The Persistent Interlocutor

Abstract

A Persistent Interlocutor (PI) is someone who, in argumentative contexts, does not cease to question her opponent’s premises. The epistemic relevance of the PI has been debated throughout the history of philosophy. Pyrrhonians famously claim that our inability to dialectically vindicate our claims against a PI implies scepticism. Adam Leite disagrees (2005). Michael Resorla argues that the debate is based on a false premise (2009). In this paper, I argue that these views all fail to accurately account for the epistemic relevance of the PI. I then briefly present an account that aims to do better in this regard, based on the modal notion of safety. On the account proposed, the PI does not violate epistemic or dialectical norms. Rather, her behaviour tends to be epistemically perverse in the sense that it wastes cognitive resources. Perhaps surprisingly, this defect turns out not to be unique to the PI.

1 Introduction

Argumentation is the practice of providing arguments, which are sets of propositions such that some provide evidential support for others. We often engage in the practice of argumentation, and may do so for various reasons. Sometimes we argue with ourselves, sometimes with others.¹ Argumentation is governed by norms: we argue well if we obey these norms, and badly if we ignore them. In this paper, I focus on the relation between the dialectical norms governing argumentation and another set of norms, i.e. the norms governing belief.²

In particular, I will focus on the persistent interlocutor (PI), a (fictional) character that, in argumentative contexts, never ceases to question the premises put forward by proponent (Leite 2005). It is an open question whether such characters violate any dialectical norms. Dialectical egalitarians maintain that they do not.³ Dialectical foundationalists maintain they do.⁴

Furthermore, it unclear whether the PI has epistemic relevance.⁵ The Pyrrhonians thought the PI prevents us from having most of the knowledge we take ourselves to have. While Leite (2005) disagrees with this conclusion, he agrees on the question of relevance: if we cannot answer her successfully, the PI will prevent us from knowing. Michael Rescorla offers a third view, arguing that even if we are unable to answer the PI successfully, this does not entail scepticism (2009).

The main aim of this paper is to show that all these views are lacking. After I make the notion of the PI more precise (Sect. 2), I outline my objections to these views in Sects. 3, 4, and 5. If I am right, we currently lack a good account of the epistemic dimensions of argumentation.⁶ In Sect. 6, I briefly suggest an account that in my view better captures the epistemic relevance of the PI. This view, based on recent work in modal epistemology, locates the problem of the PI not in the structure of her challenges, but in their content. This allows for a nuanced explanation of the epistemic relevance of the PI. In certain well-defined instances, her challenges will prevent us from knowing, in others they will not. Several advantages of this view are outlined. Section 7 concludes.

2 The persistent interlocutor

First, some preliminaries. We identified argumentation as the practice of providing arguments. We do this to rationally resolve our differences of opinion (Eemeren and Grootendorst 2003, p. 53).⁷ In its simplest form, the practice involves two roles: proponent and opponent.⁸ Proponent advances a thesis, opponent questions it.⁹ When opponent questions proponent’s thesis, she may do so in two different ways: she may provide a motivated or an unmotivated challenge.¹⁰ An unmotivated challenge merely asks proponent for reasons, whereas a motivated challenge includes reasons to believe proponent’s claim is false. Consider the following example:

Ada: The earth is round.

Brandon: Why would that be true?

Ada: Because it appears that way from the ISSR.

Brandon: I don’t think so: my astronaut buddy told me yesterday that he remembered the earth looking particularly flat last time he was aboard the ISSR.

Here, Ada is the proponent of the thesis that the earth is round, and Brandon the opponent. Brandon first poses an unmotivated challenge to the thesis, and then poses a motivated challenge to Ada’s second claim.

