1 Introduction

Broadly speaking, disagreements are worth calling “deep” to the extent they are not just deadlocked, but they’ve reached a point—a depth—where the subject of the disagreement is fundamental. A disagreement is “epistemic” when it is about something overtly epistemic in character. For example, when it is about whether a proposition is justified or an epistemic principle is true. Deep epistemic disagreements are disagreements with both features.Footnote 1

Within the existing social epistemic and argumentation theoretic literature a variety of different kinds of disagreements either have been—or could reasonably be—characterized as deep. That raises the question of whether, and to what extent, these different kinds of disagreements reflect disagreement about the nature of deep disagreement or are instead highlighting importantly different phenomena. In this paper, we discuss three different kinds of disagreement characterized by Goldman (2010), Lynch (2010, 2016) and Fogelin (2005) respectively. We make the case that, within the context of each of these kinds of disagreement, there is a breakdown in the prospects of rational persuasion through the give and take of epistemic reasons. In particular, we claim that, in each of these forms of disagreement, the prospect of one party offering an argument in support of their point of view that is epistemically cogent for the other party is substantially constrained. So, while the various kinds of disagreements that we consider here are distinct, they are variations on a common theme. Each kind of disagreement undermines the prospects of a resolution to the disagreement through an exchange of purely epistemic reasons—even though the reasons why each disagreement undermines the prospects of such a resolution varies for each case. It is this common theme that makes these kinds of disagreements worth calling deep, highlights their importance and encourages us to find new strategies for navigating them.

2 Rational Persuasion

The notion of rational persuasion is vexed, in part because the notions of rationality and persuasion are vexed. So, rather than attempting a full account of rational persuasion here, we pull out certain features of the concept that are salient for any discussion of deep epistemic disagreement as we understand it.

In A Theory of Argument Vorobej (2006) develops an account of rational persuasion that draws on Richard Foley’s discussion of epistemic rationality (Foley 1987). Foley contrasts the notion of a rationally persuasive argument with that of a sound argument,

Arguments that are sound—arguments that have true premises and a truth preserving form—might be of only little help, since perhaps not every person in the group would recognize them to be sound. The same can be said of arguments that are ‘inductively sound’—arguments that are likely to be truth preserving and that in addition have true premises. But arguments that merely are believed by all to be sound (either inductively or deductively) might not be of much help either, since they might not generate a stable consensus; it might be that if the members of the group were to reflect for a moment on these arguments, they would come to believe that they are not sound arguments after all. What is needed is not so much arguments that are sound or arguments that are merely believed by the people involved to be sound. . . . The idea, more exactly, would be to look for arguments such that were the people involved to be reflective, they would think that these arguments are likely to be truth preserving and in addition they would think that that there is no good reason to be suspicious of their premises. (Foley 1987, p. 4)

In other words, sound arguments are not the gold standard for rationally persuasive argument. Nor is it enough to merely believe the argument is sound.

Broadly stated, rational persuasion, rather than requiring sound argument, involves an attempt to persuade someone to do or believe something by offering the addressee evidence that they should recognize as providing a sufficient basis for belief or action. More exactly, it provides a person a sufficient basis for belief or action if they are being cautious, reflective and reasoning consistently with their own principles.

According to Vorobej, some person P should be rationally persuaded to do or believe something just in case the argument offered in support of the belief or action is cogent for P. Vorobej contends that “an argument A is cogent for some person P, within context C, just in case it is rational for P, within C, to be persuaded of the conclusion of A, on the basis of the evidence cited within A’s premises” (Vorobej 2006, p. 47).Footnote 2 At the core of Vorobej’s account of rational persuasion is, thus, the view that an agent recognizes an argument’s premises as providing adequate grounds to believe the argument’s conclusion and that there is good reason to think the premises are true.Footnote 3 Lynch’s account of rational persuasion overlaps with Vorobej’s in certain respects. In describing a conception of rational persuasion Lynch claims,

You rationally persuade someone on the basis of a reason that would make sense internal to her perspective. That is, to rationally persuade you of P, it is necessary that I persuade you on the basis of an irenic reason. A gives an irenic reason R to B for some P, only if were B aware of her principles, and reasoned consistently with them, B would recognize that R is a reason for P. That is, in order for A to give an irenic reason to B, B should be able to recognize—even if she in fact does not—that it is a reason from her standpoint. (Lynch 2016, p. 252)

