This contribution to the symposium on Annalisa Coliva’s Extended Rationality is largely sympathetic with the moderate view of the structure of epistemic warrant which is defended in the book. However, it takes issue with some aspects of Coliva’s Wittgenstein-inspired ‘hinge epistemology’, focussing especially on her conception of propositional warrant, her treatment of epistemic closure, her antirealist conception of truth, and the significance of her answer to so-called Humean scepticism.
Extended Rationality is a remarkable contribution to our understanding of the structure of epistemic warrant (justification). In her book Annalisa Coliva discusses the main philosophical views on the topic and forcefully advocates her own ‘moderate’ third way between ‘conservative’ and ‘liberal’ accounts. Extended Rationality is full of insightful ideas not only on the structure of epistemic warrant, but on related subjects such as epistemic closure, transmission of warrant and easy knowledge, and these ideas are woven together in a comprehensive account of the nature of epistemic rationality. Perhaps I should say from the beginning that I agree with many of the arguments and claims that are offered in the book—indeed, I would happily describe myself as a ‘moderate’ in Coliva’s sense. Fortunately for this symposium, however, I do not agree with all the details of the Wittgenstein-inspired ‘hinge epistemology’ she puts forth as her favourite version of moderatism, so in my contribution I propose to raise four different (if not all equally pressing) issues for the particular way in which she develops the moderate outlook.
First of all, however, I want briefly to draw attention to the aspects of Coliva’s view that I find most congenial. As I said, I am wholly sympathetic with her attempt to steer a middle course between the liberal and conservative accounts of the structure of perceptual warrant. That is, I agree that the possibility of acquiring perceptual warrant for ordinary propositions like
Let me turn, then, to those aspects of Coliva’s epistemological package that I find more questionable. The first has to do with her notion of propositional warrant and the way she employs it in her treatment of the connected closure principle. The distinction between propositional and doxastic warrant (or between warranted and well-founded belief) is commonplace in current epistemological literature. There is some debate on the requirements a belief must satisfy in order to count as doxastically warranted, and the orthodox view that doxastic warrant should be explained in terms of propositional warrant plus a ‘basing’ relation has recently come under some attack. However, talk of propositional warrant is usually thought to be quite harmless—at least insofar as saying that believing that p is propositionally warranted for S at t is taken as no more than a convenient way of saying that S has good reasons for believing that p at t, whether S forms the relevant belief or not. Now, the way Coliva talks of propositional warrant may seem perfectly standard, as she introduces the notion by saying that the kind of epistemic warrant that provides the focus for her book is “whatever speaks or can be considered to speak to the truth of a given proposition … independently of whether that proposition is the content of a subject’s actual belief” (Coliva 2015: 16). However, she also claims that propositional warrant is independent “of which further beliefs a given subject may actually have,” and repeatedly equates it with warrant “in the abstract space of reasons” (Coliva 2015: 16). I shall set aside for the moment the significance of this Sellars-inspired metaphor which Coliva borrows from Martin Davies (see, e.g., Davies 2009: 338). But the claim that propositional warrant is independent of any further beliefs an epistemic agent may have is obviously incompatible with the idea that what an agent is propositionally warranted to believe is partly a function of the (warranted) beliefs (s)he happens to have. This idea is far from uncontroversial—a natural way to resist it is to maintain that which propositions an agent is warranted to believe at a given time does not depend on the warranted beliefs (s)he happens to have, but merely on the propositions (s)he is warranted to believe, whether (s)he believes them or not. However, the less idealised view that what an agent is propositionally warranted to believe is at least in part a function of the (warranted) beliefs (s)he happens to have is widespread enough to make it inadvisable, I think, to build its rejection into the very definition of propositional warrant. And my worry is that Coliva’s choice may not be just a mere terminological quirk.
