sylvan (2018, p. 385; 420), DePaul (2001, pp. 176–177), Aschliman (2020), and, famously, Sosa (2007, Chap. 4). Sosa observes that human beings are “zestfully judgemental across the gamut of our experience: in art, literature, science, politics, sports, food, wine, and even coffee; and so on, across many other domains” (p. 70). Each of these are their own evaluative domains, structured around some domain-specific fundamental good(s), in relation to which evaluations within a given domain are made. These domains are insulated in the sense that if x is a fundamental good around which the critical domain D is organized, it does not follow that x has “any real value” (p. 73).

For instance, consider the chess domain, within which we evaluate chess moves, strategies, players, etc. It seems fair to assume that the fundamental good of chess, around which these evaluations are organized, is to checkmate one’s opponent. So, a chess move which is good from the point-of view of chess is a move which is conducive to the aim of checkmating the opponent. However, given this it is an open question whether even the most awesome chess move has any value simpliciter. This is because it is an open question whether the fundamental good of chess itself, to which a good chess move is conducive, has any value simpliciter. According to the IA, the epistemic domain is insulated in this sense too: epistemic evaluations do not entail or depend on claims about what is or isn’t valuable simpliciter, and that x has fundamental epistemic value does not entail that x has value simpliciter.

There are presumably different ways of understanding what, more exactly, is involved in viewing the epistemic domain as an insulated domain in this sense, and to place truth at its centre as the veritsts do. For instance, Hazlett (2013) suggests it is a matter of convention – of what we as theorists mean by “epistemic”– but it is easy to imagine that philosophers who hold that truth is the constitutive aim or norm of belief (e.g. Wedgwood 2002; Whiting, 2012) will conceive of things differently. I will not go into these matters here.³ For the purposes of this paper, what matters is that, as suggested by the paragraphs above, friends of the IA are committed to:

(NOT) If x is of epistemic value, it does not follow that x is of value simpliciter.

Note that this shouldn’t be taken to suggest that IA is the view that fundamental epistemic goods are not valuable simpliciter. The point, I take it, with adopting any version of the insulation approach, is to be able to remain neutral on questions about value simpliciter while discussing evaluations in a given domain. An x which has this type of insulated, domain-specific value may very well have value simpliciter too, but that question is orthogonal to whether it has domain-specific value. For instance, someone who takes achievements to have value simpliciter might hold that checkmating one’s opponent has value simpliciter in virtue of being an achievement (in cases where it is indeed an achievement). But this is independent of the “chess-value” of checkmating one’s opponent, and vice versa.

Several friends of the IA are veritsts, including e.g. Pritchard. And while the view is popular also among philosophers with other epistemic axiologies, this paper will focus on the prospects of veritism given IA. In particular, I will argue that that the IA is incompatible with holding that some things have non-instrumental derivative epistemic value in terms of being a (manifestation of a) fitting attitude to an epistemic good.

3 Veritism and epistemic value derivation

As noted, veritism is in principle compatible with distinct properties, activities, or states – such as curiosity, rationality, coherence, explanatory power, responsible inquiry, and so on – all being epistemically valuable. But importantly, veritism can only ascribe epistemic value to that which is related to the fundamental epistemic good of truth in the right way, i.e. in a way which allows it to derive its epistemic value. Pluralists about epistemic value challenge veritism by presenting instances of, or nuances in, epistemic goodness that allegedly cannot be explained as derivative of the value of truth. The idea is that while there might be a parsimony consideration in favour of veritism over pluralism in the sense that all else equal we ought to prefer axiologies that postulate fewer fundamental epistemic values, all else is not equal: in order to explain all that needs explanation – all instances of epistemic goodness – we need to postulate more than one fundamental epistemic good. Therefore, it is of great importance to the prospects of veritism what ways there are of deriving epistemic value.