As mentioned in section I, a variety of different sorts of disagreements have been characterized as deep within the social epistemic and argumentation theoretic literatures. What makes these different disagreements deep? We propose that these different sorts of disagreements are deep because, within the context of such disagreements, there is no possibility of rational persuasion—in the sense just outlined—that draws on epistemic reasons alone.Footnote 4 In other words, the participants to the deep disagreement will not be able to make arguments for their standpoints that are cogent for the other participants that only draw on epistemic reasons. So, participants to a deep disagreement will not recognize the reasons that the other participants to the disagreement offer in support of their conclusions as providing sufficient grounds to believe those conclusions.Footnote 5 In the following section, we illustrate how three different sorts of disagreement considered in the argumentation theoretic and social epistemic literature share this characteristic.

3 Some Varieties of Deep Epistemic Disagreement

3.1 Goldman-Style Deep Disagreement

In a 2010 paper titled Epistemic Relativism and the Epistemology of Disagreement Alvin Goldman characterizes the first kind of disagreement that we discuss here. Goldman’s discussion of disagreement emerges within the context of the debate between steadfast and conciliatory views in the epistemology of disagreement. The primary debate in the contemporary literature on “peer disagreement” pivots around the question of whether disagreement with an epistemic peer is epistemically significant. In other words, does disagreeing with an epistemic peer suggest you should adjust your belief on the matter under dispute?

Two key opposing views on the epistemic significance of peer disagreement are the conciliationist and the steadfast views. Conciliationists hold that many beliefs that people hold on a range of disputed issues need to be substantially revised (Christensen and Lackey 2013, p. 1). For instance, on some variants of this view, if an agent learns that someone they believe to be an epistemic peer disagrees with them, then they are rationally obliged to shift their doxastic attitude in the direction of their peer’s. Conciliationists regard peer disagreement to be epistemically significant. Disagreement with an epistemic peer, on this view, is evidence that one should adjust their doxastic attitude towards their peer’s on the matter under dispute. Proponents of the steadfast view hold that there is not always an obligation to change one’s doxastic attitude in virtue of learning that a peer disagrees with you. According to the steadfast view, if one agent learns that someone they regard to be an epistemic peer disagrees with them, it does not automatically provide evidence, in and of itself, that there is a good reason to revise the relevant doxastic attitude. Perhaps one should reflect on one’s evidence carefully and it turns out that this is one of the rare times that one’s peer, or oneself, have made a mistake. Between these two opposing points of view on the epistemology of disagreement there are a variety of middling positions about the epistemic significance of peer disagreement (Christensen and Lackey 2013, p. 1).

In Goldman’s discussion of peer disagreement he articulates one way to flesh out a steadfast answer to the problem of the epistemic significance of peer disagreement. Goldman purports to explain a variant of the steadfast approach he calls objectivity based relativism (OBR). In what sense is OBR a version of the steadfast approach? On Goldman’s view the following scenario is possible: two or more agents hold incompatible doxastic attitudes towards the same proposition and each of the agents is justified in believing that their respective doxastic attitudes are justified. Since both agents are justified in believing their doxastic attitudes are justified, and these doxastic attitudes are incompatible with one another, it is reasonable for each of these agents to remain steadfast in the face of the disagreement. This is the case even if the person with whom they are disagreeing is an epistemic peer.

How are such disagreements possible? To answer this question some background is required. First, consider Goldman’s notion of objective justification.

(Objective justification) Doxastic attitude D towards proposition p is objectively justified just in case the objectively right epistemic system (E-system) prescribes that agent S should adopt A towards p in light of S’s total evidence vis-à-vis p (modified from Goldman 2010, p. 192).

This definition involves the notion of an epistemic system. Goldman characterizes an epistemic system (or E-system) as follows,

(Epistemic system) An epistemic system (or E-system) is “a set of norms, standards, or principles for forming beliefs and other doxastic states” (Goldman 2010, p. 187).

