Self-knowledge of one’s current mental states often seems interesting—if not outright puzzling—for at least two reasons. First, such self-knowledge often seems to be privileged, for it seems to be systematically (though not universally) more secure than the knowledge one has of others’ mental states. Second, it often seems to be peculiar, for it seems to be systematically (though, again, not universally) obtained in a way that is only suited for delivering self-knowledge, hence, not by whatever means enable one to acquire knowledge of other minds (Byrne (2018), 4–9). The standard project in contemporary theorizing about self-knowledge is to vindicate these appearances by unearthing the special security and sources of self-knowledge. However, others have argued that we do not actually possess any privileged and peculiar self-knowledge (hereafter “PPSK”), at least when it comes to self-knowledge of propositional attitudes like belief (Gopnik 1993; Carruthers 2011; Cassam 2014). These PPSK-skeptics typically understand self-knowledge and other-knowledge of propositional attitudes as on a par in terms of their security, source, or both.

In reply, some PPSK-realists have offered competing interpretations of the putative evidence against realism about PPSK of propositional attitudes Keeling (2019b), while others have pushed back against the non-privileged and non-peculiar accounts of self-knowledge that are favoured by many skeptics (Coliva 2016; Keeling 2018; Marcus and Schwenkler 2019; Andreotta 2022). The stakes of these debates are hard to grasp if we are unsure “what, if anything, of value we fail to possess if these skeptics are right” (Peterson 2021, 365). For this reason, I will argue that PPSK of one’s propositional attitudes—chiefly, our beliefs—is instrumentally valuable for the efficiency and reliability of a widespread activity in our social-epistemic lives, that of interpersonal reasoning. Some readers may interpret my arguments as providing further support for PPSK-realism if they believe that interpersonal reasoning is in fact a highly efficient and reliable activity in our actual lives. Other readers might reach the more modest conclusion that interpersonal reasoning is a more effective enterprise to the extent that we possess PPSK, whether or not we really possess PPSK, and hence whether or not interpersonal reasoning is a particularly efficient and reliable activity for us to undertake. Either way, the significance of debates between PPSK-skeptics and PPSK-realists can be better appreciated in light of what follows.

Here is the layout for my paper. In section 1 I draw initial inspiration from two earlier accounts of PPSK’s instrumental value. The first, due to Sydney Shoemaker (1988, 1996), concludes that social cooperation in general requires each of us to possess PPSK of many of our propositional attitudes. The second, due to Charles Siewert (2003), concludes that PPSK is indispensable to social cooperation whenever this depends on justifying one’s actions to others. Justifying one’s actions to others can be one way of reasoning with others, that is, reasoning interpersonally. But it is only one way of reasoning interpersonally. I thus consider, in section 2, whether all interpersonal reasoning might benefit from PPSK. My argument is that PPSK does indeed play beneficial roles in all interpersonal reasoning. In section 3 I address objections to my account. In section 4 I consider another recent account of PPSK’s instrumental value, one that emphasizes its role in our capacity for “epistemic control,” and I show how my account complements that account. In section 5 I conclude.

1 Cooperation and Privileged, Peculiar Self-Knowledge

Is PPSK instrumentally valuable? Some philosophers have argued that it is.¹ Indeed, some have argued that it is instrumentally indispensable. Here is Shoemaker, who writes of self-knowledge by “self-acquaintance” instead of privileged and peculiar self-knowledge:²

When one is engaged in a cooperative endeavor with another, it is essential to the efficient pursuit of the shared goal that one be able to communicate to the other information about one’s beliefs, desires and intentions […] When in such circumstances one conveys one’s beliefs to another, this is not merely for the purpose of conveying what one takes to be information about the world, namely the contents of the beliefs; it is also for the purpose of giving him information about oneself which will assist him in predicting one’s behavior and so in coordinating his own behavior with it, and also to enable him to correct those of one’s beliefs he knows to be mistaken […] And here the utility of self-knowledge depends crucially on its being acquired by self-acquaintance; if I had to figure out from my behavior what my beliefs, goals, intentions, etc. are, then in most cases it would be more efficient for others to figure this out for themselves than to wait for me to figure it out and then tell them about it. (1988, 185–186)

Shoemaker argues that PPSK is indispensable for efficiently cooperating with other human beings. For, if others could know one’s mind in the same way and as reliably as one knows one’s own mind, one would be far less efficient at soliciting others’ cooperation. This is because it would just as often be up to others to figure out one’s mind, and to decide on this basis whether cooperation was worthwhile. As a result, one would frequently fail to solicit others’ cooperation of one’s own accord.

