Microsoft Word - de Grefte - 389-415.doc © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. A Modal Criterion for Epistemic Argumentation JOB DE GREFTE Department of Philosophy University of Groningen 9712 CP Groningen Netherlands j.a.m.de.grefte@rug.nl Abstract: In this paper, I adopt a pluralistic framework on argumenta- tion, where the norms governing argumentation depend on the aim with which we engage in the practice. A domain of specifically epistemic argumentation is singled out, and I argue based on recent findings in modal epistemology that this domain is governed by the modal norm of safety; where a belief is safe just in case it is produced by a method that would not easily produce a false belief. While this criterion is well- known and uncontroversial in episte- mology, it has hitherto not been applied to epistemic theories of argumentation. I show the fruitfulness of bringing this modal norm into our theory of argumentation by arguing that this allows for a novel and superior perspective of the relevance of the persistent interlocutor in argumentation theory, and on the relation between dialectical and epistemic norms more generally. Résumé: Dans cet article, j'adopte un cadre pluraliste sur l'argumentation, où les normes qui dirigent la construction et l’évaluation de l’argumentation dépendent du but de notre engagement dans cette pratique. Un domaine d'argumentation spécifiquement épistémique est distingué, et je soutiens, sur la base de découvertes récentes en épistémologie modale, que ce domaine est dirigé par la norme modale de sécurité, selon laquelle une croyance est sûre juste au cas où elle serait produite par une méthode qui ne produirait pas facilement une fausse croyance. Bien que ce critère soit bien connu et non controversé en épisté- mologie, il n'a jusqu'à présent pas été appliqué aux théories épistémiques de l'argumentation. Je montre la fécondité d'introduire cette norme modale dans notre théorie de l'argumentation en soutenant que cela permet une perspec- tive nouvelle et supérieure sur la pertinence de l'interlocuteur persistant dans la théorie de l'argumentation, et plus généralement sur la relation entre les norms dialectiques et épistémiques. Keywords: epistemic argumentation, safety, persistent interlocutor 390 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. 1. Introduction We argue with each other all the time, but what is good argumen- tation? Rhetorical (Perelman and Olbrechts-Tyteca 1969), pragma- dialectical (Eemeren and Grootendorst 2003) and epistemological (Goldman 2003; Lumer 2005) approaches each identify different norms for good argumentation. Usually, these approaches are seen as competitors (e.g. Lumer 2005). I have two aims in this paper. First, to paint a different picture; an inclusive approach to argu- mentation that conceptualizes argumentation as a multi-purpose tool subject to different norms depending on contextually deter- mined aims. On this view, rhetorical, pragma-dialectical and epis- temological approaches peacefully coexist. My second aim in this paper is to apply the inclusive approach by delineating a specific form of argumentation for which the epistemological approach is most suitable, and to contribute to the existing literature on epistemological approaches by incorporating insights from recent (modal) epistemology. Along the way, we’ll see how this approach may provide novel insights on a notorious character in argumentation theory: the persistent interlocutor. The paper is structured as follows. In Section 2, I will argue for my inclusive approach to argumentation theory, and carve out a specifically epistemic form of argumentation. In Section 3, I dis- cuss the safety-based account of knowledge. In Section 4, I apply the safety condition by developing a safety-based condition for argumentation, and show how argumentation may increase the safety of our beliefs. In Section 5, I show how the approach pro- vides novel insight on the persistent interlocutor. In Section 6, I consider objections, and Section 7 concludes. 2. An inclusive approach to argumentation According to rhetorical approaches, argumentation serves to per- suade. According to pragma-dialecticians, the aim of argumenta- tion is to resolve differences of opinion. According to epistemo- logical theories of argumentation, the aim of argumentation is to produce justified belief or knowledge. These approaches are usual- ly seen as competitors in the sense that only one of them can be right (Lumer 2005). A Modal Criterion for Epistemic Argumentation 391 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. Here, I will resist this picture. In particular, I will draw distinc- tions between arguments and argumentation, between descriptive and normative approaches to argumentation and explain my own inclusive approach to argumentation. The first distinction to draw is between arguments and argu- mentation (Dutilh Novaes 2021), or as the distinction is sometimes labelled, arguments as products vs arguments as processes or argument1 vs argument2 (O’Keefe 1992). Here, I will adopt the former terminology. Arguments, as I will conceive them, are sets of statements, such that some of them support others. I am explic- itly not opting for a characterization of arguments in terms of (sets of) propositions, because I think arguments require intentionality; propositions that have never been used to support each other are not yet arguments. Argumentation, then, is the process of producing such argu- ments. In this paper, I am interested primarily in social settings: while it may be possible to “argue with oneself,” I will focus on the norms governing settings where we argue with each other. I remain uncommitted as to the question whether these norms gen- eralize to intrapersonal settings.1 While argumentation can of course occur in much larger groups, the central case in this paper will be a setting with two agents arguing with each other. Let us call these agents agent P and agent O (for proponent and opponent). P and O could take various roles, but in its simplest form, P presents a thesis and produces an argument for it in an effort to convince O. The role of O is to ask critical questions regarding P’s argument—does the conclusion really follow; are the premises plausible?—this kind of argumentation is known as a simple, or non-mixed dispute (van Eemeren and Grootendorst 2016, Sect. 2.3). If O defends theses of her own, she too engages in argumentation and the dispute be- comes mixed. For reasons of simplicity, we will focus on non- mixed disputes in this paper. Argumentation thus, is a practice. This practice is goal-directed: people use argumentation for a reason. To be sure, this reason can 1 There is some reason to suspect they might, however. As Catarina Dutilh- Novaes has argued, argumentative practices may influence our capacity for deductive reasoning (Dultilh Novaes 2013). 