For one thing, taking propositional warrant to be independent of any belief the agent may actually have is in tension with the moderate claim that having propositional warrant for ordinary empirical propositions depends on assuming such (unwarrantable) propositions as
Coliva’s view is that this principle is not unconditionally valid. She argues that closure of warrant fails whenever the following two conditions are met: (i) assuming the conclusion of the relevant argument is necessary in order to have warrant for the argument’s premise(s) and (ii) the argument’s conclusion cannot itself be warranted, evidentially or otherwise, because it is a heavy-weight proposition (Coliva 2015: 103). However, she upholds the principle in all other cases: if no heavy-weight propositions are involved, it is true, she maintains, that, if S has warrant for P and knows that P entails Q (and competently deduces Q from P), then S has warrant for Q—and this is so whether S’s having warrant for P depends on S’s having independent warrant for Q or not. One of the arguments Coliva discusses in this connection is FINGERS (Coliva 2015: 103):
Closure principle (for warrant)If S has a warrant for P, S knows that P entails Q, and competently deduces Q from P, then S has warrant for Q. (Coliva 2015: 101)
This argument is relevant for the topic because, if an agent has warrant for (i), knows that (i) entails (iii)2 and competently deduces (iii) from (i), then, according to the closure principle for warrant, (s)he has warrant for (iii). In this case, Coliva agrees with the principle: her view is that FINGERS, like countless other arguments that do not traffic in heavy-weight propositions, yields no counterexample to the closure of warrant. Interestingly enough, however, her rationale for taking this view brings us back to the notion of warrant ‘in the abstract space of reasons.’ For she maintains that, whether or not the argument is warrant transmitting (which she thinks it is), “the fact that, at least in the abstract space of reasons … there is a warrant for (iii), which is independent of FINGERS itself, guarantees that if (i) is warranted and (ii) is known, (iii) is also warranted” (Coliva 2015: 104). This statement strikes me as odd. For one thing, even if the fact (let us grant, for the sake of argument, that it is indeed a fact) that in the abstract space of reasons there is independent warrant for (iii) does guarantee the truth of the relevant instance of the closure principle, it also guarantees the truth of the conditional (say) that if an agent has no warrant for (i), then (s)he has warrant for (iii), as well as of any other conditional with the same consequent. So it seems clear that citing the fact that in the abstract space of reasons there is independent warrant for (iii) can explain why the closure principle holds in the case at hand only at the cost of trivializing it. More importantly, however, what does it mean to say that there is a FINGERS-independent warrant for (iii) ‘in the abstract space of reasons’? Surely, it is a contingent fact that an epistemic agent considering the number of John’s fingers at a given time has independent evidence to the effect that he has ten fingers. Suppose Ann is such an agent—an agent, that is, that is reasoning her way from the premises to the conclusion of FINGERS. Suppose, further, that the only independent evidence she can rely upon to form a belief as to the number of John’s fingers is that he was recently involved in a serious car crash. Now, it seems clear that (iii) is not independently warranted for Ann at the time when she is running through FINGERS. There is, then, room to doubt that Coliva’s conception of propositional warrant is really standard; indeed, the claim that (iii) is warranted “at least in the abstract space of reasons” makes one suspect that a proposition’s being so warranted is very close, in her view, to its being warrantable in principle—a suspicion that is borne out by her willingness to treat warrant ‘in the abstract space of reasons’ as a kind of warrant that a proposition can enjoy for an agent at a time even if it is in fact inaccessible to the agent in question, being accessible perhaps “only by epistemologists” (Coliva 2015: 30, 31). Be that as it may, it seems clear at least that a notion of propositional warrant according to which Ann may be said to be independently warranted to believe that John has ten fingers in an epistemic situation like the one that has just been described abstracts away from so many features of the agent’s epistemic situation that the restricted closure principle it helps to preserve ends up concerning an epistemic good that is quite remote from most ordinary notions of epistemic warrant.
(i) John has two hands;
(ii) If John has two hands, he has ten fingers;
(iii) John has ten fingers.