Reflecting on Shoemaker’s argument, Siewert wonders whether rational animals “could engage in cooperation and assistance-seeking behaviour, even by generally acting in an attitude-revealing fashion, without representing their own minds to themselves” (2003, 139). In a different idiom: couldn’t there be creatures that are exceptionally adept at expressing—i.e., showing, manifesting, displaying—their attitudes to their fellow creatures without also possessing PPSK of the attitudes expressed, and couldn’t this enable equally efficient cooperation?³ Contra Shoemaker, Siewert supposes that there could be such creatures. Still, he is optimistic about a nearby argument:

For whether or not there can be social animals that act in a usefully self-revealing fashion while oblivious to their own psychologies, they could not engage in the practice of justifying such acts, without being able to represent, in their justifications, relevant facts about their own desires and beliefs […] Now, if the reasons we would offer did not have us acting in ways revealing our actual beliefs and desires to others, we would be much less effective in securing others’ cooperation and assistance in the satisfaction of our desires than we in fact are. (2003, 139)

On this argument, it is not that efficient cooperation always requires PPSK. Rather, such self-knowledge is required for cooperation whenever such cooperation also depends on justifying one’s actions to one’s would-be cooperators. For, lacking PPSK, our actions would often fail to cohere with the attitudes that we self-ascribe. In turn, we would be worse at justifying our actions because we would be worse at appealing to the actual beliefs, desires, and intentions that underwrite them. These inconsistencies might be noticed by others, and this might diminish their trust in us.

More recently, Jon Greco has written that:

Of course thinking about one’s first-order mental states is essential to activities involving coordination and cooperation […] In particular, giving one’s reasons, both epistemic and practical, is essential to various activities in which one must defend one’s beliefs and actions, and having a grasp on such mental states oneself is essential to reporting them to others. (2019, 52)⁴

Like Siewert, Greco claims that self-knowledge is essential for justifying one’s actions (and beliefs) to others. However, Siewert argues that PPSK is indispensable to our widespread success in these matters, whereas Greco claims that “this kind of metacognitive activity can tolerate the same fallibility that we experience in cognition generally” (2019, 53).⁵ He thus denies the importance of privileged self-knowledge (and is silent about peculiar self-knowledge). But he does not consider Siewert’s argument, and so it is hard to know whether his position would change upon further reflection.

This difference between Greco and Siewert set aside, notice that they both focus on a certain kind of interpersonal reasoning. Here, ‘interpersonal reasoning’ denotes exchanges of assertions between interlocutors, or exchanges of questions and assertions, toward a discursive end. For instance, one might reason interpersonally in order to acquire new rational attitudes, or to subject one’s already-held attitudes to the scrutiny of other agents, or to persuade other agents to adopt one’s already-held attitudes. I say that Siewert and Greco are focused on a certain kind of interpersonal reasoning because they only focus on cases in which agents reason interpersonally about one another’s actions or attitudes. In other words, neither philosopher focuses on cases in which agents aim to justify “agent-neutral” propositions to one another, these being propositions whose contents do not refer to any particular agent’s actions or attitudes. One such proposition is:

It is to be contrasted with the sorts of propositions that Siewert and Greco focus on, namely “agent-specific” propositions like:

The latter proposition, but not the former, requires agents to provide self-referential information, this being information that justifies the agent herself to act in such-and-such a way or have such-and-such an attitude. Such information will naturally include “relevant facts about their own desires and beliefs” (Siewert 2003, 139) whereas reasoning about agent-neutral propositions simply requires providing first-order evidence about the agent-independent world, e.g., evidence of rapidly melting arctic ice. But because reasoning about either sort of proposition can be conducted interpersonally, Siewert and Greco will have shown—at most—that self-knowledge matters for interpersonal reasoning about agent-specific matters (whether such self-knowledge is privileged and peculiar as Siewert claims, or not, as Greco claims). I note this here because I will argue in section 2 that PPSK plays a role in both agent-specific and agent-neutral interpersonal reasoning.

Before I get there, I want to make two preliminary points. First, the reader may have wondered whether that Shoemaker’s and Siewert’s arguments establish what they purport to since, on close inspection, they seem to emphasize the importance of privileged access but not, in addition, peculiar access. This is because each argument insists that the special security of agents’ self-knowledge is what facilitates cooperation with other people, and yet this does not obviously entail that agents must exploit a peculiar means of achieving such security. In what follows I will provide arguments for the importance of peculiar self-knowledge as well, thereby going beyond the arguments considered thus far.

