1 Introduction

In this essay I provide a solution to Moore’s paradox, which has so far been ignored in virtue epistemological literature. I demonstrate that one of the core ideas of virtue reliabilism entails a straightforward solution to this paradox. Virtue reliabilism entails that in virtue of believing a ‘Moorean’ proposition one will necessarily have an unjustified belief. Moorean beliefs and assertions are absurd in that although they can be true, they cannot be believed with justification. I should note at the outset that this kind of solution is not novel. Many have argued that Moore-paradoxical beliefs are not justified, and that this explains their absurdity. What is novel here is the way in which the solution is derived. The core claims of virtue reliabilism entail this solution. Hence virtue reliabilists do not need to rely on extra-virtue theoretical considerations to solve the paradox.

The structure of this essay is the following. In the next section I present Moore’s paradox. In the third section I lay out some of the core tenets of virtue reliabilism, and offer a view on how the notion of cognitive character should be understood. In the fourth section I demonstrate that the idea of cognitive character can be used to offer a solution to Moore’s paradox.

2 Moore’s paradox

Moore (1942, p. 543) noted that it would be absurd to say that “I went to the pictures last Tuesday, but I don’t believe that I did”. It is absurd to state that something is the case and deny in the same breath that one believes it. Such statements are absurd but they are not contradictory in any obvious way since they can be true. It is entirely possible that Moore went to the pictures last Tuesday although he does not believe it. It is worth emphasizing that the absurdity arises only in a first-personal context. It is absurd to assert a sentence with the propositional content p & ¬B(p) under a first-personal guise, or mode of presentation (p and I do not believe that p). It is not absurd to assert that p and Niko does not believe that p. Moreover, Moorean absurdities are not confined to assertions. Beliefs in propositions such as p & ¬B(p), which are believed under a first-personal guise, are equally absurd.

The paradox that is associated with Moorean absurdities has to do with the question why Moorean beliefs or assertions are absurd, given that they can be true (Green & Williams, 2007, p. 5). Here then is the paradox:

Moore’s paradox:

  1. 1.

    Moorean beliefs and assertions are absurd.

  2. 2.

    Beliefs and assertions which can be true are not absurd.

  3. 3.

    Moorean beliefs and assertions can be true.

Claims 1-3 are intuitively plausible, but jointly inconsistent. Hence one of them must be false. One solution to the paradox is to locate some kind of contradiction in the Moorean belief or assertion and hence reject 3. Moore (1993, p. 210) thought that in asserting “p but I don’t believe that p” I imply that I believe p, which contradicts what I assert, namely that I don’t believe that p.

Another prominent solution to the paradox is to hold that knowledge is the norm of belief and assertion and that Moorean beliefs and assertions violate this norm (DeRose, 2002, pp. 180–181; Huemer, 2007, p. 146; Williamson, 1996, p. 506). After all, knowledge is factive and entails belief. Therefore a belief of the form p & ¬B(p) could amount to knowledge only if one believed that p and one did not believe that p. But that is a contradiction and hence one cannot know p & ¬B(p). Given the knowledge norm, one is not permitted to believe or assert Moorean propositions, and this explains their absurdity. The thought is that Moorean beliefs and assertions are absurd since they can be true yet cannot be permissibly held or asserted.

I find this solution quite elegant and plausible, since I happen to like the knowledge norm of belief, but many have reservations regarding it.Footnote 1 One reason to be skeptical of the knowledge norm-solution is that it rests on a much too strict conception of justified belief. Proponents of the knowledge norm hold that a belief is justified and epistemically permissible just in case it meets the primary norm of belief, which is knowledge (Littlejohn, forthcoming; Sutton, 2005; Williamson, forthcoming). This means that only those beliefs that amount to knowledge are justified, and hence justification is factive: there are no false justified beliefs. Moreover, holding that only those beliefs that amount to knowledge are justified entails that Gettier cases are metaphysically impossible (Kelp, 2016, p. 82). After all, in Gettier cases a subject is supposed to have a justified true belief which doesn’t amount to knowledge.Footnote 2