Fortunately, however, there is no need to maintain that FINGERS yields no counterexample to the closure principle for warrant because (iii) is independently warranted ‘in the abstract space of reasons.’ The reason why FINGERS yields no such counterexample is rather that having warrant for (i) depends neither on having independent warrant for (iii) nor on assuming it, and so FINGERS is a perfectly warrant transmitting argument; as a consequence, any agent who has (propositional) warrant for (i), knows that (i) entails (iii) and competently deduces (iii) from (i), will thereby have (propositional) warrant for (iii). This explanation does not extend to the undesirable conditionals whose truth is guaranteed by Coliva’s claim that (iii) is warranted in the abstract space of reasons and does not invoke a notion of propositional justification that abstracts away from so many features of the agent’s epistemic situation. Whatever one thinks of the highly idealised notion of epistemic justification Coliva seems to be working with, the explanation I am proposing is perfectly consistent with her own brand of moderatism, and I suggest that she might welcome it as an acceptable fix for the problem I have been discussing.
So much for the first issue. The second has to do with closure too, specifically with Coliva’s claim that moderates, though denying the general validity of the closure principle far warrant, have “another good in stock, for which closure does not fail. Namely, the rational mandate to assume those very general propositions that are necessary in order to have empirical warrants in the first place, which stems from the endorsement of the extended rationality view” (Coliva 2015: 137). That there is such a rational mandate is a claim that I do not think Coliva succeeds in establishing to the satisfaction of the sceptic (more on this later). But let us grant, for the moment, that she is right that we are rationally mandated to assume those unwarrantable basic propositions that apparently defy closure. Now, it is not clear to me in what sense such a rational mandate would be closed under known entailment. For according to Coliva’s own lights, we are not rationally mandated to assume, but warranted to believe the countless ordinary propositions for which we have acquired empirical warrant. Is then the idea supposed to be that epistemic warrant is just a species of the genus rational mandate? Coliva never says so. On the contrary, she argues that “a rational mandate is not an epistemic warrant,” because it is not “an epistemic good that speaks to the truth of what it is meant to warrant” (Coliva 2015: 11). To be sure, by this she might just mean that a rational mandate is not as such an epistemic warrant (an epistemic good that ‘speaks to the truth’ of the relevant proposition), although some rational mandates (those that do so speak) are. But this does not seem to be the natural way to construe her view, and in any case a warrant is the sort of thing that gives someone the right to do something, while a mandate, at least in the sense that is at issue in Coliva’s discussion, is the sort of thing that puts someone under the obligation to do something. So the idea might be instead that there is a higher genus of which both epistemic warrant and rational mandate are species, or else that the good that is closed under known entailment is to be intended disjunctively as the status that is enjoyed by a proposition when it is either epistemically warranted or rationally mandated. However, both suggestions run counter the letter of Coliva’s proposal: her claim is that the good for which closure does not fail just is “the rational mandate to assume those very general propositions that are necessary in order to have empirical warrants in the first place.” The impression, then, is that here there is room—and need—for clarification. My own feeling is that the claim that there is a good for which closure does not fail is not really indispensable to the moderate account of the structure of epistemic warrant; perhaps it is just an addition that, in the end, friends of the account might conveniently drop.
Here I will not discuss all the details of the ‘extended rationality’ view, nor will I consider Coliva’s ingenious attempt to extend it outside the domain of perceptual warrant, but I will just raise a few worries about the notion of truth that is applied to those assumptions and about the way the view is put to work in her discussion of scepticism.