392 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. be as simple as to have some fun or spend some time, but argu- mentation takes some effort so people do not engage in it without some expected benefit (although they may be drawn into argumen- tation by others for their benefit). Because argumentation is a goal-directed practice, we may study it descriptively or normative- ly. That is, we may try to describe how people actually argue, or in conjunction, we may try to investigate how people can achieve their argumentative aims effectively—and provide norms for good and bad argumentation. This paper clearly falls on the normative side of this distinction. I will posit some (hopefully uncontroversial) aims with which people engage in argumentation, and investigate what it would take to achieve these aims effectively, thus deriving norms for good argumentation. The pluralism of my account follows from the fact that people may use argumentation for different aims, and that different aims will produce different norms. For example, argumentation may be used by politicians to persuade people that a certain course of action should be taken. In such a setting, effectively achieving this goal may require great rhetorical skill, and even an occasional liberal interpretation of the truth. On the other hand, argumentation in science is often used to discover unknown truths, and with such an aim a strict truth norm may be more effective. What this means is that we must relativize our norms of argumentation to the aims with which we engage in the practice, and that different norms will be associated with different aims. Argumentation is a tool, and whether a specific use of that tool can be criticized depends on what we aim to do with it. In this paper, the aim that I want to focus on is epistemic, namely the acquirement of knowledge. This aim is not the same as the aim of finding truth, since as we will see below, knowledge requires more than just truth. But ambitious as the aim may be, I submit it is one of the more important aims of argumentation. In science, and in daily life as well, we often aim at more than just luckily true belief: we aim at knowledge. And we use argumenta- tion as a tool to acquire it. Let us call argumentation aimed at knowledge epistemic argumentation. In the remainder of this paper, I will investigate one of the main norms governing epistem- A Modal Criterion for Epistemic Argumentation 393 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. ic argumentation. To that end, we will have to look in the next section at the requirements for knowledge. 3. The safety-based account of knowledge What is clear is that at least sometimes, we use argumentation to provide us with knowledge. The primary example here is science, where debates in scientific journals and conferences can be seen as temporally extended processes of argumentation with a clear epistemic aim: scientists debate hypotheses at least partly, we may assume, in order to know whether they are true. I have called this kind of argumentation epistemic argumentation. An account of epistemic argumentation requires an account of what it is such argumentation tries to achieve: knowledge. To be sure, the vast literature on this subject cannot even in main lines be summarized here. I thus opt for a modest strategy: I focus on one recent and popular account to see what would follow regarding the norms on epistemic argumentation. I do not claim this is the only way of conceptualizing knowledge, but the criterion is relatively uncontro- versial, and as we will see, it will allow us to paint a plausible and novel picture of the epistemic relevance of the persistent interlocutor. I focus on what is known as the safety-condition on knowledge. We are all familiar with the notion of safety: we can be (relatively) safe from accidents if we properly protect ourselves, or we may be unsafe from such accidents if our behaviour is reckless and we fail to take the proper precautions. To be safe means not to be at risk from some unwelcome event occurring. In the case of knowledge, the focus is on the risk of forming false beliefs. At its most simple formu- lation, a safety-condition of knowledge says that we know that p is the case (the account is in the first instance an account of proposition- al knowledge) only if we are not at risk of believing falsely. A couple of things need to be said to make this rough account more precise. First, what it takes in general to be safe from some risk occurring. Here, safety theorists like Sosa, Williamson and Pritchard take a modal approach, and I will follow their lead in this paper (Pritchard 2005; Sosa 1999; Williamson 2000). A modal account of safety says that we are safe from some risk occurring if that risk does not only fail to occur in the actual world, but also in 394 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. most nearby possible worlds, where nearby possible worlds are worlds that are similar to the present one. The more similar the worlds where the risk event occurs are to the actual world, the more easily the risk event can occur, and the less safe we are from it occurring. Tricky questions arise when we try to make the notion of world similarity more precise: how do we measure overall world- similarity? Do all respects in which worlds can differ from each other matter equally? Here I do not aim to provide an answer to such questions. Rather I take the notion of similarity as primitive; it seems we are able in practice to make fairly clear judgements of similarity about particular cases, and I take these judgements to inform our judgements of safety. For example, we use airbags in our cars because we think, I presume, that with such devices the range of circumstances in which car accidents fail to cause serious injuries (the risk event in question) substantially widens.2 On the present analysis of safety, that is why we consider cars with air- bags safer than cars without. The safety-based account thus says that one knows just in case there are no nearby (similar) worlds where one believes falsely. This needs to be qualified, however. Take the following case: I believe falsely that water is H3O. Now I happen by accident to overhear a chemistry professor explaining that it is, in fact, H2O. Supposing that not much would need to change to the actual world for me to fail to overhear this professor, there will be similar worlds where I continue to believe falsely that water is H3O. But still, it seems, by modifying my belief on the basis of reliable testimony from an expert I should be able to acquire knowledge about the chemical nature of water (if this would not be so, many of us would lack this knowledge!). What matters for whether we have knowledge thus is not whether we believe falsely at all in nearby worlds, but rather whether we believe falsely on the same basis, using the same method. To illustrate this further: I can come to know about the chemical nature of water on the basis of reliable expert testimony, but not on the basis of reading tea-leaves, even if the tea-leaves happen to predict the nature of water to be H2O, A Modal Criterion for Epistemic Argumentation 395 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. which is true necessarily.3 The reason is that tea-leaf reading would have easily led me to form a different, false belief about the chemical nature of water (presuming, of course, that this method is unreliable and will produce semi-random answers to the questions posed). We thus arrive at the following specification of the safety- based account of knowledge: Ksafe: agent S’s belief that p constitutes knowledge iff the method S used to form her belief that p does not produce a false belief in nearby worlds.4 One specification is in order before we continue to apply this norm in the next section. For it may be thought that Ksafe is a reliability condition on knowledge as it relates to the tendency of our meth- ods to produce true belief rather than false. And while I think in a general sense, standard forms of reliabilism and safety-based approaches belong to the same externalist family in epistemology, I will nevertheless here distinguish the two. In this paper, reliabil- ity concerns the proportion of cases where a method produces truth, whereas a safety concerns the modal closeness of the nearest world where the method produces error. Lottery cases famously bring out the distinction. If I believe on the tremendously long odds involved that my ticket is a loser, it is reliably formed yet not safe, for only a few different numbers would need to come up for my ticket to be a winner, and so the nearest world where I err is 2 This is not the same as saying that safety requires a low probability of the risk event occurring. Take the following example (from Neil Levy): suppose I play Russian roulette with a 100-chamber gun filled with just a single bullet. I am not safe, even if the probability of death is quite small—and we may of course arbitrarily increase the number of chambers. For more on the relation between safety, possible worlds and probability, see de Grefte, J. (2020). 