What are the norms governing argumentation? In the case above, Ada does not argue very well if she fails to respond to Brandon’s challenges, whether motivated or unmotivated, so the following norm seems plausible:

The defence norm: When challenged to defend an asserted proposition, one must either defend it or else retract it. (Rescorla 2008, p. 88)

When proponent responds to opponent’s challenge, this will of necessity be another assertion, and so it may be challenged as well. A Persistent Interlocutor is someone who keeps raising challenges for every proposition asserted by proponent (Leite 2005, p. 397). These challenges may be motivated or unmotivated, but for simplicity I will assume in this paper they are all unmotivated in the sense defined above. Let us further stipulate that a claim is successfully vindicated iff all challenges of opponent have been met. Since the PI will not cease challenging, proponent cannot vindicate her assertion against a PI.

The defence norm is a dialectical norm. But some derive dire epistemological consequences from it, and the persistent interlocutor plays a crucial role in the arguments offered. In the next sections I outline various views on the epistemic relevance of the PI, and argue that they all face serious objections. This motivates the novel view outlined in Sect. 6.

3 Pyrrhonian Scepticism

Why think the PI has epistemic relevance? We take ancient Pyrrhonian scepticism as our point of departure. According to Barnes (1990, Chap. 1), the Pyrrhonians took the widespread disagreement between philosophers of their time to motivate a global epistemological scepticism. Their argument can be summarized as follows:

P1 Knowledge requires epistemic justification.

P2 Epistemic justification requires dialectical justification.

P3 There is no dialectical justification.

C There is no knowledge. (from P1, P2, P3)

P1 is relatively uncontroversial, and will be assumed here. The idea behind P2 is that if you are unable to defend your belief when legitimately challenged, then you lack sufficient reason for your belief and it should consequently not count as justified. Or, conversely, if your belief is justified, then you should be able to defend your belief in response to a legitimate challenge by providing good reasons.

The PI becomes relevant in the defence of P3. For as we saw above, it is impossible to vindicate our beliefs against the perpetual challenges by the PI. Let us here look into this claim in a little more detail. In the face of a challenge against p, proponent has the following options for vindication:¹¹

• a. She can vindicate p by making further assertions q, r, s, etc.

• b. She can vindicate p by asserting p.

• c. She can choose not to vindicate p.

Option b is problematically circular. It is not an adequate response to a challenge to p to simply reassert p.¹² Option a is problematic because the PI will in this case just raise a new challenge to q, r, s, etc. Since the PI by hypothesis continues ad infinitum, and humans are finite beings with limited amounts of time for vindication, proponent will in this scenario never succeed in completely traversing the infinite regress of vindications. Since Pyrrhonians assume that to be able to vindicate p, these further assertions need to be vindicated as well, proponent on this option is never able to vindicate p when confronted with the PI. Since option c also does not lead to the vindication of p, it follows that it is impossible to vindicate p in the face of the PI. Since for the Pyrrhonians, such vindication is required for dialectical justification, and this is in turn required for epistemic justification and knowledge, the Pyrrhonians conclude there is no knowledge.¹³

Besides the sheer implausibility of global scepticism, the main problem with the Pyrrhonian view is that P2 is implausible. In contemporary epistemology, it is relatively uncontroversial that there is a basic distinction between epistemic justification and dialectical justification. I will focus on two arguments in support of this claim.

Our first argument depends on the distinction between internalist and externalist concepts of justification. Let us define an internalist concept of justification as one that maintains that all factors determining whether a belief is epistemically justified are reflectively accessible, and externalist concepts as those for which this does not hold. Externalist accounts of epistemic justification stress that epistemic justification involves factors ‘outside our ken’ such as the truth of our beliefs or the reliability of our methods. On many such views, epistemic justification does not require dialectical justification. According to a standard form of process reliabilism, for example, it does not matter that we are unable to produce reasons in favour of our beliefs (Goldman 1979). As long as they are formed reliably, they will be justified.