… epistemic rationality extends … to those very assumptions that make it possible to produce ordinary perceptual justifications and that therefore make it possible to have the kind of practice of forming, assessing, and withdrawing from empirical beliefs on the basis of perceptual evidence, which is itself constitutive of our very notion of epistemic rationality. If so, it turns out that we are actually mandated by epistemic rationality itself to assume “There is an external world.” (Coliva 2015: 11)
The issue I want to raise in relation to the first topic concerns the claim that the truth of the assumptions that are alleged to be constitutive of epistemic rationality is best interpreted in the minimal, anti-realist fashion described in the penultimate section of Chapter 4 (Coliva 2015: 147–150). There Coliva attempts to reconcile the acknowledgment that, when it comes to such assumptions, rationality and truth “may actually come apart” (Coliva 2015: 148) with the (anti-sceptical) claim that, in the end, their truth cannot lie “beyond our possibility of recognition” (Coliva 2015: 149). The solution she proposes is designed to escape the alternative between construing truth in a robustly realist way and construing it in a traditional anti-realist fashion. For construing the truth of those assumptions in a robustly realist way (as if it were some sort of correspondence to mind-independent facts or states of affairs) would inevitably put it beyond any possibility of recognition, while construing it in traditional epistemic terms (as if it were some sort of idealized rational acceptability, superassertibility, or the like) is precluded by the distinctive moderate thesis that heavy-weight propositions are inherently unwarrantable. Coliva’s proposal is then that their truth be understood in the “minimal way” suggested by the recognition that it “depends merely on the kind of role they play in our epistemic system” (Coliva 2015: 149). To say that heavy-weight propositions are true is to say “that what they state is how things are, given our overall Weltbild,” where the point of the qualification is that the nature of their (minimal) truth is, ultimately, of an anti-realist kind, “[n]ot because it depends on evidence, but because it is seen as dependent on how we, human beings, experience the world around us” (Coliva 2015: 149; italics in the original).
I see two problems with this proposal. The first arises from the well-known circumstance that anti-realist accounts of truth are appropriate only for those areas of thought or discourse for which it is a priori that all true propositions are in principle knowable (Wright 1992: 57–61; Lynch 2009: 42–43). The trouble is, of course, that the basic assumptions whose truth Coliva would have us understand in an anti-realist way are declaredly not warrantable and therefore not knowable (even in principle), which seems to thwart any attempt to characterize their truth in an anti-realist fashion. Coliva does not discuss this problem, but an important feature of her proposal might help to address it. We have seen that her view of the kind of truth enjoyed by heavy-weight propositions is presented not only as anti-realist, but as non-epistemic. This runs counter the popular wisdom that (non-deflationary) rejection of alethic realism boils down to acceptance of an epistemic conception of truth. However, Coliva’s suggestion is not obviously unacceptable (Künne 2003: 20–27), and accepting it might help to mitigate the present worry. Sure enough, it will not do to straightforwardly argue that the preoccupation is misplaced because it is just epistemic accounts of truth that presuppose that all true propositions in the relevant area of thought or discourse be in principle knowable. For it is not just epistemic accounts of truth, but (non-deflationary) alethic anti-realism that carries with it a commitment to the idea that truth is epistemically constrained. So more work will have to be done to show that this argumentative strategy can lead to the desired conclusion. Here I can only suggest a direction it might plausibly take. It seems to me that it will be necessary to reconceptualise the notion of an epistemic constraint so as to show that the (non-epistemic) truth of the basic assumptions that are supposedly constitutive of epistemic rationality is, after all, subject to such a constraint—not, perhaps, in the usual sense that all true propositions should be knowable in principle (scilicet, in the areas of thought or discourse in which truth is construed in an anti-realist fashion), but, say, in the weaker sense that it should not outrun epistemic rationality. The next step will of course be to show that meeting this constraint is sufficient for the property ascribed to those assumptions that are such “that what they state is how things are, given our overall Weltbild,” to qualify as a truth-property. As far as I see, this strategy is not obviously a nonstarter. Here, however, I cannot pursue it any further, and must be content with having sketched it in its broadest outline.
The second problem I see with Coliva’s proposal to characterize in an anti-realist way the truth of the basic assumptions she takes to be constitutive of epistemic rationality is connected to the fact that such assumptions are either entailed by the ordinary propositions for which they make it possible to acquire perceptual warrant (as is the case with
As a consequence, Coliva argues, “both sceptics and non-sceptics alike are required by the lights of epistemic rationality itself to assume that there is an external world” and other like propositions, whose acceptance is therefore “mandated by a shared notion of epistemic rationality” (Coliva 2015: 129–130; italics in the original).