In what follows I wish to explore whether the theoretical resources that virtue reliabilism provides suffice to explain the absurdity of Moorean beliefs and assertions. If virtue reliabilism all in itself entails a solution to Moore’s paradox, then it is a point in its favour. The virtue reliabilist solution to Moore’s paradox is, roughly, that while Moore paradoxical beliefs can be true, they cannot be justifiably held, and this explains their absurdity. This way of explaining the absurdity of Moorean beliefs and assertions is not unique. Indeed, it is very similar to the knowledge norm-solution and to the solution advanced for instance by De Almeida (2001) and Hintikka (1962). However, the ingredients from which the solution stems are novel. For instance, contra Williamson, we do not need a factive conception of justification in order to advance the solution. Contra De Almeida, the virtue reliabilist solution I offer draws on resources that epistemic externalists can accept, unlike the internalist solution that De Almeida (2001) proposes. I argue that the virtue reliabilist conception of justification, which is both externalistic and non-factive, enables us to offer the kind of solution to Moore’s paradox that many have found attractive. Indeed, this essay can be seen as taking up De Almeida’s challenge to reliabilism. De Almeida writes that “a conceptual framework that is distinctively internalist may have provided us with the solution to [Moore’s paradox…] whereas, to my knowledge, we don’t have the foggiest idea of what a reliabilist approach to [Moore’s paradox] would look like” (2001, p. 56). I suggest that at least virtue reliabilism yields a solution to the paradox.

Finally, a qualification; we will limit ourselves to examine Moore’s paradox as it pertains to beliefs, rather than assertions. I assume that the connections between assertion and belief are strong enough in order for the proposed solution to carry over to Moore paradoxical assertions.Footnote 3 That is, I assume that if it would not be proper to believe a proposition under a guise it would not be proper to assert it. For my argument to work I don’t need to assume the reverse, namely that it if it would be improper to assert a proposition under a guise it is improper to believe it. There might be reasons that prohibit asserting p, say because it would be impolite, hurtful, or disclose secret information to the FSB which don’t prohibit believing p.

3 Virtue reliabilism

The core idea of virtue reliabilism is that knowledge is a cognitive achievement, i.e. a success that is attributable to the subject’s cognitive character. Robust virtue reliabilists see this central thesis as giving both the necessary and sufficient conditions for knowledge, while modest virtue reliabilists think that it only gives a necessary condition for knowledge.Footnote 4 Virtue reliabilists differ in what they take a cognitive success to be. Those who seek to provide a virtue-theoretic analysis of knowledge understand cognitive successes as the acquisition or maintaining of a true belief whereas knowledge first virtue epistemologists understand it as knowledge.Footnote 5

Another issue that divides virtue reliabilists is the question of when a cognitive success is attributable to the subject’s cognitive character. Greco (2010) used to hold that the truth of a subject’s belief that p is attributable to her cognitive abilities if the fact that the subject exercised her cognitive abilities in coming to believe p is part of the most salient explanation why the subject acquired a true, rather than a false belief. Greco (2012, p. 19) has since changed his mind and nowadays holds the truth of a belief is attributable to the subject’s cognitive abilities just in case the way in which the exercise of the cognitive abilities resulted in a true belief would regularly serve the relevant informational needs.Footnote 6 Sosa (2007, 2009, 2010) and Turri (2011), among others, hold that a cognitive success is attributable to a subject’s cognitive character just in case the subject’s success is a manifestation of the subject’s cognitive abilities that make up her cognitive character. Here the ‘attribution-relation’ is understood in terms of a more general metaphysical relation, namely that of a manifestation of a disposition.

As luck would have it, here we need not be concerned with the difficult question of when a cognitive success is attributable to one’s cognitive character. The aspects of virtue reliabilism that I will base my argumentation on have to do with a feature that all virtue reliabilists agree on, which is that knowledge and epistemic justification require the exercise of cognitive abilities or virtues.Footnote 7 What then are cognitive abilities and under what conditions does on possess a cognitive ability?