A significant portion of contemporary epistemologists will thus reject P2. A second argument, however, shows that even internalists do well to recognize that there is a distinction between the state of epistemic justification and the ability to show that one is in that state. As William Alston argues, “many persons are justified in many beliefs without possessing the intellectual or verbal skills to exhibit what justifies those beliefs. Thus the fact of being justified is not dependent on any particular actual or possible activity of justifying” (1988, p. 273). This holds for many states in which one may be. For example, one can be in the state of being mentally ill without being able to show that one is mentally ill, even if the mental illness supervenes on one’s internal mental states only. Even if justification supervenes only on reflectively accessible states of the subject, it is possible that one is in a state of having a justified belief that p without being able to show that one is in this state. For example, because one lacks the appropriate concept of justification, or even the concept of (good) reasons. Not all people will be able to distinguish good from bad reasons, or distinguish reasons sufficient for justification from ones that fall slightly short. Still, their beliefs may be supported by excellent reflectively accessible states. Internalists need not, and may not in general want to deny justification to these beliefs. So, even on internalist concepts of justification the claim that in general, being justified entails being able to vindicate one’s beliefs in argumentation turns out to be implausible.

If P2 is false, then the sceptical conclusions of the Pyrrhonians do not follow from our inability to vindicate our beliefs in the face of the PI. Even if we cannot vindicate our beliefs in argument, they may still possess the status of being justified, and so the Pyrrhonian view on the epistemological relevance of the PI turns out to be implausible.

4 Leite’s view

Adam Leite grounds his view on epistemic justification in “ordinary conversations in which we offer reasons for our beliefs” (2005, p. 397). In terms of the argument spelled out above, Leite agrees with the Pyrrhonians on P2 (or at least on a version of P2, see below), but evades their conclusion by denying P3. For Leite, there is a class of dialectically foundational propositions that can be asserted to vindicate a further assertion but do not themselves stand in need of vindication. This class of assertions, which Leite also calls ‘terminating claims’, contains all and only those propositions for which proponent “correctly and responsibly takes there to be no reason to doubt” (2005, p. 405). On Leite’s view, it is possible to vindicate one’s claims against a PI by asserting one of these dialectically foundational propositions.¹⁴

At first sight, it seems arbitrary to stop providing justifications at any one point. How does Leite explain that some propositions are dialectically foundational? According to Leite, the Pyrrhonians presuppose, incorrectly, that in order to be justified in one’s belief that p, the belief that p must be already based on reasons. Leite relaxes this requirement, claiming instead that all that is required is to have the ability to provide reasons when called upon (Leite 2005, pp. 401–402). One can have this ability provided there are good reasons available that one can call upon, even if one’s belief is currently not based on those reasons. One can thus vindicate one’s claims to the PI by asserting propositions that one has the ability to provide reasons for, but actually providing the reasons is not required. Of course, the PI keeps asking questions, and so there is a continuous need to further provide good reasons for one’s belief. But crucially, the fact that this game may go on indefinitely does not, on Leite’s view, mean that one’s assertions at any one stage are not vindicated. To the contrary, at any stage, as long as one has the relevant ability to keep going, one’s assertions are fully vindicated. So, dialectical justification is possible, and Leite can maintain that epistemic justification requires dialectical justification without committing to global scepticism.

Because the dialectical warrant of an assertion in argumentative discourse for Leite does not depend on further assertions but rather on the possession of the relevant abilities, no regress is generated, and the justification is able to stay ‘local’ in the sense that “it requires one only to defend a target belief (or a limited set of target beliefs) with good, non-circular reasons drawn from amongst one’s justified background beliefs” (Leite 2005, p. 402).¹⁵

So far, so good. However, as said, Leite assumes P2 of the Pyrrhonian argument, the claim that epistemic justification requires dialectical justification:

I will propose an account of [epistemic] justification which does not generate a structural regress even though it retains the pre[-]theoretical thought that whether one is [epistemically] justified has something to do with one’s ability to offer reasons for one’s belief [dialectical justification]. (Leite 2005, p. 398)