Second, it should be noted that some philosophers deny that interpersonal reasoning of any kind (i.e., whether about agent-neutral or agent-specific propositions) requires self-knowledge of any kind (i.e., whether privileged and peculiar, non-privileged and non-peculiar, or any other combination). For example, Robert Brandom writes that there is “nothing incoherent in descriptions of communities of judging and perceiving agents, attributing and undertaking propositionally contentful commitments, giving and asking for reasons, who do not yet have available the expressive resources I provides” (1994, 559). If these communities lack articulate use of the first-person singular, then they cannot self-ascribe and hence self-know their attitudes.⁶ Similarly, Ladislav Koreň claims that we can reason interpersonally by exercising a “practical competence” with linguistic devices like “no,” “but,” and “so,” thus manifesting a “sensitivity” to rational connections between claims without having “metarepresentational” beliefs about the rational connections between one’s own attitudes or one’s interlocutor’s attitudes (2023, 5 (NEW PAGENUMBER)). Finally, Annalisa Coliva offers the following thought experiment:

Take a subject who is able to judge that P, give evidence in favour of it and withdraw from it if required and, therefore, has the first-order belief that P based on judgement. Suppose you ask her “Do you believe that P?” and she is unable to answer. You conclude that she does not have the concept of belief. (2016, 191)

This is a situation in which one interlocutor reasons interpersonally while, ex hypothesi, lacking the conceptual wherewithal to self-ascribe the attitudes that her assertions express. Coliva adds that any such agent will at least possess “the ability to differentiate between, for instance, believing P and P’s being the case, by being sensitive to the fact that her point of view may be challenged […]” (2016, 192, emphasis mine). On my reading, the emphasized terms suggest that such an agent utilizes pre-metarepresentational capacities in the service of interpersonal reasoning; these abilities and sensitivities enable her to reason with others without forming second-order beliefs about her first-order beliefs or her interlocutor’s first-order beliefs.

These philosophers clearly reject Greco’s claim that “thinking about one’s first-order mental states is essential to activities involving coordination and cooperation,” given that interpersonal reasoning is itself a coordinated and cooperative endeavour. But do they extend this rejection as far as to deny that interpersonal reasoning with an aim to justifying one’s own actions and attitudes requires self-knowledge or, at the very least, some form of self-representation like a self-belief? As Steven Levine makes clear in a response to Brandom, it is hard to see how they could cogently deny this. Levine begins by acknowledging the possibility of agents who reason interpersonally insofar as the assertions at issue are first-order assertions of the form “that-P,” these being expressions of agent-neutral propositions in the sense described above. As regards assertions of these propositions, “the performer can justify the statement without explicitly claiming that it is he who is justifying the statement […] because this assertion concerns an objective state of affairs that can be justified by agent-neutral reasons” (2009, 111). However:

[…] is this the case when the assertion that is being challenged concerns an agent’s own action or perception? Here what is being challenged is, for example, one’s entitlement to perform an action or one’s entitlement to claim that one’s perception is veridical. In either case, the justificatory reasons offered cannot be agent-neutral in the way that reasons justifying the assertion “that-P” are. (2009, 111)

So Levine is in league with Siewert and Greco in arguing that, when one’s own actions are challenged by an interlocutor, one cannot merely avail oneself of agent-neutral reasons. Instead, one must avail oneself of agent-specific reasons, which will include facts about one’s own psychology. The only question is whether Levine would side with Siewert in understanding these exchanges as requiring PPSK on the part of whoever seeks to justify her own attitudes, or with Greco in denying any indispensable role for such epistemically high-grade self-knowledge.

As aforementioned, I will soon argue that PPSK plays important roles in interpersonal reasoning about both agent-specific and agent-neutral propositions. But how can I be headed in this direction, having just traced a dialectic that only acknowledges a role for self-knowledge in interpersonally defending agent-specific propositions about one’s own actions or perceptions? In other words, if it is conceded to Brandom and others that agents can reason interpersonally about agent-neutral propositions without so much as a capacity for self-belief, then isn’t it foolish to contend that PPSK—let alone any other sort of self-knowledge—matters for such activity? Fortunately, there is no real problem here. My argument will be that PPSK contributes to interpersonal reasoning for agents who in fact possess the capacity for representing themselves and their beliefs in higher-order thought. This focus allows me to grant Brandom, Koreň, and Coliva their contention that some agents can reason interpersonally despite lacking this metarepresentational capacity.⁷ What I will argue is that agents who do possess this capacity, such as most cognitively developed adult human beings, are systematically vulnerable to certain deficiencies in interpersonal reasoning if they lack PPSK.

2 Interpersonal Reasoning and Privileged, Peculiar Self-Knowledge

Oftentimes, cognitively developed adult human beings have knowledge—or at least beliefs—about their own attitudes, and they often have further beliefs about how their attitudes converge with or diverge from their peers. It is often these higher-order states of mind that motivate agents to reason with one another in the first place. After all, if one agent believes that there is a discrepancy between what she believes and what her interlocutor believes, this can help to explain why she bothers to try and settle the discrepancy through an interpersonal exchange of reasons.