3 Pritchard noted the problems caused by necessarily true beliefs and decided to restrict his account of safety to contingent propositions (Pritchard 2005). Since this move is ad-hoc, a better solution is to allow the formation of false but different beliefs in the modal neighbourhood to undermine the safety of one’s belief in the actual world, as Pritchard acknowledges in his (2012, pp. 256-57). 4 The safety condition is usually understood as a necessary condition on knowledge, rather than as a necessary and sufficient one. For reasons of sim- plicity, I here favour a necessary and sufficient reading. The points of this paper can also be made on the weaker reading, however. 396 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. modally close.5 Since lottery cases are prime examples of igno- rance rather than knowledge, it seems it is safety rather than relia- bility that is required for knowledge. Having explained the safety requirement on knowledge, I de- velop a safety-based approach to epistemic argumentation in the next section. 4. A safety-based account of epistemic argumentation In Section 2, I delineated the notion of epistemic argumentation: argumentative processes with the aim of producing knowledge. In the previous section, I explained the safety condition on knowledge. In this section, I will connect these points and develop a safety-based account of argumentation. I contrast the theory to other epistemic theories of argumentation, and explain how argu- mentation may help us achieve knowledge. The question before us is how we may gain knowledge through argumentation. If knowledge requires safety, then one way in which argumentation may help us achieve knowledge is by in- creasing the safety of our beliefs. In this section, I will explain how argumentation may do this under normal conditions. In the next section we will look at the abnormal case of the persistent interlocutor, the epistemic import of which is illuminated with the help of our safety-based account. Why think argumentation helps increase the safety of our be- liefs? We have seen above that safety requires that one’s belief- forming process does not easily produce false belief, in the sense that it does not lead one to form false beliefs in nearby possible worlds. To show that argumentation can increase the safety of our beliefs, we thus need to explain how argumentation may increase the sphere of possible worlds in which one believes truly. The relevant notion of safety applies to belief-forming methods, so the first thing to establish is that argumentation is a belief- forming process at all. A belief-forming process is a process that produces beliefs. If that process occurs within the skull of a single agent, we speak of individual belief-forming processes. If it cru- cially involves other people, we speak of a social belief-forming 5 See (Williamson 2009) A Modal Criterion for Epistemic Argumentation 397 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. process.6 Since the process of argumentation as we have been characterizing it involves different people, it will be a social be- lief-forming process. Does argumentation indeed produce beliefs? According to an influential account of argumentation, the aim of argumentation is to resolve a difference of opinion (van Eemeren et al. 2004). Since opinions are plausibly regarded as beliefs here, and since to re- solve a difference in belief at least one party to the argument must change her belief, it follows that argumentation is a process that aims to produce new beliefs. If argumentation is a belief-forming process, how may this pro- cess increase the safety of our beliefs? In order to show this, we need to say a bit more on the structure of argumentative exchange. Throughout this paper, I have been drawing on the pragma- dialectical framework of argumentation developed by Frans van Eemeren and Rob Grootendorst (2016, 2004), and I will continue 6 As all processes, belief-forming processes can be described on different levels of generality. We may investigate the process of forming beliefs by eyesight, for example, or that of forming beliefs by eyesight in good lighting conditions while looking at medium sized objects from a small distance, for example. On the one hand, we seem to have a fairly intuitive grasp of the concept: when I look out the window, I form my beliefs by eyesight, when I remember to bring the milk, I form my belief on the basis of memory. But these characterizations are somewhat open, we have a choice as to characterize them more specifically, for example as the method of forming beliefs by eyesight when it is day and lighting conditions are good, or more generally, such as the method of forming my belief on the basis of sensory impressions. As opponents of theories of justification referring to belief-forming methods have long been pointing out, there seems to be no objective basis of making a choice between these more general or specific characterizations Without such a principled distinction, there seems to be no answer to the question what the process is that we use to form our belief. This problem is known as the generality problem (Conee and Feldman 1998; Goldman 1979). Besides the fact that this problem is not specif- ic to safety-based approaches, and in fact plagues all major theories of epistemic justification, I will sidestep this problem by stipulation. For the argumentation theory literature contains many fairly specific descriptions of the various roles and steps in the process of argumentation (e.g. van Eemeren and Grootendorst 2016, Ch. 4), and we may simply select one of those processes and ask whether that process will help us eliminate luck. 398 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. to do so here. This specific approach has the virtues of specificity, recency and popularity, but the features of the approach that I will draw on in what follows, in particular about the roles of opponent and proponent, are generally accepted by other approaches, so our find- ings do not depend on the specifics of the pragma-dialectical theory. As we saw above, in an unmixed dispute the role of proponent is to advance a thesis, and the role of opponent is to cast doubt on this thesis. Opponent may do so in different two ways; she may raise a motivated or an unmotivated challenge. A motivated chal- lenge contains a reason for believing the standpoint is false, whereas an unmotivated challenge merely asks the proponent to provide additional support for the standpoint. Consider the follow- ing example: P: It is raining outside O: Why would that be true? P: Because I see it raining through the window O: But aren’t experiences sometimes unreliable? P: Well, lighting conditions are good and nothing out of the ordinary seems to be taking place… Here, Proponent puts forward the thesis that it is raining outside, and Opponent tries to cast doubt on this thesis. First, by raising an unmotivated challenge, merely requesting reasons for the claim put forward. Second, by raising a motivated challenge that indicates a specific set of error-possibilities. I will now argue that in both cases, proponent’s reasons function to eliminate error possibilities. As I will understand reasons in this paper, they exclude ways in which the propositions for which they are reasons can be false. Thus, the claim that P sees it raining through the window is a reason for her claim that it is raining outside because the proposi- tion that P sees it raining through the window is incompatible with certain scenarios where it is not raining outside—indeed, in this case, most normal scenarios; something strange would have to be going on for P to see it raining outside but it not actually raining. I take this to be a rather minimal requirement on reasons that ex- presses the fact that a reason for believing p is true is always at the same time a reason for believing p is not false. If a