I argued above that this pre-theoretical insight is mistaken, at least interpreted as the claim that epistemic justification implies the ability to show that one is justified. But Leite’s claim in the quoted passage is slightly different. He does not claim that epistemic justification requires that one shows that one is justified. Rather, Leite requires that one is able to offer good reasons for one’s belief, which is weaker in the sense that it does not require a concept of justification or even of good reasons.¹⁶

How plausible is this weaker claim? Leite’s support is based on the (uncontroversial) assumption that being justified requires the ability to properly base one’s belief on one’s reasons. He proposes the following adequacy condition on proper basing:

A minimal adequacy condition for an account of the epistemic basing relation is thus that it allow (1) that the reasons for which a belief is held can be directly determined […], and (2) that one sometimes directly opens oneself to epistemic criticism and incurs further justificatory responsibilities by sincerely declaring that one holds one’s belief for particular reasons (Leite 2004, pp. 227–228).

Causal interpretations of the basing condition fail this requirement, since they cannot be determined directly, and further, one does not ‘directly open oneself to epistemic criticism’ by declaring that one has based one’s belief that p on a particular reason q, since one’s belief may instead be caused, without one’s knowledge, by reason r. One’s declared reasons thus do not provide the proper ground for criticism on the causal account.

Instead, Leite argues that the best way to accommodate this adequacy condition is an account of basing that holds that the relevant agent has the capacity to “directly establish her reasons for holding the belief through her explicit deliberation and reflection about reasons for belief” (Leite 2004, p. 232). After all, if we can establish the basis of our beliefs directly in deliberation, then this basis is always directly determinable. And secondly, by declaring those reasons to be the basis for one’s belief, one plausibly incurs a justificatory burden for those declared reasons.

The core of Leite’s argument is thus that his ability-account of basing best satisfies his dual adequacy condition. I will not take issue with this claim, but rather argue against the adequacy condition itself. Leite supports the condition with explanatory considerations. Our justificatory practices are such, according to Leite, that the adequacy condition would properly explain those practices. When we engage in argumentation, we hold people accountable for the reasons they explicitly endorse, and when we ask people what supports their beliefs, we generally tend to accept the reasons they produce as the basis of their belief.

The first thing to note about this explanatory argument is that it presupposes that our argumentative practices tell us something about the structure of epistemic justification. Since this is precisely the kind of claim Leite aims to support, the argument seems to beg the question.

Even if we do allow an appeal to our argumentative practices at this point, the argument is problematic. For it is equally possible to explain our argumentative practices without assuming Leite’s adequacy condition. A proponent of the causal account of basing, for example, can maintain that sincerely declaring one’s reasons in deliberation is one way out of many to establish the required causal connections between one’s belief and one’s reasons, and that because of this, one incurs certain responsibilities by declaring one’s reasons. On this view, our argumentative practices would be an indirect and fallible route to the reasons on which our beliefs are based. But such fallible routes can still be reliable, and if so, this would explain why we engage in these practices. This causal account fails Leite’s adequacy condition, but it is able to explain our argumentative practices just as well.

Leite might object that such external account allow people to evade responsibility for their beliefs (2004, p. 229). But a proponent of the causal account can plausibly maintain that it is sufficient to explain why we hold people responsible for their declared reasons that their declared reasons are a reliable, albeit fallible, indication of the basis of their beliefs.

If Leite’s adequacy condition is unmotivated, then so is his account of basing. That does not show it is wrong, just that lacks motivation.

However, Leite’s account of epistemic justification faces other problems. First, it is dialectically inconsistent: we can run his argument against causal accounts of basing against his own account. For when Leite discusses some counterintuitive results based on implicit biases, he maintains that declaring one’s reasons establishes basing relations “only if one’s rationality is not impaired” (Leite 2004, p. 237). Since it is often not directly determinable whether one’s rationality is impaired, it now becomes possible on any occasion to evade responsibility for one’s declared reasons. If this is problematic for causal accounts, then so it is for Leite’s own account.