For a hypothetical example, consider two interlocutors: Maya and Roman. Maya might aim to convince Roman that climate change is an existential threat to human civilization (note that this is an agent-neutral proposition: I emphasize the importance of this fact near the end of this section). My claim now is this: Maya would be in a precarious epistemic position, one that might undermine the efficiency of her reasoning with Roman, or one that might even make it better for her to not try to reason with Roman about this issue, if she did not possess PPSK.

Why so? It is easiest to begin by focusing on privilege. Here is the basic idea: if Maya were not in a systematically superior epistemic position regarding her beliefs about her attitudes than Roman was concerning his beliefs about Maya’s attitudes, then Roman could more easily—i.e., with better epistemic grounds—convince Maya that her attitudes already align with his. In convincing Maya of this, Roman would be providing second-order grounds for skepticism about Maya’s belief that she believes climate change to be an existential threat to human civilization. As a result, Maya would not even bother to reason with Roman about the first-order discrepancy, because her self-belief would change in such a way that she no longer took there to be any such discrepancy. Roman might alter Maya’s self-belief in good faith by providing evidence that it is mistaken. But in other cases, Roman might operate in bad-faith by knowingly supplying Maya with misleading grounds for the same conclusion. Indeed, if Roman’s testimony is a source of evidence all on its own then, given Maya’s lack of privileged access to her own belief, her epistemic situation upon receiving Roman’s testimony is immediately altered even if Roman supplies no independent evidence in favour of his testimony. In such cases we could say—perhaps somewhat overdramatically—that Maya has been taken as Roman’s epistemic hostage. As an epistemic hostage, Maya succumbs to Roman’s efforts (good faith or otherwise) to convince her that her self-ascribed attitudes are not really her own. Maya, being falsely convinced of this, is even cut off from opportunities to reason with agents other than Roman about climate change, given that she has been pre-emptively convinced that she does not disagree with those—like Roman—who are climate science deniers.

We might construe these situations as threats to Maya’s epistemic autonomy. I say this because, plausibly, epistemic autonomy is at least partly a matter of being able to navigate various interpersonal reasoning contexts without having one’s self-conception co-opted too easily by others. Indeed, this matters even if we are sometimes duped about the first-order issues by clever interlocutors who supply us with misleading evidence at that level of discourse (e.g., misleading statistics suggesting that climate change—of the anthropogenic variety, at least—is not taking place). An agent who is convinced by a clever interlocutor that the evidence for climate change is bad is still an agent who has assessed those reasons for herself and hence has been mislead on a basis that still deploys her own rational faculties to some degree. And while it is true that Maya might also deploy her own rational faculties in assessing Roman’s claim that her self-belief is wrong, perhaps because Roman supplied her with good reasons (by her lights, at least) to do so, the result is that Maya lacks the self-knowledge that she needs in order to recognize that there is a discrepancy between her belief about climate change and Roman’s belief about it, and this undermines her epistemic autonomy for reasoning with Roman about climate change itself.

Now, as aforementioned, this account of PPSK’s instrumental value is most clearly geared toward privileged self-knowledge, since it is an argument about what happens when the epistemic security of Maya’s self-beliefs is, as a general matter, no better than that of Roman’s perspective unto Maya’s mind. But the account can extend to peculiarity as well, at least if we construe the relationship between privilege and peculiarity in such a way that Maya’s privilege is due to the peculiar way in which she knows her own mind (cf. Peterson 2021). For, if her self-beliefs are not generally acquired by a peculiar means that is generally available to her, then nothing prevents individuals like Roman from seizing upon the very same means to acquire knowledge of Maya’s mind, and this makes it harder to understand why Maya’s self-beliefs are, in general, so epistemically secure that Roman’s contrary claims or beliefs do not give Maya strong reason to change what she believes about herself.

To bring this point into sharper relief, we can consider a putative foil for my account, namely Quassim Cassam’s Inferentialist account of self-knowledge. According to Cassam, both self-knowledge and other-knowledge of agents’ attitudes are acquired through inferences. On his view, there remains an epistemic asymmetry between self-knowledge and other-knowledge, but this asymmetry simply “boils down to a difference in the kinds of evidence that are available in the two cases” (2014, 150). More specifically, the evidence that one has about one’s own attitudes is superior to the evidence that one has about others’ attitudes because it includes sensations, memories, and other non-attitudinal mental goings-on that are not so easily accessed by one’s peers. Applying this view to interpersonal reasoners like Maya, we might say that Maya’s self-knowledge of her attitudes is privileged to some degree even if the same method—inference—is used by both Maya and Roman in coming to form beliefs about Maya’s attitudes. So there is nothing peculiar about Maya’s route to self-knowledge. But now one might insist that Maya cannot be easily taken as an epistemic hostage by Roman, even though she lacks a peculiar way of knowing herself, simply because she has especially good evidence about herself.