Note that the definition of objective justification employs the notion of an objectively right E-system. Whether or not such a system of epistemic norms exists is a contentious issue. However, Goldman’s primary point does not require the existence of an objectively right E-system. Moreover, Goldman doesn’t argue in support of one particular E-system. Rather, Goldman’s purpose is to show that a reasonable disagreement might exist at the higher-order level even if it doesn’t at the lower-order level. That is, even if one assumes there is just one objectively right E-system, it is possible for the following situation to arise: one agent is objectively justified in believing that they are objectively justified in believing p and the other agent is objectively justified in believing that they are objectively justified in disbelieving (or in withholding belief from) p. And that, importantly, is consistent with it being the case that only one of the agents is, in fact, objectively justified in believing (or disbelieving) that p. That is, both agents might be objectively justified (by the correct E-system as it were) in believing their lower-level attitudes are justified (by that system) without it being the case that both of those attitudes are justified. Supposing that higher-order justification can contribute to the reasonableness of belief, then it would be reasonable for one agent to believe p and for the other to believe not-p (or to disbelieve p) or withhold judgment from p.

How might such a disagreement emerge in practical terms? According to Goldman, this situation is possible when agents possess distinct epistemic positions. Epistemic positions are distinct instantiations of correct schematic epistemic norms. An instantiation of an epistemic norm can be modeled by offering different specifications of values for variables used in the statement of the norm.

To further explain the notion of epistemic position, consider the example of two distinct educational contexts discussed in Goldman (2010). One educational context is within a religious community that is skeptical of evolutionary theory. The other context is within a scientific community in which evolutionary theory is regarded as highly reliable. The thought is that students occupying these distinct instructional contexts occupy different epistemic positions relative to certain testimonial norms of the objectively correct E-system.

As Goldman points out, there is wide agreement among epistemologists that a correct E-system includes some kind of testimonial norm in which it is reasonable to believe the testimony of authorities. The exact statement of a correct testimonial evidence norm is controversial, but there is wide agreement on a basic schema that involves the principle that a random speaker, or a recognized authority, should be believed in the absence of defeating evidence (Goldman 2010, p. 197). Thus, schematically a testimonial norm may look something like the following,

(T-norm) For agent A, and recognized authority R, if R says that p, then A should ceteris paribus believe p.Footnote 6

Doubtless this statement of the testimonial norm is open to criticism and requires qualifications, but (T-norm) serves to illustrate the point that distinct epistemic positions can generate distinct instantiations of some schematic epistemic norms. Plausibly, teachers who are acknowledged by a community as having the authority to teach are recognized authorities. It is reasonable for students to believe what their teachers tell them. Thus, the student in the religious instructional context occupies a different epistemic position than the student in the scientific instructional context. The recognized authorities, relative to the distinct epistemic positions, make conflicting claims on, for instance, the reliability of evolutionary theory, or on the origin of species. Thus, distinct values assigned to the A and R variables in (T-norm) generate distinct instantiations of (T-norm).

Given this example, a natural question is whether agents who occupy these distinct epistemic positions are genuine epistemic peers? We think it is not implausible that students operating within these distinct epistemic positions could possess the sort of evidential and inferential symmetry required for epistemic peerhood. However, honing in on a different example than the one Goldman considers may serve to more clearly illustrate how this sort of disagreement could emerge among epistemic peers.

Consider the norm governing beliefs grounded in deductive reasoning. Again, the exact formulation of this norm is controversial. However, it is highly plausible that some such norm would be part of the correct E-system. One formulation of such a norm would be as follows,

(D-norm) If \({\mathcal{S}}\) holds a reasonable belief that p, and S competently deduces, in accordance with an admissible logic L, from p that q, while retaining her knowledge that p, then (epistemically) S should believe that q.Footnote 7

Imagine one exceptional graduate student learns logic from Brouwer, an ardent advocate of the view that intuitionistic logic is the one true logic, and another exceptional graduate student learns logic from Timothy Williamson, an ardent advocate of the view that classical logic is the one true logic. Both graduate students engage in extensive discussion and deliberation with their supervisors. Their understanding of several problems in philosophy and mathematics are shaped by their experiences learning from these master logicians. Over time, they each become highly convinced of their supervisors’ respective views and see things largely in line with their respective supervisors. They both understand the arguments for and against the correctness of classical and intuitionistic logic. However, Brouwer’s student thinks that, for instance, excluded middle and double negation-elimination are not generally valid and Williamson’s student thinks that these classical principles are valid. Both graduate students believe one another to have reasonable responses to all the other’s arguments and can draw on a range of literature and respected authorities that agree with them. Plausibly Brouwer’s student is in a different epistemic position with respect to D-norm than Williamson’s student. In the former case, competent deductions do not generally involve inferring in accordance with DNE, for instance, and in the latter case they do. So, different instantiations of D-norm are generated by these logicians’ distinct epistemic positions. Each of these distinct positions offers a different value for the L variable in D-norm. In the case of Brouwer’s student, the value of the L variable is intuitionistic logic. In the case of Williamson’s student, it is classical logic. For Brower’s student, logic alone does not provide a basis for rationally believing claims supported by some classically valid inference principles. This is the case even if the premises of such inferences are known to be true. However, for Williamson’s student any classically valid inference would provide rational grounds for believing the conclusion, presuming the premises are known to be true. In fact, for Williamson’s student, it would be unreasonable not to believe the conclusion given that the antecedent conditions on D-norm were satisfied.