Secondly, Leite is forced to deny justification to adults, children and animals who do not have the required abilities. While Leite is explicit about this restriction of his account of justification (2004, pp. 243–245), it is a heavy price to pay. At the face of it, it seems implausible that the justification of simple perceptual beliefs works differently for animals, small children and reflective adults. We look at a glass, perceive that it contains water, and form the corresponding belief. The justification in these cases seems the same, even if the child lacks abilities that I have.

Leite might maintain that the restriction is supported by the adequacy conditions on the kind of basing required for epistemic justification. But we already saw that these were unmotivated. In contrast, causal accounts allow for uniform perceptual justification in children and adults. These accounts are thus both more simple and more versatile in that they are able to explain epistemic justification both in reflective, deliberative contexts as well as in more unreflective ones through a common account of basing.¹⁷

I submit the above provides sufficient reason to reject Leite’s account of basing, and with it, his account of epistemic justification and its relation to dialectical justification.

5 Rescorla’s view

The final view to be discussed is that of Michael Rescorla (Rescorla 2008, 2009). Rescorla’s view contrasts with that of the Pyrrhonians and that of Leite in that he does reject P2. Given what we have said above, that is an advantage. While Rescorla accepts that dialectical vindication is impossible (P3), he evades global scepticism because P2 is denied. At first sight, this is an attractive view that seems to respect the logical independence between dialectical and epistemic justification.

Ultimately, however, Rescorla’s account fails to adequately capture the epistemic relevance of the PI. Even if she does not violate any dialectical norms, we have the clear intuition that the PI is doing something wrong.¹⁸ But what? Rescorla locates the fault of the PI in the sphere of argumentative goals rather than norms. It is a constitutive goal of argumentation, according to Rescorla, to achieve rapprochement: which is achieved when parties to the argumentation “isolate mutually acceptable premises relevant to the truth of disputed propositions” (2009, p. 57). Since the PI will question every premise put forward by proponent, she fails to achieve such convergence. On Rescorla’s view, the PI engages in the dialectical practice only deviantly; rather like someone who is playing tennis but flouts the constitutive goal that one should try to win.

This explanation is unsatisfactory for several reasons. First, it leaves unexplained the normative connection between argumentation and epistemology:

Reasoned discourse is not just rational cogitation transplanted into the public sphere, and rational cogitation is not just reasoned discourse internalized. They are two distinct modes of rational activity, with two very different architectures. (Rescorla 2009, p. 50)

While this may be true, it leaves unexplained why norms on knowledge may be relevant in argumentation. On Rescorla’s view, argumentation is one thing, and cognition another. Given that they are governed by different rules, there is no reason on Rescorla’s account to expect the one practice to further the other. Of course, such an account may be added, but as it stands, it is missing. Below, I provide a more complete picture. And once that is clear, the assumption that rapprochement is a constitutive goal of argumentation is not needed.

Secondly, at least part of what’s wrong with the PI’s behaviour is specifically epistemic, rather than dialectical in nature. For Rescorla, the PI fails to achieve rapprochement, which is a dialectical, not an epistemic criticism. But what the Pyrrhonians saw right is that we sometimes use argumentation to achieve knowledge, and that the PI frustrates this specifically epistemic goal as well. We need more than just a dialectical critique.

Third and finally, it simply does not seem to be true that rapprochement is a constitutive goal of argumentation. The problem is that that people non-deviantly engaged in argumentation sometimes lack the required goal. If rapprochement is a constitutive goal of argumentation, then to engage non-deviantly in the practice, people must have the goal to achieve rapprochement. But it is possible that people lack this goal and still engage in the practice non-deviantly. I provide two examples.