Instrumental value is the most obvious and non-controversial type of derivative value: x has instrumental epistemic value if x is an effective means to acquiring something with fundamental epistemic value.⁴ For instance, reliable perceptual faculties are epistemically valuable because they are effective means to acquiring true beliefs. By contrast, x has non-instrumental epistemic value if x’s epistemic value is not (exhaustively) explained by its being an effective means for acquiring fundamental epistemic goods. A fundamental epistemic good has non-instrumental epistemic value, but a derivative epistemic good x may in principle also be a non-instrumental epistemic good (more on this below).

Instrumental epistemic value can take the veritist quite far in her endeavour to explain instances of epistemic goodness – many properties, traits, states, etc. are arguably effective means for promoting accuracy, directly or indirectly. But there are several problem cases, where there is either epistemic value additional to that conferred by truth, or epistemic value in the absence of truth-conduciveness. Epistemic virtues can exemplify both types of problem. Many epistemologists will want to say that traits such as open-mindedness, epistemic self-reliance, or curiosity, are epistemic virtues even if they fail to be generally truth-conducive. Additionally, while some responsibilist⁵ virtue epistemologists think that an intellectual virtue must be reliably truth-conducive (e.g. Zagzebski 1996; Driver, 2001), it is more common to claim that while many virtues may in fact be reliable, their epistemic value is not exhausted by their instrumental value (e.g. Baehr 2006; Code, 1987; Crerar, 2021; Montmarquet, 1992).

For another problem case where there allegedly is epistemic value in the absence of truth-conduciveness, take Goldman’s (1988) bifurcation of the justification concept, where he distinguishes strong (i.e. undefeated reliabilist) justification from weak justification. A subject’s belief formed through process or method M is weakly justified if M is unreliable but S’s does not believe that M is unreliable, and is blameless in taking M to be reliable (e.g. because there is no other reliable method available to S which shows M to be unreliable). Weak justification clearly does not have instrumental epistemic value with respect to believing truly, since it is specifically not reliably truth-conducive. But if weak justification is a form of epistemic justification – which Goldman claims it is – it must plausibly be of epistemic value.

Finally, the most well-rehearsed in the literature is the so-called swamping problem: if the value of an epistemic good (e.g. knowledge or understanding) is to be explained in terms of its instrumentality to truth, but truth is already present (as with knowledge and, on some accounts of it at least, understanding), then that further value is “swamped”. But then veritists cannot explain why e.g. knowledge or understanding has more epistemic value than mere true belief. Some veritists claim to have solved the swamping problem (e.g. Goldman & Olsson 2009; Olsson, 2007), others push back and insist that intuitions about the surplus value of knowledge or understanding compared derive from the non-epistemic (e.g. practical, prudential) value that these items have in addition to epistemic value (Ahlström-Vij, 2013).

Whatever the fate of the swamping problem, there are evidently going to be quite a few instances of epistemic goodness that apparently cannot be explained as instrumental to truth. These cases present problems for veritism, as it is a theoretical cost to deny that these are instances of epistemic goodness. In principle, it is open to veritists to bite the bullet of every problem case thrown their way. Just how costly that will be presumably depends on how good the attempts to explain away the seeming epistemic value as only apparent, or as non-epistemic, are. But veritists in fact often take a different tack. Note that the typical problem case arises on the assumption that instrumentality is the only way in which epistemic value can be derived. However, many epistemologists, including several veritists, clearly think that there are other ways. Goldman, himself a veritst, does not think that admitting weak justification as an epistemically good-making property threatens veritism. He notes that the relation between true belief and weak justification is “not a straightforwardly instrumental one”, but also that the best version of veritism plausibly “permit[s] a plurality of relations to the specified [fundamental] value”, and “subjective justifiedness exemplifies a distinct type of relation to the core veritistic desideratum” (2002, p. 66). In describing how many kinds of epistemic pluralism (including Goldman’s pluralism about justification) are in fact compatible with veritism, Pedersen (2017, p. 59) appeals to the notion of being “subjectively truth-connected”. Kvanvig (2003), who in the end plugs for epistemic value pluralism, nevertheless distinguishes between “effective” (i.e. instrumental) and “intentional” means as two kinds of derivative⁶ epistemic value.