reason rules out A Modal Criterion for Epistemic Argumentation 399 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. every scenario in which the target proposition is false, we say it provides deductive support for its conclusion, if it merely excludes some scenarios in which the target proposition can be false, the support provided is inductive.7 On this account, both deductive and inductive reasons exclude error-possibilities.8 What goes for the case of P and O goes for argumentation gen- erally. If epistemic argumentation requires proponents to provide reasons for their claims, and if reasons are propositions whose truth rules out some possible ways in which the supported proposi- tion could be false, then argumentative exchanges like the one above exclude error possibilities. Argumentation is thus a social belief-forming process that can eliminate error-possibilities. Of course, arguments fail to provide support for their conclu- sions if their premises are false, and so the propositions advanced as reasons will only eliminate error-possibilities if they are true.9 My claim is not that all reasons eliminate error-possibilities, but only that they are able to do so. To support this claim, it is suffi- cient to show that true propositions can be reasons and that true propositions are incompatible with certain error possibilities. If this much is admitted, then argumentation may contribute to the achievement of knowledge. As we have seen, in the course of argumentation specific error-possibilities may be eliminated. This contributes to the acquirement of knowledge because knowledge, 7 Note that this conception allows for very weak inductive reasons. Even if there are more p-worlds where q is false than there are p-worlds where q is true, p is still counted as being an inductive reason for q as long as p rules out some worlds in which q is false. However, this is compatible with the natural interpre- tation that in this case, while p provides some reason for believing q is true, p provides more reason for believing q is false. I thank an anonymous reviewer for pressing me on this point. 8 That still leaves the abductive case. This is a complicated case, since the nature and justification of abduction is contested (Douven 2017). While I do believe even abductive arguments exclude error-possibilities (after all, explaining why something is the case seems to involve explaining why it is not not the case) I leave the discussion of abductive arguments for another occasion, and restrict myself to deductive and inductive cases here. 9 This is an externalist conception of argument: an argument provides support only if it actually eliminates certain error-possibilities, not if we merely think it does. It follows that some arguments may be convincing even if they do not provide proper support. 400 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. as we saw above, requires that one’s belief-forming process does not easily produce error. Because the process of argumentation as described in the pragma-dialectical theory leads to the exclusion of specific error-possibilities, proponent’s belief in her claim will generally be safer after having gone through the process of argu- mentation than it was before. Again, the claim is not that all argu- mentation will necessarily have this result. But when argumentation excludes nearby error-possibilities, it contributes to the safety of arguer’s beliefs, and so contributes to the acquirement of knowledge. I want to stress again that my claim is not that argumentation always, or even predominantly leads to knowledge. The epistemic function may not be the primary function of argumentation, and additionally, the increase in safety of our beliefs may still be insuf- ficient to amount to knowledge. But my aim in this section is to argue that argumentation can help us acquire knowledge, and for this an explanation of the way in which argumentation may in- crease the safety of our beliefs is sufficient. Argumentation is a tool that may legitimately be used for many purposes, my claim is merely that the epistemic purpose is among them. It should be noted here as well that the pragma-dialectical framework that I have been using to argue for the claim that argu- mentation can produce knowledge is not itself an epistemic ac- count of argumentation, in the sense that accounts of Goldman (2003) or Lumer (2005, 2020). According to such epistemic ac- counts, the primary function of argumentation is epistemic; usually the achievement of knowledge. As explained in Section 2, I adopt a pluralist account of (the aims of) argumentation. For my purpos- es, it is sufficient that argumentation can be used to achieve knowledge, and that is a claim that is compatible with both prag- ma-dialectical and purely epistemic accounts of argumentation. For even the pragma-dialecticians would allow that sometimes in resolving our differences of opinion, knowledge is generated.10 Seen in this light, the claim that argumentation can help us ex- clude error possibilities may seem fairly obvious. After all, what is argumentation other than a game of giving and asking for reasons, and what are reasons other than considerations that show a stand- 10 I thank an anonymous reviewer for pressing me on this. A Modal Criterion for Epistemic Argumentation 401 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. point is true, i.e., not false? What is the value of specifically intro- ducing the safety condition in this context? In general, I think it allows for a better, more detailed account of the relation between epistemic and dialectic norms. In particular, I will argue in the next section that our epistemic account can solve a well-known problem in argumentation theory. 5. The persistent interlocutor and epistemic perversity In a well-known set of papers, Adam Leite and Michael Rescorla debate the question how to normatively assess the behaviour of a Persistent Interlocutor (PI). According to Leite, a persistent inter- locutor is someone who, in argumentative contexts, does not cease to ask questions (2005, 397). In our terminology, a persistent interlocutor is an opponent who whenever proponent produces a reason in support of her claim, asks for further reasons supporting this claim. While asking for reasons is a legitimate move of oppo- nent according to most theories of argumentation, we have the feeling that someone who keeps asking for reasons is doing some- thing wrong, argumentatively speaking. The question is what this something is. First, let us pin down opponent’s unconditional right to ask for reasons, which Rescorla calls dialectical egalitarianism: Dialectical Egalitarianism: all assertions require defence when faced with brute challenges (Rescorla 2009, p. 146). Here a “brute” challenge is what we have been calling unmotivat- ed challenges. According to Dialectical egalitarianism, proponent always has to respond to such challenges by providing reasons for her view. The debate between Rescorla and Leite is primarily a debate about dialectical egalitarianism. Rescorla argues we should be egalitarians, while Leite argues we should instead by dialectical foundationalists, where dialectical foundationalism is simply the denial of egalitarianism; the claim that there are certain assertions that do not require defence when faced with brute challenges. 