So, presuming that norms similar to (T-norm) and (D-norm) are part of the objectively right E-system, the different epistemic positions occupied by different agents renders it prima facie reasonable for one agent to believe a proposition that it is reasonable for the other agent to withhold belief from (presuming there is no other evidence with respect to the proposition in question). Even though we are working with one and only one correct system of E-norms, one person is justified in believing that it is justified to believe a proposition that the other person is justified in believing that it is justified to disbelieve (or withhold belief from).

There are, of course, a variety of issues raised by Goldman’s account. For our purposes, however, two are salient. The first is that while Goldman motivates his account by way of the problem of peer disagreement, it can also be understood as involving an explicitly epistemic disagreement as defined above. The peer disagreement concerns the question of whether the epistemic peers involved can or cannot continue to reasonably disagree over the belief that p given the total relevant evidence. But note that, e.g. Goldman’s account suggests that the agents involved in an OBR scenario disagree not just about whether p, but also about whether they are both justified in believing that p. This disagreement is properly epistemic in our sense, since it concerns whether a belief has a given epistemic property, that of having objective justification. Secondly, it is important to highlight that it is unclear how OBR type argumentative discussions could result in one agent rationally persuading the other of their point of view by drawing on epistemic reasons alone. For their ability to make epistemically cogent arguments to the other about the subject matter in question will be hampered by the different instantiations of the relevant epistemic principles. The reason their ability to make epistemically cogent arguments is hampered is because the different parties to the disagreement interpret the relevant epistemic principles in different ways. When one party draws on the relevant epistemic principles it will not rationally persuade the other of the same conclusion. For example, consider two agents who occupy slightly different perceptual positions (plausibly an aspect of their epistemic position) such that one perceives a particular patch of a tomato as red and another perceives the same patch of the tomato as orange. Further suppose that both employ the epistemic principle, “presuming normal conditions, one should believe that something is the color it appears to be.” Any attempt for one of the participants to persuade the other about the patch’s color by drawing on this principle would not produce an argument that is epistemically cogent for the other.

Note that, given all we have said here, it is possible, in some circumstances, for one or other of the participants in an argumentation to change their mind, perhaps due to other sorts of reasons or factors. Other factors could, for example, result in one party adjusting their epistemic position with respect to (T-norm). Such an adjustment could shift who is regarded as an appropriate authority on the issue under consideration. However, holding the epistemic positions of the parties to the disagreement fixed, rational persuasion that draws on epistemic reasons alone seems unlikely. This is the case because, in considering the (T-norm) example at least, a key component of each epistemic position involves who is appropriately recognized as an authority on the origin of species. Making changes to these features of the party’s epistemic position changes key features of Goldman’s example. Thus, so long as these features are held fixed—that is, so long as their epistemic position is held fixed—it is not clear what reasons could be leveled that would result in rational persuasion using epistemic reasons alone.

For these reasons we characterize Goldman-style disagreement as a variant of deep epistemic disagreement. Goldman-style disagreement may be regarded as a level-1 deep disagreement. We call it level-1 deep disagreement because it is a sub-norm disagreement. The disagreement is “sub-norm” in the sense that it does not require any disagreement over E-norms. Rather, it can emerge from a scenario in which the parties to the disagreement are in complete agreement over the correct E-system, but disagree because of their different epistemic positions.