Veritsts have rarely specified what this type of “intentional” or “subjective”, non-instrumental relation to truth consists in. Arguably, they need to do so if we are to grant that the problem cases can indeed be handled. Kurt Sylvan’s (2018) recent work is exemplary in this regard. Like Goldman, Sylvan thinks the best version of pluralism admits of more than one value derivation relation, and argues that veritists should take on board a form of non-instrumental derivative epistemic value constituted by the proper appreciation, or valuing of, an epistemic good.

The idea that proper appreciation, or more generally fitting attitudes, have non-instrumental value has been around for a long time in value theory (Brentano, 1889; Moore, 1903).⁷ Non-instrumental value is commonly equated with final value, which is in turn often glossed as value “for its own sake” or value “as an end in itself”, in contrast to value as a (mere) means to some end (i.e. instrumental value). Sylvan draws in particular on Thomas Hurka’s (2001) recursive definition of value. Hurka is concerned with securing a way for ethical consequentialists to endorse the final⁸ value of moral virtues. He proposes that “base-level goods” that have final value generate “higher-level goods” of having appropriate attitudes to those base-level goods, and these appropriate attitudes also have final value. For example, if pleasure is finally valuable, it is finally valuable to love pleasure. “Loving” some x can take several forms: desiring or wishing for x to obtain, pursuing and attempting to preserve, as well as taking pleasure in, x. Conversely, it is finally good to “hate” that which is finally bad, and moreover of final disvalue to love that which is finally bad, and to hate that which is finally good. As Hurka takes care to stress, the value of e.g. loving pleasure is not (merely) instrumental to the value of pleasure: loving pleasure is valuable independently of whether not doing so actually brings about any pleasure (2001, Chap. 1).

Sylvan summarizes this in what he calls Hurka’s Principle:

(Hurka’s Principle) When V is a non-instrumental value, proper ways of valuing V have some derivative non-instrumental value (2018, p. 383)

Not everyone will agree that fitting attitudes have final value. Howard (2018), for instance, claims that it is implausible to say that it is good for its own sake to envy the enviable or be annoyed by the annoying. However, it is not clear that this is an objection to Hurka’s principle. Hurka only claims that (at least) some fitting attitudes are valuable in themselves, namely those that are proper ways of (dis)valuing things with non-instrumental (dis)value. In any case, if as a veritist you’re not moved by something like Hurka’s Principle, the challenge is to present an alternative account of non-instrumental value derivation and we have yet to see that.

Now, Sylvan’s key idea is to enlist Hurka’s Principle in the interest of veritism as follows. Truth, being the fundamental epistemic good, by definition has non-instrumental epistemic value. And because it is proper – and hence non-instrumentally valuable in turn – to value what has non-instrumental value, it is non-instrumentally valuable to value, i.e. have certain pro-attitudes to, truth. The value of “valuing” – which is to be broadly understood to “include any way of being positively oriented toward something in act, attitude, or disposition” (Sylvan, 2018, p. 384) – truth is not instrumental: it is valuable in itself in the sense that its goodness does not derive from, and is independent of, any capacity to bring about e.g. true beliefs. But it is not fundamental, since it is explained in terms of a further epistemic good. Simply put, the valuing of truth is good because its intentional object is good. This last sentence will be central to my discussion in the next section. Therefore, let me just note that neither Hurka nor Sylvan are in the business of offering a fitting attitudes or buck-passing account of value. That is, the proposal is not to analyse value in terms of fitting attitudes, but simply a claim about what has value. In particular that (a) fitting attitudes are valuable for their own sake and (b) though non-instrumental, this value is derivative in the sense that it depends on the first-order goods that they are directed towards.