Virtue reliabilists predominantly hold that abilities are dispositions to succeed while in suitable conditions (Greco, 2010; Littlejohn, 2014; Pritchard, 2012; Sosa, 2010; Turri, 2011). But not all dispositions to form, say true beliefs count. Only those doxastic dispositions that are a proper part of one’s cognitive character can count as abilities of the agent (Greco, 1999, p. 287; Palermos, 2014, p. 1940; Pritchard, 2012). This is the key feature that distinguishes virtue reliabilism from process reliabilism. Several virtue reliabilists have argued that a reliable doxastic disposition is not knowledge-conducive if it is not a proper part of the agent. The case often cited in this connection features a subject, call him Alvin, who has a brain lesion that causes him to believe that he has a brain lesion, though he has no reason to think, apart from the deliverance of the brain lesion, that he has a brain lesion.Footnote 8 Alvin’s belief is the manifestation of an infallible doxastic disposition but yet the belief seems to be neither justified nor to amonut to knowledge. Virtue reliabilists hold that Alvin’s belief lacks a positive epistemic standing because the disposition constituted by the brain lesion is not a part of Alvin’s cognitive character (Breyer & Greco, 2008, p. 174; Greco, 2010, p. 151; Palermos, 2014, p. 1938; Pritchard, 2012, p. 263).

When is a doxastic disposition a part of a subject’s cognitive character? Greco (2010, p. 150) has argued that in order for a reliable disposition to be a part of the subject’s cognitive character it must be.

  1. (i)

    stable,

  2. (ii)

    not strange, and,

  3. (iii)

    integrated into the subject’s cognitive character.

These conditions are designed to secure the idea that in order for a doxastic disposition to be a part of the subject’s cognitive character the disposition has to be the agent’s disposition. Outputs of such dispositions are in sense owned by the subject, in that she is responsible for those outputs, and she can deserve praise or blame for them. Earlier we noted that virtue reliabilists hold that in cases of knowledge the cognitive success that the subject attains is attributable to her cognitive character. Within the literature on responsibility and attributability it is common place to hold that an action is attributable to a subject just in case it reveals or expresses the subject’s character. This is so even outside virtue reliabilism. Shoemaker (2015, p. 59), for instance, writes that “An agent is attributability-responsible for any specific attitude (volitional or non-volitional) just in case it expresses the agent’s deep self, that is, just in case it is causally dependent on, and its content is harmonious with, at least one of the agent’s cares, commitments, or care-commitment clusters.” The notions of character and attributability are hence widely thought to be interconnected. It is via the notion of cognitive character that virtue reliabilists are able to maintain that we are responsible (at least in the sense of attributability) for our beliefs.Footnote 9

The claim that a disposition can qualify as a cognitive ability only if it partially constitutes one’s cognitive character is one of the core ideas of virtue reliabilism. This is because virtue reliabilists seek to understand epistemic normativity in terms of the agent’s properties. For instance, Turri and Greco (2021) write that “[Virtue epistemology] explains a cognitive performance’s normative properties in terms of the cognizer’s properties, such as whether a belief results from hastiness or excellent eyesight, or whether an inquiry manifests carelessness or discrimination. For virtue ethics the relevant properties are moral traits, and for [virtue epistemology] intellectual traits.” In what follows I will focus on spelling out condition (iii) in greater detail since it is the one that my argument hinges on. Furthermore, condition (iii) is most widely shared by virtue reliabilists, and hence focusing on it is justified.Footnote 10 What then does cognitive integration require?

Cognitive integration of one’s doxastic dispositions is a matter of how those dispositions interact with each other. It is a function of how well the different aspects of one’s cognitive system act together to produce successful attempts within the cognitive realm. This kind of idea is widely shared among virtue reliabilists. Greco (2010, p. 152), for instance writes that

cognitive integration is a function of cooperation and interaction, or cooperative interaction, with other aspects of the cognitive system.

Palermos (2014, pp. 1941–1942) holds that

the only necessary and sufficient condition for a process to count as knowledge-conducive is that it cooperatively interacts with the rest of the agent’s cognitive character. [The] process of cognitive integration gives rise to a coherentist effect both on the level of processes (how the beliefs are generated) and on the level of content (how the beliefs themselves combine).Footnote 11

And Sosa (1991, p. 240) writes that

A reason-endowed being automatically monitors his background information and his sensory input for contrary evidence and automatically opts for the most coherent hypothesis even when he responds most directly to sensory stimuli. […] The beliefs of a rational animal hence would seem never to issue from unaided introspection, memory, or perception. For reason is always at least a silent partner on the watch for other relevant data, a silent partner whose very silence is a contributing cause of the belief outcome.