One line of objection to understanding Goldman-style disagreements as deep disagreements is that it, some may claim, results in an over inclusive account of deep disagreements. Why? The thought underlying this objection is that it may be possible to have a Goldman-style deep disagreement that is relatively confined, and does not involve the parties disagreeing over a large class of domains. Consider again the dispute about the color of the patch on the tomato. If the agents with varying perceptual perspectives with respect to the tomato’s color came to engage in a discussion about its color, they would disagree about the claim that the tomato is red. One of the agents might argue as follows, “Look! The tomato appears red.” The other agent might then reply by saying, “it certainly does not, it is clearly orange.” However, as explained earlier, the agents, in this circumstance, would be hampered in their capacity to provide epistemically cogent arguments that would persuade one another that the tomato is red or orange respectively. Similarly, some constructive mathematicians may disagree with classical mathematicians as to the set of valid inferences. This means that some arguments that the classical mathematician regards as proofs, the constructive mathematician may not (or, at least—depending on the flavor of the constructivist—may regard the classical proof as less informative). In spite of this, both mathematicians may also be anarcho-syndicalists, Libertarians about free will, panpsychists and agree on many other subjects. So, deep disagreements need not extend to a wide variety of domains. Their core feature is an inability to persuade through epistemically cogent arguments. Therefore, in our view, including these kinds of disagreements in the class of deep disagreements is not over-inclusive, but the proper result.Footnote 8

3.2 Lynch-Style Deep Disagreement

How do we determine what reasons justify an epistemic principle, like (T-norm) or (D-norm), as being a member of the objectively right E-system? As noted, Goldman does not problematize this issue. Instead, Goldman presumes one correct system of E-norms. From there he proceeds to explain how, even in the presence of an objectively right E-system, reasonable disagreement among epistemic peers could occur. However, presumably an important outstanding issue is what sorts of considerations would justify regarding some E-norm as a member of the correct system of E-norms.

While Goldman doesn’t address this problem, Lynch (2010, 2012, 2016) does. In particular, Lynch raises the issue of how an epistemic principle can be justified for someone that is skeptical of the correctness of that principle. So, while one form disagreement emerges because distinct epistemic positions generate distinct instantiations of principles, another form of disagreement emerges because of skepticism over some epistemic principle. Since there is not an agreement as to the correct epistemic principles, this sort of disagreement is not an instance of sub-norm disagreement.

In certain cases, uncontroversial principles can be drawn upon to resolve differences of opinion over correct principles. For example, suppose two hunters disagree over whether certain features of a track indicate that the track was left by a coyote or by a domestic dog. One hunter follows the principle of inferring that these features are characteristic of a dog track, the other that they are characteristic of a coyote track. However, both agree that a respected elder hunter knows about such matters, so they consult the elder’s opinion to settle their disagreement. In this case, a disagreement about one epistemic principle—whether the relevant characteristics indicate a dog or coyote track—is settled by using a more fundamental epistemic principle that both hunters agree with—that the elder hunter knows about such matters and her opinion can be counted on.

In the preceding example, the principle that the elder’s testimony is reliable on these matters is not controversial. Application of this norm settles the dispute. However, occasionally a disagreement occurs that turns on the application of what Lynch calls a fundamental epistemic source principle (FESP). A FESP is type of epistemic principle that does not only indicate that a certain source of evidence is reliable. According to Lynch, a FESP “is a principle such that it can’t be shown to be true without employing the source that the relevant principle endorses as reliable” (Lynch 2016, p. 250). Unlike the disagreement over the footprints just considered, a disagreement that turns on a FESP cannot be resolved by appealing to a more fundamental epistemic principle.

Common examples of FESPs discussed in the philosophy literature are the principle that induction is reliable or that observation is reliable. David Hume famously established that it is not possible to show that induction is reliable without using induction itself. In arguing for a FESP “we hit bedrock,” we encounter a point at which we can’t muster epistemic reasons that would rationally persuade a person doubtful of the reliability of the FESP in the first place.

Lynch explicitly calls disagreements that turn on conflicting FESPs deep disagreements. Lynch claims that these disagreements are deep because it is not possible to provide epistemic reasons that the skeptic of the principle would recognize as reasons for or against the principle in question. Why is this not a possibility? The reason is that the FESP that one person accepts the other doesn’t. FESPs, by definition, need to employ the source of evidence that they claim is reliable in a justification of the FESP. Thus, a person who doesn’t accept some FESP won’t regard its source of evidence as reliable. And, as a result, they won’t recognize the reasons that might be given in favor of the FESP’s truth as rationally persuasive. Therefore, a person who does not accept some FESP won’t find any argument for the FESP epistemically cogent. That is to say, no argument for the FESP that draws on epistemic reasons alone could be rationally persuasive to such a skeptic. After all, the skeptic would not recognize the premise that invoked the source in question as providing credible reasons to think that the principle in question was true. Such arguments would not provide sufficient reason to endorse the principle, and thus would not be rationally persuasive to the skeptic.