402 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. While the debate about the dialectical egalitarianism is a debate about the norms governing argumentation, it has potential episte- mological consequences. One of the main reasons for foundation- alism is that egalitarianism seems to lead to widespread scepti- cism. As the Pyrrhonians already saw, every assertion can be challenged by a persistent enough interlocutor. Since the Pyrrhoni- ans also held that to be epistemically justified in holding a belief one needs to be able to successfully defend this belief in argumen- tative contexts, the impossibility of satisfying a persistent interloc- utor means that the beliefs discussed in such contexts will not be justified. Since the persistent interlocutor’s ability is content- independent, large swaths of our beliefs will turn out not to be justified. On a standard view where knowledge requires justifica- tion, widespread scepticism follows. It is partly because of these dire epistemic consequences that foundationalists like Leite find room for propositions that can be asserted but cannot legitimately be challenged by the persistent interlocutor. Such a set can serve as the basis, or ‘foundation’ of all our other assertions, and we can defend all of our other beliefs successfully by drawing on them. In other work I provide a more detailed analysis of the debate between Leite and Rescorla, and why I think that ultimately, both views are problematic (n.d). Here I am mainly interested in the exposition of my epistemic theory of argumentation, so I shall merely summarize the main points of critique. For Leite, I follow Rescorla in objecting that Leite’s view unduly conflates epistemic and dialectical justification. Just because one is not able to provide an adequate defence of one’s belief (dialectical justification), that does not mean it is not epistemically justified, as the many exter- nalists about epistemic justification would agree. The main prob- lem with Rescorla’s view, however, is that it makes the behaviour of the persistent interlocutor epistemically irrelevant. On his view, the PI’s behaviour does not violate any dialectical norm. She also, however, does not violate any epistemic norm. Rescorla is an epistemological foundationalist, so he thinks that our beliefs are ultimately justified on the basis of a set of foundational beliefs. On his view, the behaviour of the PI is epistemically irrelevant since nothing she says in conversation bears directly on this structure. In A Modal Criterion for Epistemic Argumentation 403 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. terms of our example above: if proponent’s belief that it is raining outside is properly based on her belief that she sees it raining outside, and this latter belief is properly foundational, then her belief is justified, no matter what opponent may say, and no matter whether proponent is then able to successfully defend her belief. As Leite puts the challenge to this view, however, we “want a way to say ‘Shut up!’ which isn’t dogmatic but instead has the right sort of normative and justificatory force” (2005, p. 398). As I understand this comment, Leite means that we want to explain the specifically epistemic defects of the PI. Since for Rescorla the structure of dialectical and epistemic justification are disconnected, it is not immediately apparent how he could provide this explanation. I believe our epistemic theory of argumentation does better in the sense that it evades both these criticisms. First, unlike Leite, the view does not maintain that epistemic justification requires dialectical justification. On the view explained above, in order to provide knowledge, the process of argumentation does not need to exclude all error-possibilities, only those that are metaphysically nearby. This means that the persistent interlocutor’s request for reasons is only epistemically relevant if those reasons exclude nearby error. A request to eliminate far-off error-possibilities like the possibility that we are brains in vats or deceived by evil de- mons may be perfectly legitimate according to the norms of argu- mentation, but since knowledge does not require these error- possibilities to be eliminated, such argumentation will be epistem- ically irrelevant. Insofar as we use argumentation to acquire knowledge, that is, insofar as we are concerned with epistemic argumentation, responding to the persistent interlocutor is not necessary and we may acquire knowledge even without an answer to all brute challenges. So, on our view, successful epistemic justification does not require successful defence against the persis- tent interlocutor. Scepticism is averted even without reliance on dialectically basic propositions. On the other hand, the present view quite naturally explains how we may criticize the persistent interlocutor on an epistemic basis. While the persistent interlocutor does not violate any dialec- tical or epistemic norms, she does frustrate the epistemic goal of argumentation. We sometimes do use argumentation to acquire 404 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. knowledge as I have explained above, and if the persistent inter- locutor in these contexts requests reasons that would eliminate far- off error possibilities, she asks the participants to the debate to spend cognitive effort on an activity that is epistemically irrele- vant. We cannot, but also need not answer such challenges in order to acquire knowledge.11 This criticism of the persistent interlocutor is partly pragmatic in character: I have argued that her behaviour is problematic be- cause it frustrates some of the aims with which we may engage in argumentation. It is however, also partly epistemic because the aim that is frustrated is the aim of acquiring knowledge. I believe this view provides a better explanation of the epistemic dimension of the persistent interlocutor than Rescorla’s view; better because unlike Rescorla, the present view is able to explain how the persis- tent interlocutor may hinders our quest for knowledge even while admitting that dialectical and epistemic justification are distinct. All of this is not to deny that argumentation may serve other purposes than that of producing knowledge. For such other pur- poses, raising far-off error-possibilities in argumentation may be beneficial. But from the epistemic point of view, there is some- thing amiss with discussing such issues; this discussion takes up valuable cognitive resources and contributes nothing to the ac- quirement of knowledge. Accordingly, we shall call argumentation that involves the discussion of far-off error-possibilities epistemi- cally perverse, in contrast to epistemically legitimate argumenta- tion that involves nearby error-possibilities. 11 An anonymous reviewer has objected that my criteria rules out all debates on fundamental issues, since questioning our fundamental premises tends to in- volve far-off error possibilities rather than nearby ones. In response, I would like to emphasize that my criterion is an externalist one. The world is a certain way, and given that way, certain errors are easily possible, and others are not. If our most central beliefs about the world are false, then on my account it is perfectly legitimate to discuss them since they are false in very near worlds. But if they are not, then discussing them is epistemically perverse. So, it would be fine on my account for medieval scientists to question the assumption that the earth was flat, but it is (presumably) epistemically perverse for current flat- earthers to question that it is round. Thanks to the anonymous reviewer for pressing me on this point. A Modal Criterion for Epistemic Argumentation 405 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. 