Hence the doxastic dispositions of a virtuous agent are not isolated from each other in that they could operate completely independently from each other without being sensitive to each other’s outputs. The main mechanism via which our doxastic dispositions integrate into our cognitive character is via constantly confirming each other’s outputs. This is something that occurs on a daily basis. Suppose, for instance, that you are sitting eyes closed in your armchair and can feel the heat of the fire as it radiates warmth onto you. You hear the crackling of the wood and smell the bittersweet aroma of the resin as it burns. Taking a sip of barley wine, you open your eyes and see the flames dancing in the fireplace. All of these experiences part-take in confirming your belief that you are sitting near the fire (well maybe the taste of the barley wine does so only very indirectly). The claim that our doxastic dispositions converge in confirming particular propositions is not a radical one. Many of our experiences are multi-modal, in that that multiple sense modalities are responsible for our phenomenological state. And it’s not just the case that our sense modalities confirm each other’s outputs. Rather, in many cases our sense modalities affect the outputs and operations of our other sense modalities.Footnote 12,Footnote 13

Note that cognitive integration, at least in mundane cases, does not require any kind of perspective on the coherence of one’s character. Instead it is solely a matter of how one’s cognitive abilities interact. Given that virtue reliabilists tend to discuss cognitive integration in order to distinguish their view from process reliabilism, it is easy to think that they would hold that cognitive integration would require a perspective on one’s own cognitive abilities. For instance, Pritchard writes that the doxastic disposition that is constituted by Alvin’s brain lesion could be integrated to his cognitive character if he gained good reasons to think that he has such a brain lesion (Pritchard, 2010a, p. 226). But this is an exception, rather than a rule. Reflective endorsement of the truth conduciveness of our doxastic dispositions is almost never required for cognitive integration. In ordinary cases it suffices that our doxastic dispositions constantly confirm each other’s outputs, as when multiple sense modalities part-take in a single experience.

Finally, there are constraints in what ways our cognitive abilities should interact with each other in order to form a whole that can be called the cognitive character. The purpose of our cognitive abilities is to yield knowledge. This is something that even those who hold that cognitive abilities are abilities to form true beliefs accept. After all, it is knowledge which is the achievement, the purpose of cognitive activity, not mere true belief.Footnote 14 If the external circumstances are benign, epistemically virtuous believing results in knowledge. There is something off with one’s doxastic dispositions if they yield true beliefs, but interact in such a way that the beliefs they yield could not amount to knowledge.Footnote 15 If it is in principle impossible that the beliefs that are outputs of one’s doxastic dispositions amount to knowledge, if the beliefs were true, then the doxastic dispositions that produced the beliefs are not properly integrated into one’s cognitive character.

We can then lay out the following condition for cognitive integration:

INTEGRATION: S’s doxastic disposition D, is integrated to her cognitive character only if D, and the other dispositions D* that partially constitute S’s cognitive character, could result in beliefs that amount to knowledge if the beliefs were true, when triggered while in suitable conditions for their exercise.

It is worth to emphasize that INTEGRATION is not as strong a condition as one might initially fear. First, virtue reliabilists hold that cognitive abilities are abilities relative to certain appropriate environments (Beddor & Pavese, 2018; Greco, 2010; Sosa, 2010). The fact that one’s doxastic dispositions would not act in concert if the subject was located in inappropriate conditions for the exercise of some of those dispositions does not mean that the subject would thereby violate INTEGRATION. Counterfactual scenarios in which one’s dispositions act against each other due to the fact that the conditions are not appropriate for the exercise of one’s doxastic disposition(s) are simply irrelevant when assessing whether a doxastic disposition is a proper part of one’s cognitive character.

Second, one might think that there are obvious counterexamples to INTEGRATION. Suppose for instance that you are exposed to the Müller-Lyer illusion. Based on your visual perception alone you would judge that the lines are of different lengths. But if you run your index finger and thumb at the ends of the lines you can perceive that the lines are of the same length via proprioception, and plausibly this over-rides the deliverances of your visual system, and you will not form the belief that the lines are of different lengths. Similarly, if you knew about the Müller-Lyer illusion (which you no doubt did), your theoretical reason would prohibit you from forming a belief on the basis of your visual perception. So in this kind of cases our cognitive abilities are in fact acting in concert, rather than in tension with each other, since our other faculties prevent us from trusting our other faculties in this kind of situations. We do not end up with two contradictory judgments, since reason is always on the watch in the case of a virtuous agent.