We call this sort of deep disagreement level-2 deep disagreements, or principle-based deep disagreements. These disagreements occur above the sub-principle level since there is not agreement over a correct E-system. Nevertheless, when these disagreements involve fundamental epistemic principles, they cannot be resolved through one party rationally persuading the other party on the basis of epistemic reasons alone.

3.2.1 Objections and Replies to the Level-1/Level-2 Deep Disagreement Distinction

As discussed above, the primary difference between sub-principle (level-1) and principle-level (level-2) deep disagreement is the cause of these two different kinds of disagreement.Footnote 9 The latter form of disagreement emerges because of disputes over correct epistemic principles. The former turns on variations in epistemic positions.

There is an important concern that someone might raise with this distinction that we will now consider. To articulate this concern requires some reflection on the way epistemic principles, or epistemic norms, were characterized earlier in this paper. Implicitly, we understood epistemic principles as being schematic in the sense that they contain “slots,” so to speak. These “slots” take different values supplied by the agent’s epistemic position. So, for instance, the “credible expert” slot in (T-norm) can take different values, and have different extensions, given distinct epistemic positions. Call this the schematic conception of epistemic principles. However, someone may prefer a characterization of epistemic principles in which concepts involved in the statement of the principle have a fixed extension. In other words, the principle is such that the extension of any concept employed within the epistemic principle is fixed and doesn’t vary with variations in epistemic position. For instance, relative to one context the epistemic principle would render evolutionary biologists as credible experts. In the other context, religious authorities could be credible experts. This view about epistemic principle regards, what we earlier characterized as different instantiations of (T-norm), as distinct epistemic principles all together. According to this understanding of epistemic principles, there is not really agreement over (T-norm) among those who come from different educational backgrounds. Rather, these agents are reasoning in accordance with distinct epistemic principles. Call this the non-schematic conception of epistemic principles. If the non-schematic conception is correct, then the disagreements discussed earlier that revolved around (T-norm) or (D-norm) are ultimately analyzed in terms of principle level disagreement rather than sub-principle level disagreement. The core of this concern is whether the level-1/level-2 distinction can be sustained across different conceptions of the level of generality at which epistemic principles are best conceptualized.Footnote 10 If epistemic principles are best conceptualized non-schematically, then the level-1/level-2 distinction cannot be sustained.

In order to reply to this concern, we contend that there are explanatory advantages associated with the schematic conception of epistemic principles. What are these advantages? Within certain argumentative exchanges arguers are likely to locate their difference of opinion as rooted in a difference in epistemic position (or some similar concept), while in others they are likely to locate their difference as rooted in a difference over epistemic principles themselves. For instance, the latter kind of understanding of the disagreement is likely to emerge among the participants to a dispute in which there is a strong agreement over (T-norm), but differences emerging over who should qualify as a credible expert based on coming from different communities (i.e. a secular community or a evangelical Christian community, for example). If all deep disagreements are disputes over epistemic principles, then arguers would be mistaken when they judge themselves to agree over principles, but disagree in their interpretation due to coming from different backgrounds or past experiences. There is at least a prima facie value in an account of deep disagreement that takes this argumentative behavior at face value.

However, in response to this reply, it is worth acknowledging the possibility arguers could be wrong in regarding themselves as agreeing over principles, but as differing because of their epistemic positioning. However, the burden of proof, in our view, rests with those who would regard this natural argumentative behavior as inherently misguided. The non-schematic account of epistemic principles requires all deep disagreements to be at the principle level. Thus, arguers who understand themselves to agree over epistemic principles but disagree due to worldview, background beliefs or being from different communities would be mistaken as to the nature of their disagreement on the non-schematic view. The account developed here takes it at face value that arguers are right in circumstances in which they locate the cause of the disagreement in divergent epistemic positions, rather than in a difference of opinion over epistemic principles.