First, the school teacher who, in order to get her students to recognize the premises on which their arguments are based, decides to keep questioning their assumptions (much as the PI does). Is the school teacher acting deviantly in this case? It seems to me that she is not; that this is a perfectly legitimate way of using argumentation for didactic reasons. Note that the schoolteacher in this case has an explicit intention not to achieve rapprochement. Rescorla is committed to her behaviour being dialectically deviant, but this seems the wrong verdict.

A second, more dramatic, example is provided by Socrates. As Socrates describes his maieutic method in the Theaetetus:

I have, in common with midwives, the following characteristic: I’m unproductive of wisdom, and there’s truth in the criticism which many people have made of me before now, to the effect that I question others, but don’t make any pronouncements about anything myself, because I have no wisdom in me. (Plato, 1973, p. c 105)

It does not seem like a stretch to interpret the claim that Socrates has no wisdom as a claim that he will not commit himself to any proposition in the argumentative exchange with his interlocutor. If this is true, and if it further holds that his only contribution to the exchange is a constant questioning of his conversational partners, then his behaviour fails to achieve rapprochement. Yet, it seems Socrates is arguing not only correctly, but non-deviantly.

These examples support the claim that sometimes we may argue non-deviantly without the goal to achieve rapprochement. If this much is admitted, then Rescorla’s analysis of the PI is unsatisfactory. Rapprochement is not a constitutive goal of argumentation (although I do not dispute that it often is a goal in argumentation). Failing to achieve rapprochement is not what makes the behaviour of the PI deviant. We need to look elsewhere. This is my project in the rest of the paper. As it will turn out, the PI is behaving deviantly in an epistemic rather than a dialectical sense.

6 A better alternative

In this section, I sketch an alternative that aims to evade the objections above. I cannot provide a full defence; my aim is merely to make it seem plausible that the above objections can be met.

The account takes its main idea from recent work in modal epistemology (Pritchard 2005; Sosa 1999; Williamson 2009). The guiding thought underlying this program is that knowledge requires the absence of the kind of luck at issue in Gettier-cases. Veritic luck, as this kind of luck is called, is at issue when you form a true belief in ways that could have easily produced a false belief instead. To evade this kind of luck, knowledge requires safety:

(Safety) S’s belief is safe iff in nearly all (if not all) near-by possible worlds in which S continues to form her belief about the target proposition in the same way as in the actual world the belief continues to be true. (Pritchard 2007, p. 283)

The crucial point to note here is that the safety condition does not require the relevant method to be infallible, but rather only requires that it does not produce a false belief easily, which is modelled as false belief in a nearby world where one uses the same method.¹⁹ One other way to put this is to say that knowledge requires a local kind of infallibility rather than a global infallibility.²⁰ It follows that distant error scenario’s like evil demons and Matrix-like contraptions are irrelevant for the question whether our beliefs are safe. In what follows I shall assume that knowledge requires safe methods rather than globally infallible methods.

Let us now connect this to our present discussion. Note first that while argumentation may fulfil many different functions (and which is why Rescorla is right that there is no essential connection between argumentation and epistemology), nevertheless one central function of argumentation is epistemic (Goldman 2003). A prime example here would be our collective scientific enterprise, where we argue with each other at conferences and in scientific journals in the hope of filtering out those claims that may achieve the status of knowledge. If argumentation is a method that may produce knowledge, safety requires that the argumentative process must not easily produce false belief.

The crux to seeing that argumentation can help us eliminate easily possible error lies in considering the role of reasons in argumentation. Let us assume that a reason q for p is a proposition that logically supports p. Thus, the propositional content expressed by the claim that the earth is a sphere is a reason for believing that we will never fall of the edge because the former provides logical support for the latter.

Logical support comes in degrees. In the above case, the relation is one of entailment (spheres do not have edges to fall off from), but the relation may be logically weaker, such as when I put forward the claim that I have seen a hundred white swans and no black ones in support of the claim that all swans are white.