Third, INTEGRATION does not rule out the possibility that beliefs that are necessarily false could not be products of cognitive abilities, as for instance when one believes the erroneous, yet plausible looking deliverance of a malfunctioning calculator which seems to be in working order. In such a case the fact that the subject cannot know the proposition in question is not rooted in the way in which her cognitive abilities interact, but in the fact that the proposition is necessarily false. INTEGRATION is meant to capture the idea that there is something off with one’s doxastic dispositions if it is in virtue of the way in which they interact that one cannot know. INTEGRATION is violated only when the subject’s cognitive abilities interact in such a way that they preclude their deliverances from being knowledge.Footnote 16

Virtue reliabilists hold that beliefs that are products of our cognitive abilities are thereby justified (Greco, 2002, p. 311; Sosa, 1980, p. 23; 1991, p. 189). Beliefs that are outputs of doxastic dispositions that do not qualify as cognitive abilities are not justified. So far so good. But could we say something more about the notion of justification at play here? Do virtue reliabilists think that there is a norm on belief, and that one satisfies that norm if, and only if one believes through a cognitive virtue? In that case, justification would ultimately be a matter of norm compliance.Footnote 17 Greco (2010) certainly does not think along these lines since he argues against rule-based accounts of epistemic normativity in a chapter entitled Against deontology. Rather, Greco (2010, p. 7), and other virtue reliabilists such as Sosa (2021 ch. 2), understand epistemic normativity in terms of performance normativity. This is the kind of normativity that attaches to any performance or attempt with an aim, and it concerns the evaluation of a performance as the kind of performance it is (Sosa, 2021, p. 24).

Is this kind of normativity evaluative rather than prescriptive? Prescriptive norms (also known as deontic norms) specify what one ought or ought not to do; what one is permitted or forbidden from doing. Prescriptive norms are thought to be action guiding, and violating them leaves one open to blame. Evaluative norms in contrast have to do merely with what is good, bad, valuable or disvaluable. Such norms are not action guiding but might give rise to prescriptive norms indirectly. (McHugh, 2012, pp. 9–10)

Many think that genuine norms are prescriptive, rather than evaluative, since only prescriptive norms specify what we ought to do (Chrisman, 2020, p. 5137). Evaluative norms, the thought goes, are better seen as standards with respect to which we can assess the goodness or badness of certain things. As such, evaluative norms do not answer the question what you ought to do in a given situation.

Several passages by virtue reliabilists suggest that performance normativity is merely evaluative. For instance, Greco writes that

If knowledge has an evaluative dimension – if epistemology is a normative discipline – then a central task of epistemology is to provide an account of the normativity involved. (Greco, 2010, p. 4 my italics)

Sosa echoes this in writing that

Telic “normativity” has thus its own distinctive character. It involves the assessment of attempts as attempts. Such assessment is relative to the agent and their aims. The apex of such normativity is the fully “creditable” attempt, whose success is fully attributable to the agent. This sort of assessment is sealed off from more substantive axiological or deontic assessment. (Sosa, 2021, p. 24)Footnote 18

The idea that performance normativity, or telic normativity as Sosa prefers to call it, is evaluative is also evidenced by the fact that performance normativity is insulated from moral evaluation. For instance, the moral impermissibility of an assassination does not impact our assessment of how skillful the assassin was in executing the deadly plot when we assess the performance of the assassin as an assassin. But do the evaluative norms that virtue reliabilism gives rise to generate or imply prescriptive epistemic norms?

While it is contentious whether we can derive prescriptive norms from evaluative ones there are prominent theories that make such derivations. Take for instance act consequentialism, according to which an act is morally right if, and only if it maximizes the good. Here the prescriptive norm (one ought to maximize the good) is directly related to the evaluative norms (the standards which determine the good). I am not certain how prescriptive epistemic norms should be derived from evaluative epistemic norms. Perhaps we ought to live good lives (in a prescriptive sense) and in order to do that we must be epistemically virtuous, which in turn requires that we exercise our epistemic virtues. In any case, we should bear in mind that satisfying the evaluative epistemic norms that virtue reliabilism gives rise to, can be necessary for satisfying the prescriptive epistemic norms that bind us (assuming there are such norms).