6. Three objections Let us take stock. So far, I have argued for a safety-based account of epistemic argumentation. This account of argumentation ex- plains how argumentation may serve to produce knowledge. In the previous section, I explained how the account can provide a new and better perspective of a notorious character in argumentation theory. In this section I will complete the defence of my account by considering some potential objections. I first discuss two general objections to epistemic theories of argument, drawing on the excellent exposition of Scott Aikin (2008).12 The first objection runs as follows: because epistemological theories are highly controversial within epistemology, it is best to get as much as one can for argumenta- tion theory independently of the controversial theories. Having one's entire case for an epistemic theory of argument hang on a highly controversial premise in, say, meta-epistemology is recipe for a theoretical disaster. (Aikin 2008, p. 140) This objection may be run against our theory: in my defence of a modal account of epistemic argumentation, I have relied on the principle of safety. If this principle is highly contested, my theory will inherit this contestability. Luckily for us, the safety condition is not highly contested. While historically, some alternative modal proposals have been made, such as Robert Nozick’s sensitivity condition (1981), it seems clear by now that the safety principle offers a superior formu- lation of the modal profile of knowledge (Sosa 1999). One of the main reasons for this is that virtually everyone agrees that Gettier cases fail to constitute knowledge due to the presence of certain kinds of knowledge-undermining luck, and that the safety principle 12 Aikin has his own solutions to these problems. They resemble the present approach in that Aikin recognizes a potential pluralism of the norms governing arguments depending on the ends with which the arguments are used. My approach thus falls in the same category of views as that of Aikin. It differs from Aikin’s approach, however, in that I focus on the externalist epistemic norm of safety, while Aikin remains uncommitted as to which epistemic norm applies to argumentation, but seems in general more sympathetic towards internalist norms (see my response to the second objection below). 406 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. is specifically designed to exclude this kind of luck (Pritchard 2005). So, while there are disagreements over the precise formulation of the safety principle, and over the question whether knowledge requires more than just safe belief, the fact that knowledge requires some form of safe belief is relatively uncontroversial. While in this paper I have opted for a specific formulation of the safety condition to be able to show how argumentation may increase the safety of our beliefs, I believe what I have said can easily accommodate slightly different formulations of the safety principle. The main point I have made in this paper is that focusing on the modal requirements for knowledge supports the epistemic theory of argumentation because it explains how argumentation may produce knowledge, and sheds new light on the legitimacy of the persistent interlocutor’s behaviour. Furthermore, it bears emphasizing that my general view of ar- gumentation is pluralistic: we can argue with different aims, and relative to those different aims different norms will apply. In this paper, I talk only of epistemic argumentation: argumentation with the aim of acquiring knowledge. That we use epistemic principles to shed light on this kind of argumentation seems necessary: what will be good norms for acquiring knowledge through argumenta- tion will very much depend on what knowledge is. Of course, the general analysis of knowledge is controversial. But I hope to have shown that we can get quite a lot of theoretical clarification out of focussing on less controversial features of knowledge like the safety condition. That we may not have a complete definition of knowledge means that we will not be able to specify the norms for epistemic argumentation exhaustively.13 But it does not mean that we cannot specify any norm for epistemic argumentation. The second objection that I want to discuss also starts from the observation that my story depends on a particular epistemic norm. Contrary to the foregoing critique, however, it takes issue with the externalist nature of my safety condition specifically. Again, we can refer to Aikin’s specification of the objection: 13 Although see de. Grefte, J. (2021) for an attempt at a definition. A Modal Criterion for Epistemic Argumentation 407 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. The practicability objection runs that epistemic theories are either too thin to answer the needs of offering advice or offer useless ad- vice, because epistemic theories risk ignoring the attitudes of those who need correction. (Aikin 2008, p. 134) The complaint here is that using epistemic theories as criteria for good argument is a bad idea because such criteria will not be able to provide advise that arguers can follow, since the epistemic theories are often thin (so that it is hard to draw practical conse- quences from them) and moreover, in order to have any conse- quences in the first place, these principles need to be recognized by the participants as such, which may not often be the case, given the narrow class of people with epistemic knowledge. My first response to this objection is that safety accounts of knowledge are not too thin to draw practical consequences from. Aikin’s example of a problematically thin notion is that it would only be able to provide advise like “construct arguments that pro- vide good epistemic reasons” (Aikin 2008, p. 132). Without any specification of what a good epistemic reasons are, this advice is fairly useless. However, it is exactly the notion of good epistemic reasons that is given some clear content by a safety-based approach. Good epistemic reasons are reasons that exclude nearby error- possibilities in the sense described above. Since we can determine which reasons exclude which error-possibilities, this is advice that is practically followable; we aim for reasons that exclude as much error as possible, until the error-possibilities excluded are clearly far-off, and therefore irrelevant to our epistemic pursuit. Of course, we may in principle be wrong about what the actual world is like, and we may disagree about which worlds are similar enough to the actual one to be relevant for knowledge. But at least we know what we are looking for. We know at least roughly what it would take to rule out nearby error. That the boundary is vague between nearby and not-nearby error means that it may not always be possible to determine whether a particular challenge raised by PI is problematic. But that does not prevent us from criticizing the PI if she is raising clearly far-off error possibilities. Even if we cannot always be certain that we currently satisfy the criterion, the criterion is not devoid of content and can thus be applied in argu- mentative contexts. 408 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. My second response is twofold. First, even if arguers are not explicitly aware of many epistemic principles, they may still be implicitly aware of them. I do not take the present account to be revisionary, in the sense that I believe that most arguers most of the time already try to exclude easily possible error. As I have argued above, reasons can exclude error-possibilities, whether people providing these reasons are aware of this feature or not. As long as the reasons they provide exclude nearby error, these prac- tices provide knowledge, whether the people involved in argumen- tation are aware of these norms or not. Secondly, I want to stress that I am concerned in this paper with epistemic argumentation, that is argumentation directed at the acquirement of knowledge particularly, not argumentation in general. And follows from the aim of this kind of argumentation that the norms on knowledge apply to this kind of argumentation as well, whether participants are aware of these norms or not. Consider the fact that knowledge requires truth. Participants may not be aware of the fact that the beliefs produced by their argumentation are true or not, but argu- mentation that produces false belief is defective qua epistemic argumentation. Such argumentation does not fulfil its goal, and it can thus be criticized as epistemic argumentation. Similarly, ar- gumentation that does not produce safe belief can be criticized as epistemic argumentation, whether the participants to the discus- sion have any means of determining the safety of their arguments or not.14 The final objection I want to discuss is