… the notion of epistemic rationality shared by sceptics and non sceptics alike—like many other notions—does not hang in the air, but depends on our practices. In particular, the notion of epistemic rationality depends on the basic practice (or method) of producing, assessing, and withdrawing from ordinary empirical beliefs, such as “Here’s a hand,” “This wall is red,” and so on, interpreted as being about mind-independent objects, based on the deliverances of our senses.
To be sure, showing that assuming such propositions is mandated “by a shared notion of epistemic rationality” is not showing that we are warranted to believe them. That is why the response to Humean scepticism provided by the ‘extended rationality’ view is described by Coliva as merely “indirect,” that is to say, as a response that does not provide the warrants required by the sceptic, but argues instead that the request to provide them “depends on too narrow and unmotivated a conception of epistemic rationality …, which confines epistemic rationality only to warranted beliefs.” Against this conception, she maintains that epistemic rationality “actually extends to those assumptions which, while unwarrantable, make the acquisition of perceptual warrants possible in the first place” (Coliva 2015: 150). So the upshot of the response to Humean scepticism provided by the ‘extended rationality’ view is claimed to be that the “devastating consequences” that sceptics would like to draw from the acknowledgement that those basic assumptions are unwarrantable “can in fact be blocked” (Coliva 2015: 127). Somewhat unexpectedly, however, the consequences Coliva has in mind do not concern ordinary empirical propositions, but the very assumptions that, according to the moderate view, enable us to acquire perceptual warrant for propositions like
The reason why Coliva maintains that the sceptical challenge to the possibility of acquiring warrant for ordinary empirical propositions can be easily answered is of course that, on the moderate account of the structure of perceptual warrant, such possibility merely depends on the acceptance of the assumptions in question. And of course she has already argued in the first two chapters of Extended Rationality that the moderate account is superior to all its competitors, including the conservative view with its claim that the acquisition of warrant for ordinary empirical propositions depends on the possession of independent warrant for those assumptions. So it turns out that the Humean sceptic whose stance is at issue in the section of the book where Coliva claims that sceptics themselves must accept such assumptions is the sort of ‘non-conservative’ sceptic that would “grant, at least for the sake of argument, that no warrant for ‘There is an external world’ is needed to have an evidential justification for ‘Here is a hand’” (Coliva 2015: 198 n. 2; see also 128, 199 n. 18). The stance of the different sort of sceptic who would draw relativistic conclusions from the arbitrariness of those unwarranted assumptions is discussed in a later section, where the claim that “there may be … many different systems of assumptions, which are mutually incompatible and yet are on a par, that give rise to different and equally valid systems of justification” (Coliva 2015: 140) is argued to be ultimately untenable.