We are now in a position to present our solution to Moore’s paradox that stems directly from the core ideas of virtue reliabilism. I will demonstrate that subjects who entertain Moorean beliefs thereby believe through doxastic dispositions that are not epistemic virtues. While the proposition p & ¬B(p) might be true while believed under a first-personal guise, the belief itself will necessarily be unjustified.

4 Back for ‘more’

Recall Moore’s paradox, now formulated in terms of beliefs:

  1. 1.

    Moorean beliefs are absurd.

  2. 2.

    Beliefs which can be true are not absurd.

  3. 3.

    Moorean beliefs can be true.

To solve the paradox we must reject either 2 or 3. The virtue reliabilist solution is to reject 2. Beliefs that are true can be absurd in virtue of being such that by believing them one necessarily manifests doxastic dispositions that do not qualify as cognitive abilities or virtues. Hence by having a Moorean belief one has an unjustified belief. Moorean beliefs are then absurd since one cannot be epistemically virtuous while believing such propositions though they can be true. Given what was said above, it is easy to see that virtue reliabilism entails that one cannot be justified in believing p & ¬B(p) under a first personal guise. To prove this assume for the sake of the reductio that J(p & ¬B(p)). Given virtue reliabilism only those beliefs that are products of one’s cognitive abilities qualify as justified. In order for a belief to be the product of one’s cognitive abilities it must be the product of a doxastic dispositions that would manifest as knowledge if triggered while in suitable conditions for their exercise if the beliefs were true. Suppose that p & ¬B(p) is true. Therefore, if a subject is justified in believing p & ¬B(p), it must be possible for her to know p & ¬B(p). Knowledge is factive, entails belief and distributes over conjunction. Therefore K(p & ¬B(p)) entails B(p) & ¬B(p). But that is an outright contradiction, and hence one cannot know that p & ¬B(p). Since the doxastic dispositions that yield as output the belief that p & ¬B(p) are not acting in concert with each other, the dispositions are not a proper part of one’s virtuous epistemic character. Hence the belief is not a product of a cognitive ability and it cannot be justified.

I contend that Moorean beliefs of the form p & ¬B(p) are absurd since the proposition can be true, but yet it cannot be justifiably believed due to its content. This is absurd because it is generally thought to be the case that epistemic justification cannot rule out believing specific propositions due to their content if that content is not contradictory or at least false. But in the case of Moorean propositions the content is such that it cannot be believed via doxastic dispositions that act in concert with each other. Therefore virtue reliabilism entails that there are true propositions that cannot be believed with justification. Moorean beliefs are a case in point. Thus virtue reliabilists would have us reject the second premise of Moore’s paradox.

So far I have argued that subjects who hold Moorean beliefs necessarily have unjustified beliefs. A question that I have only hinted at is whether the virtue reliabilist conception of justification is robust enough to capture the kind of responses that intuitively seem appropriate towards subjects who have Moorean beliefs. Earlier we noted that the kind of normativity that virtue reliabilism gives rise to is perhaps best conceived as being evaluative rather than prescriptive. Beliefs that result from the exercise of our epistemic virtues are good when evaluated as beliefs that aspire to be true or knowledge. Unjustified beliefs in contrast are bad when evaluated as beliefs that aspire to be true or knowledge. So at least we can say that one will have bad or botched belief from an epistemic perspective in virtue of believing Moorean propositions. Such beliefs are therefore epistemically irrational. But if the evaluative epistemic norms that virtue reliabilism gives rise to are directly connected to prescriptive epistemic norms we might be able to say something stronger. Perhaps subjects with Moorean beliefs have the kind of character that diminishes their possibility of living a good epistemic life, and we ought to lead good epistemic lives. Therefore, it wouldn’t just be bad from an epistemic perspective to have Moorean beliefs, one also ought not have such beliefs. I leave it as an open question whether such a connection exists between evaluative and prescriptive epistemic norms (and indeed whether there even are prescriptive epistemic norms).

5 Conclusion

I demonstrated that one of the core ideas of virtue reliabilism entails a straightforward solution to Moore’s paradox. Subjects with Moorean beliefs necessarily have an unjustified belief. The absurdity of Moorean beliefs lies in the fact that while such beliefs can be true, they cannot be believed with justification.