specific to approaches utilizing safety conditions. Some may object to our condition on 14 Another way to make the same point is in terms of the internalist vs external- ist distinction in epistemology. Especially if we include truth as a criterion on knowledge, it is fairly uncontroversial that knowledge is an externalist notion in the sense that whether we know is something that does not supervene on our mental states or that we can establish by reflection alone. If knowledge is an externalist notion, and epistemic argumentation aims at knowledge, then we will not be able to determine by reflection alone whether our argumentation produc- es knowledge. Still, this is crucial for whether our argumentation meets its aims, and so for whether our argumentation meets the norms of epistemic argumenta- tion. It follows that the norms of epistemic argumentation are at least partly externalist norms. A Modal Criterion for Epistemic Argumentation 409 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. epistemic argumentation that the notion of safety is not precise enough to serve in a fruitful clarification. While we may have an intuitive grasp of notions like world similarity and belief-forming methods, these notions are notoriously hard to make precise. As we have been arguing so far, a belief is safe if and only if it is produced by a belief-forming method that would not easily produce false belief. This means our analysis of (epistemic) safety crucially rests on the notions of belief-forming method on the one hand and the notion of easy possibility (spelled out in terms of worlds similar to the actual world) on the other. Our current objector argues that these notions are too vague to serve their purpose. In response, I will look at both notions in turn and argue that while they are vague, this does not detract from their theoretical usefulness. First, consider the notion of belief-forming method. The objec- tion to using the notion in an epistemic analysis is not necessarily that we cannot define belief-forming methods in a precise way, for we may say that belief-forming methods are processes that take inputs of type A and output beliefs of type B, and fill in A with visual percepts and B with beliefs about the external world, and then the resulting analysis is relatively precise. The charge of vagueness here rather concerns that it is indeterminate what presicification we should choose on any occasion. Consider this example: I look at a tree and form the belief that there is a tree there. We may fill in for A the following: 1) Visual percepts 2) Visual percepts with tree like appearances 3) Visual percepts with tree like appearances in good lighting conditions 4) Visual percepts with tree like appearances in good lighting conditions on April 19, 2021. These are various correct descriptions of the inputs of my belief- forming method. We similarly have choice on how to fill in B in our schema; we may put in ‘beliefs,’ ‘beliefs about the external world,’ or ‘beliefs about trees,’ for example. Crucially, the safety of the method may differ under these various descriptions: it is plausible that forming any beliefs on the basis of any visual per- cepts may more easily go wrong than the formation of tree beliefs 410 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. on the basis of visual percepts with tree-like appearances under good lighting conditions. In our possible worlds framework, this can be explained in the following way. When we check for safety; we have to keep the method fixed and check whether there are nearby worlds where that method is used but it produces a false belief. If we define the method very broadly, then more nearby worlds may be relevant and so the probability for false beliefs in nearby worlds increases if we define our methods broadly. Of course, the objection considered here is a version of the well-known ‘generality problem’ for reliabilism. This makes my defence somewhat easier, because it is now not exclusively on my shoulders to find a solution to this problem. In fact, as several authors have argued recently, the generality problem is a problem for a wide range of distinct epistemologies (e.g. Bishop 2010; Comesaña 2006). While one can take this as so much the worse for those epistemologies, the fact that many theories from different epistemological orientations all run into some form of the problem means that it may be inescapable. I suspect it is, since it seems clear that whether we know depends on the way we have formed our beliefs. The same belief that will never amount to knowledge when we arrive at it through tea-leaf reading may nevertheless constitute knowledge if formed in a different way, say by consult- ing an encyclopaedia. Reflection on cases like this suggests that the difference between knowledge and merely true belief is consti- tuted at least in part by the method of belief-formation. If this is true, then no matter what exact analysis is given, any plausible epistemological theory will need to refer, in some way, to belief- forming methods, processes or ways in which we form our beliefs. The generality problem is an important epistemological problem to be solved, but not one that counts against any particular epistemo- logical theory. While I do not aim to solve the problem here, I think a potential dissolution to the problem may actually be suggested by looking at various non-epistemic forms of safety. First note that for these other kinds of safety, a similar problem arises: crossing the street may be relatively safe in general, but not when we look primarily at our cell-phone, listening to loud music. How we describe the way we do things matters for their degree of safety. While the A Modal Criterion for Epistemic Argumentation 411 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. generality problem is thus a potential problem even outside epis- temology, the fact that we have no principled way of describing our methods of action has not, and should not have prevented us from improving the safety of those methods. Looking left-right- left when crossing the street increases our safety, whether we have a principled way of describing the method or not. The hope im- plicit in a safety-based epistemology is that once it is clear that we aim to increase the safety of our beliefs, it will be able to similarly identify pieces of conduct in epistemology to do so. Like the ex- ample of looking left-right-left, the prescribed behaviour may be fairly particular, and only increase the safety in specific situations (in left-driving countries it will not produce the same degree of safety). But we may also hope to uncover more general methods that help us increase the safety of our beliefs in a wide range of circumstances. As I argued above, argumentation may be just such a method. A second source of vagueness may come from the notion of world similarity. As indicated above, our notion of safety depends on how similar the nearest possible world is where the target event occurs. While the notion of similarity is itself a notion that has resisted philosophical analysis, the notion of complete world similarity may seem even more difficult to make any sort of pre- cise sense of. Yet, the notion has been fruitfully applied in philosophy of lan- guage in the analysis of counterfactuals, among other things (Lew- is 1973; Stalnaker 1968). And also, our psychological capacity for counterfactual thinking, so central to many of our reasoning and planning abilities (Williamson 2005), for example, seems to de- pend on some way of ranking possibilities in terms of how easily they could have occurred.15 We thus seem to have an intuitive, coarse grained method of answering questions like which worlds are more easily possible 15 This is not exactly the same capacity as the capacity for probabilistic reason- ing. Consider playing Russian roulette with one bullet in a thousand chambers: probabilistically, in this situation one does not easily lose. There seems to be a clearly different sense of easy possibility, however, in which in this situation it is easily possible that one shoots oneself. 