In endorsing the non-instrumental value of fitting attitudes, veritsts arguably have the resources to explain several otherwise problematic instances of epistemic goodness as derivative relative to truth. Goldman’s weak justification, and related instances of blameless but unreliable belief, can be explained as epistemically valuable in virtue of being manifestations of this proper valuing of accuracy. Epistemic virtues, on the responsibilist view, are characterised by their motivational profile which is often described in terms of “loving epistemic goods”, and the veritist could just insert “truth” as the relevant fundamental one which it is proper to value, thus accounting for the non-instrumental value of epistemic virtues.⁹ For sure, one may quibble over details in any particular case, but it is clear that the veritst needs some non-instrumental relation of value derivation in her explanatory toolbox. Sylvan’s account in terms of fitting attitudes is the most, perhaps the only, developed alternative on the market.

4 No reasons to value in insulated domains

Sylvan is not the first epistemologist to enlist the notion of non-instrumentally valuable attitudes. As noted, responsibilist virtue epistemologists characterise epistemic virtues as traits that manifest a love of epistemic goods. The value of these loving attitudes are crucial to the responsibilists’ account of the value of epistemic virtues: especially for those who reject reliability as a requirement on virtues of course, but even responsibilist who endorse it tend to hold that the value of an epistemic virtue is partly non-instrumental and is due to this intentional commitment to epistemic goods. Notably though, responsibilist virtue epistemologists typically do not subscribe to the IA. While they seldom take a stance on what the (fundamental) epistemic goods are, they typically take epistemic goods to be valuable simpliciter, and epistemic virtues to be genuine virtues – in Ridge’s Geachian terminology, they endorse predicative readings of “epistemically valuable” and “epistemically virtuous”.

In contrast, Sylvan claims to endorse the IA, and notes that this requires a modification of Hurka’s Principle, so that it applies to domain-specific value:

(Extended Hurka Principle) When V is a non-instrumental value from the point of view of domain D, fitting ways of valuing V in D and their manifestations have some derivative non-instrumental value in D (2018, p. 385).

Sylvan does not take the step from Hurka’s Principle to the Extended Hurka Principle to be a big deal. However, I think – and shall presently argue – that the Extended Hurka Principle should be rejected insofar as domain-specific value is taken to be independent of value simpliciter, as per (NOT).

Sylvan maintains in support of the Extended Hurka Principle that it predicts the plausible claim that the appreciation of beauty is non-instrumentally valuable from the aesthetic point of view – even if we assume that beauty does not have final value simpliciter. But beauty is a very common example of a value simpliciter, and I think many will find it difficult to bracket that assumption in investigating whether they find the claim plausible. Let us instead consider a domain where the fundamental good is not a plausible candidate for value simpliciter and see whether the Extended Hurka Principle seems plausible.

As Sosa observed, humans can and do evaluate just about anything. We can evaluate, say, torture techniques, strategies, actions, as good or bad from the torture point of view. It seems fair to assume that a fundamental good of the domain of torture-criticism is causing unbearable suffering in the victim. Being a fundamental domain-specific good entails having non-instrumental value in the domain, so a victim’s unbearable suffering is non-instrumentally good from the torture point of view. But what has derivative value in the torture-domain?

Certainly, things can have instrumental value in the torture-domain. Once we are clear on the fact that the aim of causing unbearable suffering in the victim is constitutive of the torture-domain – much as checkmating one’s opponent is a constitutive aim of chess – it is a straightforward, descriptive matter to evaluate agents, actions, properties, etc. with respect to whether they are efficient means of bringing about this aim. We can say that these things are good means, relative to that end. For instance, practicing my waterboarding technique is instrumentally valuable, because it will likely result in my realising the fundamental torture-goods more efficiently and/or to a greater degree. More generally, assuming that x is what we are after, means for x are good with respect to the aim of attaining x – and this is quite independently of the nature or value simpliciter of x. Instrumental goodness is merely a question of efficiency, and any x automatically “generates” derivative instrumental value for everything which is a means to x.¹⁰

The Extended Hurka Principle predicts that proper ways of valuing unbearable suffering of the victim are also derivatively, non-instrumentally valuable from the torture point of view. It is not obvious what would count as a proper way of valuing in this domain, but encouraging and taking pleasure in are two stock examples of fitting attitudes to something good, so it seems the Extended Hurka Principle might predict that it is non-instrumentally valuable from the torture point of view for e.g. a torturer to take pleasure in the causing of insufferable pain in the victim.