However, it could happen that such additional evidence is unavailable to Maya in any number of cases, for what reason can be given for thinking that Maya will always have access to special evidence, given that access to evidence in general is a contingent matter of one’s epistemic position relative to a body of information? Peter Carruthers—another prominent Inferentialist—takes it that we have privileged self-knowledge of non-propositional-attitudinal mental states (2011), and contends that this can be used as a basis for inferring our propositional attitudes. However, privileged access to these other mental states can only provide a basis for inferring our propositional attitudes when we are in such mental states, and yet this itself is a contingent matter. Moreover, even stipulating that Maya has systematically better evidence about herself than Roman has about her, we would also need a general assurance that Maya infers the correct conclusions from this systematically superior evidence. It could happen that Maya has privileged access to the evidence about what she herself believes but cannot reliably utilize this evidence. At the very least, it could happen that she is, in general, no better at utilizing this evidence than Roman is at utilizing his evidence about Maya’s attitudes. Indeed, philosophers like Carruthers seem to embrace this point when they claim that Inferentialist views best explain failures of self-knowledge.

Finally, Inferentialist views are vulnerable to what I call an efficiency concern and a gridlock concern. The efficiency concern is that, absent peculiar access, it could be generally appropriate for Roman to ask Maya to supply the grounds for her self-beliefs, and for Maya to ask Roman to do the same, just to be sure that they were operating in a case where Maya really did have (and had effectively utilized) this superior evidence. Engaging in this second-order interpersonal reasoning would significantly slow down their efforts to get to the first-order issues, thus rendering interpersonal reasoning about first-order issues a less efficient activity. The gridlock concern is that the second-order issue might not get resolved at all whenever both parties fail to reach a verdict about what Maya believes. One might attempt to circumvent these concerns by arguing that Maya’s inferences are subpersonal or non-conscious, such that she cannot be expected to articulate them to Roman. But inferences that are not available for peer-review are also inferences that Maya might be required to lower her trust in, thus calling her self-beliefs into question all over again. To be sure, if some sort of Inferentialism is true, it may follow that agents like Maya often have better evidence and draw better inferences about their own attitudes than their interlocutors can draw about her attitudes, but the points I have been making suggest that such access will be worse for Maya than any form of access that renders the special epistemic security of her self-beliefs a non-contingent matter.

Now, even though I have been critiquing an Inferentialist rejection of peculiar access, I want to reiterate a general lesson for all would-be skeptics about such access. The lesson is that, if the same method—whether inferential or otherwise—is used for acquiring both self-knowledge and other-knowledge, then epistemic privilege will seem to be highly contingent. For, if two agents can come to know one agent’s mind by the same means, then there need be no systematic barrier to their doing with equal epistemic pedigree. In the context of my account, this would mean that there is no strong assurance that agents are systematically warranted in retaining their self-beliefs when challenged by their interlocutors. And this, in turn, would mean that there is no general assurance that interpersonal reasoning about the world, rather than about the interlocutors’ minds, can proceed smoothly. The efficiency and gridlock concerns also generalize: if Maya and Roman share the same method for arriving at a view about Maya’s mind, then Roman might endeavour to interrogate Maya about whether their current context is one in which she has exercised the method more effectively, whether the method is inferential or not. This would slow down and (potentially) gridlock the discourse at the second-order level. Crucially, though, I am not claiming that the systematic protection provided by PPSK against these concerns is universal in scope. For my purposes, PPSK’s instrumental value will have been demonstrated if it is our standard sort of self-knowledge. This would ensure that one is not systematically, generally, or universally vulnerable to innocent-yet-erroneous self-belief change, bad-faith epistemic hostage-takers, or to the efficiency and gridlock concerns, thus improving interpersonal reasoning’s reliability and efficiency as a tool in our social-epistemic toolkit for understanding our shared world.

So goes my account. If correct, it shows that PPSK is instrumentally valuable for interpersonal reasoning, at least among those who are in a position to form beliefs about their own attitudes in the first place (again, a child who has yet to acquire the concept of belief cannot be erroneously convinced that she shares a belief with someone else). Notably, the account applies whether we imagine interpersonal reasoners as aiming to debate an agent-neutral proposition or an agent-specific one. I initially described Maya as aiming to convince Roman that climate change is an existential threat—this being an agent-neutral proposition—whereupon Roman steers the discourse to the second-order level in order to convince Maya that she does not really believe this in the first place. But the content of the proposition was incidental to the example. Had the proposition’s content been agent-specific, e.g., about Maya’s particular climate-focused actions or the belief-desire pairs that rationalize her actions, Roman might have proceeded in the same way. So, my account has a broader scope than Siewert’s: it applies to agent-specific and agent-neutral interpersonal reasoning.