Here, however, one cannot help the feeling that something has been neglected. For the nub of Coliva’s attack on conservatism is in fact that the conservative ideal is unachievable (see, e.g., Coliva 2015: 83–84). And this is a point which may well bring grist to the moderate mill (that is, if it is admitted, as it is admitted by conservative, liberal and moderate epistemologists alike, that in certain conditions our sensory experience is capable of producing warrant for ordinary empirical propositions), but which is unlikely to upset a ‘conservative’ Humean sceptic, who (unlike Coliva’s ‘moderate’ Humean sceptic) merely aimed to show that no perceptual warrant for ordinary empirical propositions is forthcoming in the first place. So it seems to me that the way Coliva ignores the sort of Humean sceptic that would insist that having independent warrant for propositions like
To see how this may be so, let us grant (as I am happy to do) that the moderate account of the structure of perceptual warrant is indeed superior to its liberal and conservative rivals, and that it is superior to them for precisely the reasons offered in Extended Rationality—first and foremost, because it provides an explanation of our capacity to surpass our cognitive locality which does not beset us with the exceedingly exacting task of producing a warrant for the basic assumptions that underlie the acquisition of perceptual warrant for ordinary empirical propositions. However, we have seen that it is not the moderate account as such, but the wider framework of the ‘extended rationality’ view that allows Coliva to argue that sceptical objections presuppose the adoption of too narrow a notion of epistemic rationality (Coliva 2015: 120, 130, 150). According to the sceptic, “it is epistemically rational to believe only evidentially warranted propositions” (Coliva 2015: 130), while on the ‘extended rationality’ view it is also epistemically rational “to accept those unwarrantable assumptions that make the acquisition of perceptual warrants possible in the first place and are therefore constitutive of ordinary evidential warrants” (Coliva 2015: 129). Now, it seems to me that at this juncture a Humean sceptic might legitimately ask whether the arguments offered on behalf of the ‘extended rationality’ view establish anything stronger than the conditional that if there are any evidentially warranted propositions, then it is epistemically rational to believe them and to accept those (unwarrantable) assumptions that make it possible to acquire warrant for them. For a Humean sceptic might happily concede this conditional and yet insist that, if the upholder of the ‘extended rationality’ view is to show that sceptics and non-sceptics alike are required by epistemic rationality to accept such assumptions, what (s)he needs to show is not just that the practice of producing, assessing and withdrawing from ordinary empirical beliefs on the basis of the deliverances of our senses depends on accepting them, but that it does yield perceptual warrant for those beliefs—and for those beliefs interpreted, in Coliva’s own words, “as being about mind-independent objects.”
Let me sum it up this way. Coliva expressly restricts her discussion to the challenge of the non-conservative Humean sceptic who acknowledges the warrant-conferring power of the practice in question. However, it is far from clear that her forceful criticism of the conservative account of the structure of perceptual warrant provides a refutation of the position of the non-relativist sceptic who would rather question such warrant-conferring power. For the arguments she offers for preferring moderatism to conservatism all turn on the impossibility of producing a warrant (be it a priori or a posteriori, evidential or non-evidential) for those basic assumptions that, according to conservative epistemologists, put us in the position to acquire perceptual warrant for ordinary empirical propositions only if they are themselves warranted; so, to repeat, such arguments bring grist to the moderate (as opposed to the sceptical) mill only if it is admitted that there are conditions in which the practice in question is indeed capable of producing warrant for ordinary empirical propositions. If that is denied or doubted, it seems to me that Coliva’s arguments cannot be taken to justify a preference for non-sceptic moderatism over sceptic conservatism.
So I do believe that Extended Rationality makes out a very strong case for the superiority of the moderate view over its non-sceptical rivals. However, if arguing that liberal views are unable to account for our capacity to surpass our cognitive locality and that the conservative ideal is actually unattainable may support the conclusion that the moderate position has, in her own words, “all the credentials to qualify as the best possible explanation of how we can have the perceptual warrants we think we have” (Coliva 2015: 85), what needs to be shown to meet the (ordinary) Humean sceptical challenge is, in addition, that we do have the perceptual warrants we think we have. This is a task that the ‘extended rationality’ package in its present form does not seem to me to be able to accomplish. My tentative conclusion is then that Coliva’s treatment of Humean scepticism may well be an illuminating exercise of ‘damage limitation’ which enables us to explain to ourselves how we can acquire the perceptual warrants we think we have; but it is an exercise which, as far as I can see, falls short of establishing to the satisfaction of (ordinary) Humean sceptics that we are epistemically within our rights in thinking that those are warrants that indeed we have.
Davies M. (2009). “Two Purposes of Arguing and Two Epistemic Projects.” In Ravenscroft I. (ed.), Mind, Ethics, and Conditionals: Themes from the Philosophy of Frank Jackson, 337–383. New York: Oxford University Press.
I prefer not to use the term employed in Coliva’s book (“hinge propositions”) to avoid suggesting that by accepting a moderate account of the structure of epistemic warrant one necessarily incurs in all the Wittgensteinian commitments that are associated with that term.