This lacks the intuitive pull of Sylvan’s aesthetics example. My hypothesis is that this is because beauty has high plausibility as a candidate for value simpliciter whereas the unbearable suffering of a torture victim does not.¹¹ But why should this matter, if we are considering the epistemic domain as insulated? Why can things derive domain-specific instrumental but not non-instrumental value from domain-specific goods?

The short answer is that the fitting attitudes at issue here – proper ways of valuing and disvaluing – involve genuine normativity, or objective reasons.¹² If x is good, I have reason to value x, whether or not I actually do. Ethicists tend to agree that goodness is that which it is appropriate to be positively oriented towards – some go as far as analysing goodness in those terms (so-called fitting attitudes accounts of value), but many more affirms it as a true description of goodness and that is all we need here. The plausibility of Hurka’s Principle turns centrally on this normative connection between goodness and certain attitudinal orientations: it seems correct that we ought – in a genuine sense – to value what is good. But we should not expect insulated domains generally to deliver such a connection because the “goods” around which they are structured need not be, and often are not, genuinely good and so they do not generate reasons to value them. Again, certain attitudes are valuable because they are fitting, and it is only fitting to value what is good.

I expect it will now be objected that domain-specific goods do generate reasons to hold certain attitudes to them – but these reasons are domain-specific, just like the “good” in question. But can we really make sense of the needed notion of a domain-specific reason? Domain-specificity is sometimes phrased in terms of a “point of view”, but it is not at all clear to me what this means when it comes to fittingness. What is the “point of view” of a domain? It is presumably not someone’s point of view, e.g. that it is proper from the torture point of view to value suffering cannot simply be a matter of it seeming fitting to someone who is deeply immersed in the domain and has adopted its constitutive aim. To the extent that the constitutive aim of the torture-domain has a point of view, we might perhaps say that it is valuable “from the point of view” of that aim to adopt certain pro-attitude towards it because doing so is an efficient means to realizing the aim. For instance, perhaps taking pleasure in suffering makes one a more efficient torturer. But this only gives the attitude instrumental value – we were after goodness in virtue of fittingness, or propriety.

What one would need to do – and what I think will be very, very difficult – is to pin down a sense of objective, agent-neutral propriety, but with a somehow limited scope so that it applies only in an insulated domain. The best attempt I can come up with is to appeal to something like a domain-specific norm which dictates that one ought to value or disvalue certain things for their own sake.¹³ Since it is unclear what the norms are, or would be, in the torture-domain, let’s switch to a somewhat different example of valuing something which is (or so it seems to many of us) not a good candidate for a value simpliciter. Say that in some conservative Christian contexts, norms prescribe that it is fitting to hold certain pro-attitudes to sexual chastity (e.g. respecting, wishing for it to obtain). Plenty of people embedded in those contexts agree that it is fitting to value chastity, they criticise those who fail to value chastity, etc. But, as noted, many will agree that chastity is not valuable simpliciter. Yet, is there then not a sense in which we can say that from the point of view of this type of conservative Christian domain, it is fitting to value chastity, even if it is not fitting (and hence not valuable) simpliciter? This is a perfectly familiar type of case: those who engage in in the domain of neo-Nazism consider it fitting to pursue and appreciate the domain-specific fundamental “good” white supremacy; within the homophobic domain, there are norms according to which it is fitting to despise the domain-specific fundamental “bad” homosexuality. Might it not be just this that is meant by domain-specific fittingness?