3 Addressing Objections

In this section I reinforce my account by addressing four objections. The first objection is that legitimate challenges to our self-knowledge are in fact quite frequent, and that this provides evidence against the claim that PPSK frequently serves as an epistemic shield against erroneous self-belief change in our actual interpersonal reasoning practices. The second objection is that PPSK does not suffice to ensure that interpersonal reasoning is a reliable route for rational attitude adjustments.⁸ The two final objections are specific defenses of the claim that factors beyond PPSK can protect interpersonal reasoners against erroneous self-belief change in interpersonal reasoning contexts.

The first objection turns on familiar cases of self-deception. Self-deception is ordinarily taken as a failure of self-knowledge in which an agent self-ascribes an attitude that she in fact lacks. Those who take us to have privileged self-knowledge surely ought to say something about this familiar phenomenon. If one does not take privileged access to be universal in scope, then it is at least logically possible to accommodate such cases. Alternatively, one might deny the ordinary view of self-deception by arguing that it does not involve false self-ascriptions (Bilgrami 2006; Coliva 2016). The apparent trouble for my account, however, is that accusations of self-deception are frequent and potentially epistemically legitimate in many cases, and yet these might be precisely the moves that our interlocutors use in order to convince us that our self-beliefs are false. If accusations of self-deception are epistemically legitimate and widespread, and if these accusations can spur agents to adjust their self-beliefs, then what protection does PPSK really provide here?

To begin my response, I want to reiterate a point from my introductory remarks about the dialectical ambitions of this paper, namely that readers need not be convinced that I have unearthed PPSK’s actual functional role for interpersonal reasoners at this world. Secondly, when I say that PPSK provides an epistemic shield against challenges to one’s self-beliefs in interpersonal reasoning, I do not deny that people might sometimes fail to take advantage of this shield—PPSK offers epistemic protection that may not be psychologically appreciated. Beyond these somewhat concessionary responses, the devil must reside in the details, since any further response depends on how we understand the cases at issue. Thus, consider a case in which Maya avows a love of comic books and Roman replies: “you do not love comic books; you’ve just tricked yourself into thinking that loving comic books makes you interestingly different.”⁹ What might bring Maya to accept this accusation? Well, Maya might fixate on the thought that her interlocutor has better evidence about her mind than Roman has about it. If she wondered about her own evidence, and wondered about its inferential role in supporting her self-beliefs, she would be supposing her own self-ascription to be vulnerable to the same epistemic standards that Roman uses to evaluate her self-beliefs. If her self-knowledge is peculiar, however, she will not fixate on this thought, because her self-ascription is not based on the same epistemic standards.

In fact, our actual manner of proceeding tends to bear this out: one’s interlocutor judges one to be self-deceived about one’s love of comic books, and one responds not by attempting (and possibly failing) to offer higher-quality evidence about what one believes, but by offering reasons about why comic books are loveable. Indeed, one possible explanation of privilege and peculiarity is that one’s own take on the reasons for or against adopting some attitude (typically) determine one’s adoption of it. And if one self-ascribes this attitude with full knowledge of the first-order reasons that one takes to support it, one is entitled to make this self-ascription even if other people have evidence contravening one’s self-ascription (Bilgrami 2006; Coliva 2016).¹⁰

Moreover, if we have PPSK, other challenges to our self-beliefs may also be illegitimated, these being challenges where other agents do not accuse us of being self-deceived but, rather, as having made innocent (or “brute”¹¹) errors about ourselves—errors that could only be made on the basis of innocent inferential or observational mistakes.

Here is another, final sense in which the devil is in the details. The objection under consideration is that accusations of self-deception are common, and that these accusations might frequently lead to (reasonable) changes in one’s self-beliefs. However, while such cases may indeed be common, they may only be common in the sense that all of us are occasionally susceptible to them. On this explanation of their commonality, no single agent is liable to be the reasonable target of an overwhelmingly large number of self-deception accusations. There is something suspicious about anyone, even one’s therapist, who would unrelentingly accuse one of self-deception across myriad cases by saying things like “you do not believe that \(-P\), nor hope that \(-Q\), nor desire to \(\phi\), nor love \(S\)”. This suspicion may well reflect a fact about us: that we have enough PPSK to be reasonable in not giving in to too many accusations of self-deception—accusations which, if legitimate, would force us to change our self-beliefs.