If reasons speak in favour of the truth of proposition p, then they speak against its falsity, and this is the sense in which all reasons exclude error possibilities. It is good to be clear on the exact nature of this thesis. I claim that to provide a reason for p is equivalent to ruling out possibilities where p is false. Suppose I support my claim that it is raining outside by noting that I see it raining through the window. I claim that my perceptual experience here provides a reason for believing it is not false that it raining. That is, my experience excludes certain error possibilities where it is not raining, in this case perhaps some cases where I am plainly mistaken about the rain and I therefore would lack the same experience if I were to look outside. Of course, my experience does not exclude all error possibilities. It is still compatible with elaborate deception. But that is fine, I am not claiming that all reasons are conclusive reasons. I am merely claiming that all reasons exclude some error possibilities.

In the deductive case, the claim that the earth is round excludes all error possibilities for the claim that we will never fall off from the edge in the sense that its truth is incompatible with the falsity of the conclusion that we will never fall from the edge. In the inductive case, the truth of the claim that I have seen a hundred white swans and no black ones is compatible with the falsity of the claim that all swans are white. But the reason does exclude some error-possibilities, including ones where all swans are black or some other colour. So, no matter whether they provide inductive or deductive support, our reasons always exclude certain scenarios in which our beliefs are false.²¹

It follows that all requests for reasons are requests to exclude some error possibilities. As the perceptual case above makes clear, most reasons will tend to exclude nearby error only.²² It is simply very hard to exclude all possibilities where we might be wrong.

We now have the material to explain in which sense the behaviour of the PI may be epistemically deviant. Remember, the PI continuously raises challenges of the simple form ‘Why p?’ for each proposition p asserted by proponent. Given the above, if the proponent of p proceeds to give reasons for p, they are in effect eliminating possibilities where p is false.²³

Now, either p is be false in some worlds close to the actual one or this is not the case (no world is closer to the actual world than the actual world itself, so false propositions could all easily be false). Trivially, if p is not false in any nearby possible world, then responding to a challenge to p will not exclude any nearby error (about p). In this case, the challenge to p is epistemically irrelevant in the sense that it contributes nothing to our knowledge that p. This is not true in case p is false in some nearby possible worlds. In this case, a challenge to p is epistemically relevant, for we need to exclude these nearby error possibilities in order to know that p.

So, a challenge to p is epistemically relevant only if it helps us eliminate nearby error. Since it does consume cognitive resources to engage with challenges to propositions that could not easily be false, I call such challenges epistemically perverse. They consume valuable cognitive resources and do not contribute to the acquirement of knowledge.

Insofar as the PI raises epistemically perverse challenges, here behaviour is epistemically deviant. Crucially, I do not claim that all requests made by the PI are epistemically perverse, only that when they are, they are epistemically problematic. So, rather than rejecting the PI’s behaviour outright, we arrive at a more nuanced picture where her behaviour is problematic only in certain cases. What is problematic is not the strategy of the PI, but rather the specific challenges that she may raise.

Perhaps surprisingly, epistemic perversity is not a defect unique to the PI. As we have defined her, a PI will not stop challenging one’s premises. On diagnosis proposed here, these challenges are epistemically relevant only insofar they concern propositions that could have easily been false. A merely stubborn interlocutor, on the other hand, does stop the process of challenging, but only after a relatively long time. There is a gradual difference between stubborn and persistent interlocutors, in the limit, the stubborn interlocutor becomes a truly persistent one. While we never encounter truly persistent interlocutors, I take it most of us have some experience with merely stubborn ones. Now, in contexts where we have only a finite amount of time available, a stubborn interlocutor may hinder our epistemic advancement just as a truly persistent one might. The present diagnosis allows us to provide a uniform explanation. As long as stubborn interlocutors raise challenges to propositions that could easily be false, these challenges need to be answered for us to know. But when they challenge propositions that could not easily be false, they commit the same error of epistemic perversity.