I think these examples fall short of establishing a sense of objective, agent-neutral propriety with limited scope, for the following reason: arguably, they are all examples of where people who endorse and live in accordance with the norms take them to latch on somehow to genuine value. That is, the reason there seems to be a sense of fittingness at issue in the domain of neo-Nazism is that from the point of view of neo-Nazism, white supremacy is valuable simpliciter (or at least is connected to value simpliciter), and similarly for the other examples in the paragraph above. Differently put, the norms dictating propriety of certain attitudes seem to assume (albeit wrongly in these cases) that the intentional object of the attitudes have genuine (dis)value. In contrast to people who are knowledgeable about coffee or archery (Sosa’s examples of insulated evaluative domains), people who will criticise others with reference to, and who live in accordance with, the propriety norms of e.g. neo-Nazism or conservative Christianity will typically not admit that these are insulated domains of evaluation. In other words, these domains where we seem to find local notions of fittingness are not clear examples of insulated evaluative domains, such as the domains of chess, coffee, or archery. In those properly insulated domains, the notion of fittingness seems to play much less of a role, and to the extent that such domains do contain norms that dictate that certain valuing attitudes are fitting to take to e.g. the good shot or to checkmating one’s opponent, this can likely be explained by a correlation to something that is plausibly of value simpliciter, e.g. some might think it fitting to admire the good shot because it constitutes an achievement in the relevant domain and achievements valuable simpliciter. These fittingness norms are local only in the sense of what the relevant type of achievement is, not in the sense that the good in question is domain-specific.

The upshot is as follows: if domain-specific propriety can only be found in domains that are taken to overlap with value simpliciter, this supports my hypothesis that the non-instrumental value derivation through fitting attitudes is not compatible with the insulation approach to epistemic value. It is hard to show that there are no plausible cases featuring evaluative domains that include norms that dictate certain pro-attitudes to the domains’ central “goods” that at the same time are clear examples of insulated evaluative domains also perceived as such. But my intention with the above reasoning is to at least shift the burden of proof to the friend of IA, and urge them to present a plausible example where we can readily make sense of domain-specific fittingness, in the same unproblematic sense that we can all readily make sense of instrumental goodness in a domain quite independently of whether we or anyone in fact adopts the constitutive aim of that domain. Pending such an example, I conclude for now that the Extended Hurka Principle is – given (NOT) – not an innocent extension of Hurka’s Principle and that it should be rejected by insulationists. For veritists, this means that if they want to appeal to the non-instrumental value of fitting attitudes in explaining instances of epistemic goodness as derivative of the value of accuracy, they need to give up the IA. If they instead insist on hanging on to it, they must forego this strategy for explaining several compelling instances of alleged epistemic goodness.

4.1 A weaker IA?

Someone might at this point object that some philosophers who endorse the IA phrase the disconnect between epistemic value and value simpliciter in terms of final value specifically, and that they might not want to commit themselves to anything as strong as (NOT). Indeed, Sylvan writes¹⁴:

I am only interested in defending a version of Veritism on which accuracy is the sole fundamental value from the epistemic point of view. It is compatible with this view that accuracy might have no non-instrumental value from any non-epistemic point of view (2018, p. 385, my emphasis).

So, what if the assumption behind the Extended Hurka Principle is not (NOT) but instead

(WEAK-NOT) If x has non-instrumental epistemic value, it does not follow that x has non-instrumental value simpliciter.

I have argued that the Extended Hurka Principle is false given (NOT), and so veritists who endorse (NOT) cannot appeal to the derivative non-instrumental value of ways of valuing truth. For ways of valuing truth to be of non-instrumental value, truth must be valuable simpliciter. But what about (WEAK-NOT)?