The second objection to my account is that PPSK does not improve the reliability of interpersonal reasoning even if it provides us with epistemic warrant to ignore (many) accusations of mistaken self-belief. Cases in favour of this objection are easy enough to set up. For example, maybe Maya claims that climate change is an existential threat to human civilization and Roman gives insufficient epistemic uptake to her assertion because he is prejudiced against women. Indeed, in this case, Maya may be the victim of a “testimonial injustice” (Fricker 2007). But I want to offer two observations. First, although the factors preventing Roman from reasoning with Maya have nothing to do with Maya’s self-knowledge or Roman’s beliefs about Maya’s self-perspective, this does not change the fact that Maya would have an additional problem on her hands if Roman were generally in an epistemic position to make Maya erroneously change her self-beliefs. Second, to the extent that Roman’s prejudiced behaviour does not prevent Maya from knowing herself, she is still in a position to congregate with less prejudiced individuals and to reason with them (or even to reason with Roman indirectly by reasoning with someone that Roman is not prejudiced against, and getting that individual to convey Maya’s reasons to Roman). This point also applies to another concern, namely that Roman might simply say that he agrees with Maya when he does not (this being an inverse version of the epistemic hostage-taking tactic). Maya may not be able to rationally challenge this claim if Roman has PPSK, unless she has reason to deem him insincere, since she will then have strong reason to take Roman at his word. Once again, though, this would not put Maya in the position of being made to form a false belief about what she herself believes about the world, and hence she would not be prevented from discoursing with other agents about the contents of her beliefs about the world.

I now address two objections to the effect that something other than PPSK can explain why we are protected against epistemic hostage-taking. According to the first objection, what really protects Maya against Roman’s nefarious machinations is the same thing as what explains her first-person authority, where what explains this is something other than PPSK. Roughly, “first-person authority” denotes two claims: (1) it is epistemically rational to presume the truth of speakers’ present-tense self-ascriptions of mental states, and (2) it is typically epistemically irrational to interrogate the epistemic grounds of speakers’ present-tense self-ascriptions (hereafter just ‘self-ascriptions).¹² Now consider an “expressivist” explanation of first-person authority which contends that speakers’ self-ascriptions ought to be presumed true and be insulated from requests for epistemic support because they express and thus show the self-ascribed mental states to one’s hearers (Bar-On 2004). This explanation is available even if speakers do not also possess PPSK of the mental states that their self-ascriptions express. The objection, then, is that Maya’s first-person authority gives Roman a strong reason not to challenge most of her self-ascriptions, such that PPSK is explanatorily superfluous in explaining why Roman is not likely to give Maya an erroneous basis for changing her self-beliefs.

Now, for all I have said, Maya’s self-ascriptions may be first-person authoritative in virtue of what they express, whether or not Maya also has PPSK of what they express. Nevertheless, I argue that without also possessing PPSK, Roman could purposefully ignore the first-person authority of Maya’s self-ascriptions in a bid to convince her that her attitudes converge with rather than diverge from Roman’s. He might (rightly) take Maya to have expressed her first-order belief through a self-ascription but still claim that her self-belief is false. Hence, PPSK protects Maya against being manipulated by bad faith interlocutors who ignore her first-person authority, however that is to be explained, because PPSK ensures the general (and systematically superior) reliability of her self-beliefs relative to Roman’s beliefs about her attitudes. PPSK is what gives Maya an epistemic warrant for holding steadfast against his machinations, even if he was already unjustified in challenging her self-ascription challenged her self-ascription in the first place.¹³ Moreover, PPSK protects Maya against erroneous self-belief change even if Roman, innocently, fails to recognize that her self-ascription expresses the very attitude that it is about.

The final objection to my argument is that Maya can get away with merely assuming that she generally has PPSK, such that she is generally entitled to not defer to interlocutors who challenge her self-beliefs (whether in good or bad faith). More substantively, it might be argued that Maya possesses a distinctively strong practical warrant for holding steadfast when faced with accusations of mistaken self-belief, even if she lacks a distinctively strong epistemic warrant for doing so.

The trouble with this objection is that it is hard to see what could ground Maya’s practical warrant for holding steadfast if it is not really, at bottom, the same as (or itself grounded by) epistemic warrant for doing so. This is because a merely practical warrant here would go against her epistemic wellbeing in any number of cases. Specifically, if she did not systematically know herself better than others know her, then she would often—perhaps even typically—have an epistemic reason to discourse with others about whether her self-beliefs are true, and this would be in tension with her practical warrant for avoiding such discourse. In other words, it is only if Maya really has PPSK, thus having epistemic warrant for holding on to her self-beliefs, that holding steadfast against her interlocutors’ countervailing assertions does not inadvertently prevent her from indulging many epistemically legitimate disagreements about what her attitudes are. It is only if she really has PPSK that not entering these disagreements is by and large good for her to do.