Of course, sometimes stubbornness is an epistemic virtue. Our account can explain this as well. A long-winded discussion at a conference may be epistemically relevant as long as the propositions challenged could easily be false. But the discussion degenerates into epistemic perversity as soon as denials of the propositions challenged involve far-off error possibilities.²⁴

Even ‘normal’ interlocutors may raise epistemically perverse challenges. Any request to eliminate far-off error possibilities is epistemically perverse. When we are debating the merits of vaccination policies, for example, challenges are epistemically relevant only insofar as the propositions challenged could easily have been false. Knowledge does not require one to rule out the possibility that vaccines are a tool of ““global elites” [that] torture children to harvest the chemical adrenochrome from their blood, which they then inject in order to stay healthy and young” (Friedberg 2020). Importantly, here, as with the challenges raised by the PI, the problem is not with the strategy of raising challenges but with their content.

Even if on my account, there is nothing special in principle about the behaviour of the PI compared to these other interlocuters, I can explain why people have thought the PI to be specifically problematic. While not a strict necessity, infinite challenging will tend to involve challenges to propositions that cannot easily be false. Argumentation works dialectically because we support more controversial conclusions with less controversial premises. When these premises are challenged, we support them with still more plausible premises. The PI’s challenges will thus tend to involve challenges to propositions of ever increasing plausibility; propositions for which we have ever less reason to believe they could easily be false. In this sense, the PI is more likely to engage in epistemically perverse behaviour then opponents who accept proponent’s reasons at some point.²⁵

Our account evades the objections raised above. First, it allows us to provide a genuine epistemic criticism of the PI without conflating epistemic and dialectical norms. Dialectically speaking, there is nothing wrong with the behaviour of the PI. She is not violating any norms of argumentation. Since we may use argumentation for purposes other than the acquirement of knowledge, her behaviour may be perfectly in order. Examples include cases like that of our philosophy teacher and that of Socrates above. Where Rescorla is committed to saying these characters violate the constitutive aim of argumentation, and so engage in the practice deviantly, the present account evades this conclusion. Rather, the behaviour of the PI is problematic precisely in contexts where knowledge is on the line: her deviancy is epistemic rather than dialectical in nature.²⁶

Second, contra the Pyrrhonians, the account does not presuppose that epistemic justification requires dialectical justification. In fact, I have provided an explanation why knowledge would not require dialectical justification, at least not when confronted with a persistent interlocutor, or anyone else challenging propositions that cannot easily be false. Insofar as such figures raise far-off error possibilities, dialectically justifying your assertions to them contributes nothing to the acquirement of knowledge. One can have knowledge on the basis of argumentation even without a proper answer to the challenges of a PI.

Third, and contrary to Leite’s account, our account is compatible with externalism about epistemic justification, and, relatedly, with epistemic justification of the beliefs of young children and animals more generally. I argued above that argumentation is one way to acquire knowledge, not that it is the only way. The account is compatible with a safety-based view of justification that holds that beliefs are justified if and only if they are produced by safe methods (de Grefte 2018, 2021). Argumentation may be one such method, and simple perception may be another. Our account thus allows for a uniform and elegant epistemology.

7 Conclusion

Let us recap. In this paper, I argued against three main views on the epistemic relevance of the persistent interlocutor. The Pyrrhonians and Adam Leite can be criticized on the grounds that they presuppose a dubious connection between epistemic justification and dialectical justification, albeit in different ways. Michael Rescorla can be criticized for his explanation of the deviancy of the PI in terms of rapprochement.

In response to these problems, I briefly presented a novel account of the epistemic relevance of the persistent interlocutor. Based on recent work in modal epistemology, I argued that the behaviour of the PI tends to be epistemically perverse in the sense of wasting cognitive resources. The account is sketched, rather than presented in detail and fully supported, a task that I have to leave for another occasion. Even if this particular solution is rejected, I hope to have shown that there is an open question about the epistemic relevance of the persistent interlocutor, and with that, an open question about the epistemic dimension of argumentation.