412 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. than others. The coarse and intuitive nature of this method means that there will be borderline cases, where it is not clear whether a world where our belief is false is similar enough to count as easily possible. But it is an open question whether this is problematic in the context in which the notion of safety is used in epistemology. If it is not clear whether our belief-forming method may have easily produced a false belief, it may be equally unclear whether the resulting belief constitutes knowledge. That there is no exact degree of safety required for knowledge seems a fact to be ex- plained rather than an objection to an epistemic theory. Forming beliefs about medium-sized objects in utter darkness is too unsafe a method to preclude knowledge—that much seems clear—but looking at the same objects in broad daylight seems safe enough. That there are going to be cases in between where it is just not clear whether we acquire knowledge about these objects just by looking seems fairly obvious. One upshot of the present paper is that the same goes for the method of argumentation: sometimes the support provided will clearly be strong enough for producing knowledge, sometimes it will clearly not be strong enough, and sometimes, it will be unclear whether an argument is strong enough to provide knowledge. The fact that it is sometimes vague whether arguments provide knowledge does not detract from the usefulness of the epistemic theory of argument. Insofar as epistemic argumentation is con- cerned, arguments are legitimate only if they are safe enough to produce knowledge. If it is unclear whether this condition is satis- fied, it will be sufficient to provide further support until it is clear that the resulting beliefs, if true, would constitute knowledge. Of course, we cannot exclude all error-possibilities, but the whole point of a safety-based analysis of knowledge is that this is not required: we need to exclude only those error-possibilities that are easily possible, and this is usually within our limits. 7. Conclusions In this paper, I first outlined an inclusive theory of argumentation. I then focussed on epistemic argumentation specifically, and pro- posed a safety-based norm for this kind of argumentation. The A Modal Criterion for Epistemic Argumentation 413 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. approach is more modest than extant epistemic theories of argu- mentation by restricting the normative force of the epistemic norms on argumentation to epistemic argumentation specifically: if we do not aim to argue for knowledge, then we are not bound by the epistemic norms, but once we try to achieve knowledge through argument, the norms take effect. I have spelled out in some detail what I take to be the most important epistemic norm on argumentation, the safety condition. Since current epistemic approaches to argumentation are usually spelled out in reliabilist or evidentialist terms, a safety-based approach is novel. I have explained the best way to cash out the notion of safety, and shown how the notion may be fruitfully applied to shed light on the epis- temic and argument-theoretic validity of the behaviour of the notorious character of the persistent interlocutor. Finally, I have responded to what I take to be the most obvious objections to the present theory. While a full defence of the safety-based approach goes well beyond this paper, I hope to have shown that it is a serious contender in the ring of epistemic theories of argument, and provides a fruitful addition to argumentation theory generally. References Aikin, Scott F. 2008. Three objections to the epistemic theory of argu- ment rebutted. Argumentation and Advocacy. 44(Winter): 130-42. Bishop, Michael A. 2010. Why the generality problem is everybody’s problem. Philosophical Studies: An International Journal for Philos- ophy in the Analytic Tradition 151(2): 28-98. Comesaña, Juan. 2006. A well-founded solution to the generality prob- lem. Philosophical Studies: An International Journal for Philosophy in the Analytic Tradition 129(1): 27-47. Conee, Earl and Richard Feldman. 1998. The generality problem for reliabilism. Philosophical Studies: An International Journal for Phi- losophy in the Analytic Tradition 89(1): 1-29. Douven, Igor. 2017. Abduction. Stanford Encyclopedia of Philosophy, 1- 22. Metaphysics Research Lab, Stanford University. Dultilh Novaes, Catarina. 2013. A dialogical account of deductive reasoning as a case study for how culture shapes cognition. Journal of Cognition and Culture. 13(5): 459-82. 414 de Grefte © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. Dutilh Novaes, Catarina. 2021. Argument and argumentation. In Stan- ford Encyclopedia of Philosophy, ed. Edward N. Zalta. Metaphysics Research Lab, Stanford University. Fall 2021 ed. Eemeren, Frans H. van and Rob Grootendorst. 2003. A systematic theory of argumentation: The pragma-dialectical approach. Cambridge: Cambridge University Press. Eemeren, Frans H. van, and Rob Grootendorst. 2016. Argumentation, communication, and fallacies: A pragma-dialectical perspective. Routledge. Eemeren, Frans H. van and Rob Grootendorst. 2004. A systematic theory of argumentation: The pragma-dialectical approach. Cambridge University Press. Goldman, Alvin I. 1979. What is justified belief? In Justification and knowledge, ed. G. S. Pappas. Dordrecht: D. Reidel Publishing Com- pany. Goldman, Alvin I. 2003. An epistemological approach to argumentation. Informal Logic. 23(1): 51-63. de Grefte, J. 2020. Towards a hybrid account of luck. Pacific Philosoph- ical Quarterly 101(2), 240-255. de Grefte, J. n.d. The persistent interlocutor. de Grefte, J. 2021. Knowledge as justified true belief. Erkenntnis, 1–19. Leite, Adam. 2005. A localist solution to the regress of epistemic justifi- cation. Australasian Journal of Philosophy. 83(3): 395-421. Lewis, David K. 1973. Counterfactuals. Cambridge, MA: Harvard University Press. Lumer, Christoph. 2005. The epistemological theory of argument--how and why? Informal Logic. 25(3): 213-243. Lumer, Christoph. 2005. Introduction: The epistemological approach to argumentation--A map. Informal Logic. 25(3):189-212. Lumer, Christoph. 2020. Knowledge and the epistemic function of argumentation. In Reason to Dissent. Proceedings of the 3rd Europe- an Conference on Argumentation, eds. Catarina Dutilh Novaes, Hen- rike Jansen, Jan Albert van Laar and Bart Verheij, 219-224. College Publications. Nozick, R. 1981. Philosophical explanations. Cambridge, Massachu- setts: Harvard University Press. O’Keefe, Daniel J. 1992. Two concepts of argument. In Readings in argumentation, eds. William L Benoit, Dale Hample and Pamela J Benoit. Foris Publications. Perelman, Chaïm and Lucie Olbrechts-Tyteca. 1969. The new rhetoric: A treatise on argumentation. Notre Dame, University of Notre Dame Press. A Modal Criterion for Epistemic Argumentation 415 © Job de Grefte. Informal Logic, Vol. 42, No. 1 (2022), pp. 389–415. Pritchard, Duncan. 2005. Epistemic luck. New York, NY: Oxford Uni- versity Press. Pritchard, Duncan. 2012. Anti-luck virtue epistemology. The Journal of Philosophy. 109(3): 247-79. Rescorla, Michael. 2009. Epistemic and dialectical regress. Australasian Journal of Philosophy 87(1): 43-60. Sosa, Ernest. 1999. How to defeat opposition to Moore. Nous. 33(s13):141-53. Stalnaker, Robert C. 1968. A theory of conditionals. In Studies in logical theory, vol. 2, ed. Nicholas Rescher. Oxford: Blackwell. Williamson, Timothy. 2000. Knowledge and its limits. Oxford, NY: Oxford University Press. Williamson, Timothy. 2005. Armchair philosophy, metaphysical modali- ty and counterfactual thinking. Proceedings from the Aristotelian So- ciety 105(1): 1-23. Williamson, Timothy. 2009. Probability and danger. The Amherst Lec- ture in Philosophy 4: 1-35.