It might now be complained, relatedly, that I have merely established the importance of an especially strong epistemic warrant for our self-beliefs, but that this need not amount to PPSK. In other words, Maya might have an especially strong epistemic warrant for her self-beliefs, but these self-beliefs need not be especially reliably true.¹⁴ Indeed, such warrant may also suffice for avoiding the efficiency and gridlock concerns described in section 2. But I think a similar response applies here. For, if Maya has especially strong epistemic warrant for her self-beliefs but this warrant does not amount to self-knowledge in at least most of the cases in which she possesses this warrant, then in any number of cases she will still miss out on an epistemic good—that of true warranted self-belief—whenever she declines to engage with interlocutors who claim that her self-beliefs are false. Moreover, it is hard to understand how she could possess this special epistemic warrant for her self-beliefs if she did not actually possess self-knowledge in most of those same cases. After all, this would be tantamount to having epistemic warrant for self-beliefs that were not correspondingly likely to be true, and this systematic mismatch between truth and warrant would call the warrant itself into question.¹⁵

4 Interpersonal Reasoning and Epistemic Control

In section 2-3 I argued that PPSK provides us with widespread (even if not universal) protection against situations in which others provide epistemic reasons for us to change our self-beliefs, whether our interlocutors are operating innocently or in bad faith, and that this helps to ensure the efficiency and reliability of interpersonal reasoning. In this section I show that my account complements another recent account of PPSK’s instrumental value.

According to Jared Peterson, PPSK is instrumentally valuable because it facilitates “epistemic control,” which is a matter of being able to “keep private or disclose particular facts about one’s mind to others” (2021, 368). Take privacy first. If you have PPSK, then you can reliably conceal your attitudes from others. For example, a teacher might fail to motivate a student’s learning if the student knows that the teacher is pessimistic about the student’s progress. But if the teacher has PPSK of her pessimism, then she has greater epistemic control over whether the student discovers this. Therefore, the teacher has greater control over the student’s motivation to continue studying. For an example about disclosing rather than concealing one’s mental states, Peterson says that “[a]n estranged lover might want a former partner to know in a highly epistemically secure manner that she still loves him” (2021, 369).

He also says that epistemic control:

[…] allows societies to function in a much more productive, organized, and amicable way. When we accomplish group objectives in an efficient and peaceful manner we do so in large part by keeping private that which would be counterproductive to the group’s efforts, and/or revealing our thoughts, beliefs, desires, etc. that are valuable for other members of a group to know. (2021, 371)

Peterson and I are both happy to emphasize the social importance of PPSK. I am also willing to say that PPSK provides a way to reliably disclose one’s attitudes to others. However, Peterson does not also acknowledge the additional possibility of, e.g., expressing one’s love itself as a reliable way of putting one’s former partner in a secure position with respect to one’s mind, where this expressive capacity may or may not depend on an agent’s self-knowledge.

More significantly, though, I submit that PPSK’s role as a shield against erroneous self-belief change is independent of its role in enabling one to disclose or conceal one’s attitudes from others. To be able to better conceal one’s attitudes is to prevent others from discovering what attitudes one has, but this may not matter to interlocutors who do not care (or are simply mistaken) about the facts and, instead, aim to convince you of a certain belief about yourself. Similarly, having an especially epistemic secure way of disclosing your attitudes is something that interlocutors could ignore (as argued in section 3 when discussing first-person authority). Thus, one may be tempted to deny that the instrumental value of PPSK for interpersonal reasoning is a matter of epistemic control.

However, one might be just as easily inclined to regard this as an instance of epistemic control after all, since my account claims that agents with systematically superior knowledge of their self-beliefs thereby exercise greater control over their social-epistemic lives as interpersonal reasoners. Accordingly, the lesson to be drawn may be that we ought to broaden our view of PPSK’s contribution to epistemic control, such that epistemic control encompasses (i) control over which attitudes one discloses to others,¹⁶ (ii) control over which attitudes one conceals from others, and (iii) control over which attitudes one is able to self-attribute, with especially secure epistemic warrant, in the face of disagreement about one’s attitudes, while attempting to reason with others.

5 Conclusion

I have argued that privileged and peculiar self-knowledge contributes to our capacity for interpersonal reasoning about the world around us. To the extent that agents possess PPSK of their attitudes, interpersonal reasoning is a more reliable route to discursively navigating our shared world, and this explains one way in which PPSK is instrumentally valuable.

For the record, I happen to believe that phenomena like epistemic hostage taking are not widespread, and I regard PPSK as at least a partial explanation of this fact. I take myself, therefore, to have contributed to the debate between PPSK-skeptics and PPSK-realists not merely by illuminating the debate’s stakes, but also by taking an anti-skeptical stand within that debate. This being said, I reiterate that others may not be persuaded to go as far as me in this regard, such that the core contribution of this paper is best viewed as an account of how being a PPSK-skeptic or PPSK-realist should affect one’s corresponding conception of our interpersonal reasoning competencies.