Someone who endorses (WEAK-NOT) can indeed say that proper ways of valuing truth have non-instrumental value – at least if they are prepared to follow Hurka’s account beyond Hurka’s principle in particular. As Hurka notes, once his recursive definition of value is taken on board, a number of other “higher order” (i.e. derivative) values are generated too – beyond the one captured by Hurka’s principle. In particular, if x is a means to a final good y, it is fitting – and hence non-instrumentally valuable – to love x as a means to y (2001, p. 17). That is to say, valuing truth can have non-instrumental value even if truth has only instrumental value simpliciter. But importantly, it is only fitting, and hence only valuable in this sense, to value truth for what it is, namely as a means (to, ultimately, something with non-instrumental value). This is clearly not what Sylvan had in mind, and it remains to be shown that the veritist could handle the problem cases of e.g. subjective justification and unreliable virtues as ways of valuing truth instrumentally to some (non-epistemic)final good.¹⁵ But in principle, can (WEAK-NOT) enable the veritists to have their insulationist cake and eat their non-instrumentally valuable ways of valuing too?

Well, no. Because saying that proper ways of valuing truth have non-instrumental derivative value – even if one refrains from assuming that truth has final value simpliciter – still requires one to make assumptions about truth’s value simpliciter. In particular, you can, as per (WEAK-NOT), stay neutral on whether truth is finally valuable simpliciter and therefore proper to value, or whether it is merely a means to something that has final value simpliciter and therefore is proper to value as a means to that further value. But in order to claim that proper ways of valuing truth have non-instrumental value, you must assume that at least one of these disjuncts are correct. That is, it matters to epistemic axiology that one of them must be correct, and moreover it matters which one is correct – because that determines which ways of valuing truth are non-instrumentally valuable (i.e. whether it is proper to value truth for its own sake or to value it as a means). This violates the core idea underlying the IA, which was that theorising about epistemic axiology can proceed without assumptions about value simpliciter. So, (WEAK-NOT) does not capture the core idea of the IA, and one can only get non-instrumental value derivation given (WEAK-NOT) by violating that core idea, viz. by making assumptions about how epistemic value relates to value simpliciter.

5 Why insulate?

I have argued above that the IA robs veritists of a very useful explanatory tool. That gives veritists a good reason for giving up the IA, all else equal. But is all else equal? That is, are there any potentially overriding reasons for veritists to endorse IA, and risk foregoing the explanatory tool provided by fitting attitudes to epistemic goods? Unfortunately, those who subscribe to IA very rarely disclose their reasons for doing so. Nevertheless, in this section I will consider a couple of potential concerns that might motivate the IA and evaluate them. My tentative conclusion will be that the overall picture from the previous section remains: veritists should not be insulationists.

First, a veritist might doubt the very notion of value simpliciter. This would be in the vein of Geach’s suggestion that value-terms like “good” and “bad” have only attributive use, i.e. it makes no sense to speak of something as good (or e.g. valuable) period, a suggestion taken up by a few other philosophers (e.g. Foot 1985; Thomson, 1997), although for forceful criticism see e.g. Hurka (1987) and Pigden (1990). But this doesn’t motivate accepting IA: if there are no values simpliciter, what are we then isolating epistemic value from?

Second, a veritist might accept the notion of value simpliciter but doubt that truth has such value, hence wanting to endorse (NOT).¹⁶ Here it is important to distinguish between two different worries. On the one hand, the veritist might doubt whether truth has any value simpliciter at all. Although I think this is a relatively uncommon view, for veritists who do hold it, it arguably constitutes a reason for endorsing the IA. But it also blocks these veritists from appealing to the non-instrumental value of valuing truth. On the other hand, and I think this is much more common, the veritist might merely doubt that truth has final value simpliciter. But as I argued in Sect. 4.1, that does not in itself amount to accepting the IA. As long as one is prepared to grant that truth has at least instrumental value simpliciter, it is possible to be a veritist without being an insulationist, and indeed this position (unlike the IA) – allows one to hold that it is epistemically valuable to value truth (albeit as a means).

Third, the veritist might worry that epistemic value is not a genuine kind of value simpliciter. This appears to be Ridge’s (2013) reason for recommending the attributive reading of ‘epistemically valuable’. He suggests that kinds of value are distinguished in terms of the kinds of reasons associated with them. For instance, something is aesthetically valuable if it has properties which provide reasons for appreciating it. On this picture, identifying epistemic value as a distinct kind of genuine value requires explaining what sorts of reasons are associated with specifically epistemic value. Ridge notes with frustration that epistemologists have largely failed to address this question but he tries out the suggestion that epistemic reasons are reasons for believing, and finds it wanting. This in itself gives us little reason to conclude that there is no sensible way of delineating an epistemic kind of value simpliciter. For one, Ridge’s paper only investigates on what reading the claim that knowledge is distinctively epistemically good¹⁷ is most plausible (and whether it is plausible enough to pose an ex ante explanatory requirement on theories of knowledge). The predicative reading (and the “reasons to believe” gloss) is rejected because it does not render the claim in question plausible. Moreover, in addition to only evaluating it in this one respect¹⁸, he only evaluates this one proposal of how to delineate epistemic value, and there are presumably other and perhaps better proposals that need to be considered.

Of course, one might have other reasons for doubting that an epistemic kind of value simpliciter can be delineated. But even this does not take one straight to the IA. One may affirm that being of epistemic value entails being of value simpliciter, but not of a distinctly epistemic kind. Epistemic value might be intimately connected with prudential value for instance, but there are presumably many other options to explore as well. This position is also incompatible with (NOT), and again, a veritist holding this position can also avail herself of the derivative non-instrumental value of fitting attitudes (albeit of e.g. truth as related to other things of value simpliciter).

Insofar as one wants to move away from pure instrumentalism about epistemic value derivation, one cannot remain neutral on these relations, and theorising about epistemic axiology and its internal structure cannot happen in isolation from questions of value simpliciter. Such theorising becomes intimately connected with, indeed part of, value theory. This may or may not be a problem for veritists (compared to e.g. epistemic value pluralism, and to other forms of epistemic value monism). It is interesting to note that all these potential motivations come from substantial views or worries about simpliciter values, considering the key idea behind IA: that such considerations do no matter to theorising about epistemic value. Perhaps that suggests that veritists themselves are worried that it will be a problem to make veritism plausible once the IA is rejected. I have suggested in this section that several views on how epistemic value relates to value simpliciter, all compatible with rejecting the IA, are in principle available to veritists, but how viable they are in the end remains to be seen once veritists start to spell out how truth and the epistemic relates to (other) things of value simpliciter.¹⁹

6 Conclusions

For several compelling instances of alleged epistemic goodness that are not instrumental to attaining truth, veritists must either present an account of how these instances can derive their value non-instrumentally from the value of accuracy, or claim that these are not, after all, instances of epistemic goodness. Since the latter would constitute a serious blow to veritism, Sylvan helpfully proposes that veritists can appeal to the derivative value of certain fitting attitudes in explaining these instances of epistemic goodness. I have argued that this strategy is incompatible with the widespread assumption (among veritists and others) that having epistemic value does not imply or depend on having value simpliciter. Therefore, veristists who are moved by problem cases featuring epistemic goodness that cannot be explained in terms of instrumentality to truth, and who want to avail themselves of non-instrumental value derivation have a strong reason to abandon the insulation approach to epistemic value. Spelling out just how epistemic value relates to (other kinds of) value simpliciter may bring its own problems for veritists, but there are presumably several different options in principle available for them to explore.

It is worth highlighting the following wider implication of what I have argued: anyone who wants to affirm that we have reasons to value epistemic goods because they are epistemic goods, and that it is thereby of non-instrumental value to do so, needs to steer clear of the insulation approach to epistemic value. Instrumentalists about epistemic value derivation, in contrast, could happily continue to endorse it. However, if being a means to bringing about a more fundamental epistemic good is the only way in which something can be of derivative epistemic value, the pressure to endorse epistemic value pluralism increases – that is, in order to explain all instances of epistemic goodness, one might need to postulate more things as fundamentally epistemically valuable besides e.g. truth.²⁰