3.3 Fogelin-Style Deep Disagreement

In a paper titled The Logic of Deep Disagreement, Robert Fogelin also characterized a type of disagreement as “deep.” According to Fogelin, normal disagreements occur within a framework of shared background beliefs. These shared background beliefs include facts relevant to the issue under dispute and about what sort of inferences and standards of argument evaluation are adequate (Fogelin 2005, p. 6). For example, do the participants assume classical inferences are valid? Is inference to the best explanation an acceptable mode of inference? If so, what determines whether one explanation is better than another? Do the participants share beliefs about what the basic facts are with respect to the dispute at hand?

Typically arguments occur in situations where the participants have a wide range of agreement on the basic facts and the standards of good reasoning. Deep disagreements occur, according to Fogelin, when there is an absence of relevant shared background beliefs (Godden and Brenner 2010, p. 43) and a clash of, so-called, “framework propositions” (Fogelin 2005, p. 8). If we don’t agree on the facts, or about how to reason well about them, then the facts to which I appeal, and the reasons I provide, are not likely to be registered as persuasive to you.

At first glance, Fogelin’s account seems similar to Lynch’s. And in some ways it is; but there is a significant difference between the kind of disagreement that Fogelin has in mind and the other two kinds we’ve discussed.

For the purposes of illustrating these similarities and differences, it will help to step out of the epistemological realm for a moment and consider Fogelin’s example of an analogous deep moral disagreement over whether abortion is morally permissible. In such a disagreement, it is plausible that the parties share some basic moral principles in common. For instance, the parties could agree about the intrinsic value, or sanctity, of human life. However, according to Fogelin, in spite of this agreement, in many cases the disagreement cannot be resolved by appeal to facts or by appeal to moral principles (Fogelin 2005, p. 8). Fogelin says,

One diagnosis of the situation is tempting. What I have called deep disagreements are generated by conflicts between framework propositions. They remain recalcitrant to adjudication because the sources of the disagreement—the framework propositions—are allowed to lie in the background, working at a distance. The way to put the debate on a rational basis is to surface these background propositions and then discuss them directly. (Fogelin 2005, p. 8)

If we assume, as seems likely, that Fogelin would draw a similar lesson in the case of epistemic disagreement, then FESPs are another example of the sort of thing that Fogelin means by “framework propositions.”Footnote 11 And, if we consider an analogous diagnosis for deep epistemic disagreement to the diagnosis Fogelin entertains for the moral case, then we can assume he would think that some deep disagreements pivot on a dispute over specific FESPs. To that extent, a Fogelin-style account of deep epistemic disagreement would seem very similar to Lynch-style deep disagreement. However, Fogelin is clear that he is unsatisfied with this account of deep disagreement. According to Fogelin, it is unclear that the disagreement is best understood as pivoting around specific framework propositions. Fogelin explains,

On the one side someone will hold that at conception... an immortal soul enters the fertilized egg and with this, personhood is attained. Why should one believe anything like this? Well, this is part of a wider tradition, grounded in revelation, and sustained and deepened by faith. When we inquire into the root of deep disagreement, we do not simply find isolated propositions... but instead a whole system of mutually supporting propositions (and paradigms, models, styles of acting and thinking). (Fogelin 2005, p. 9)

Fogelin-style deep disagreements do not turn on the truth of specific framework propositions. These disagreements involve clashes among overarching frameworks themselves. And these frameworks, Fogelin suggests, are frameworks of understanding. As such, they determine not only what people believe, but how they understand the terms they use to communicate those beliefs. Hence a deep disagreement, in Fogelin’s sense, involves a disagreement not just over the truth of a principle or principles but in how to understand the relevant principles. Put differently, a disagreement is deep on Fogelin’s characterization when the dispute’s participants possess radically different conceptual frameworks, resulting in divergent and conflicting views as to what the facts are, and about how to properly evaluate the evidence pertaining to the issue under dispute. But while these consequences mean that the argument can’t be settled by appeal to cogent epistemic considerations, the dispute is informed and animated by different conceptual frameworks—that is, by different traditions, conventions and overarching modes of interpreting and understanding reality, society, politics, morality and knowledge.Footnote 12

Thus, Fogelin-style disagreement is a level above principle-level deep disagreement. It occurs at the level of conflicting conceptual frameworks and as a result, over how to understand networks of claims, not (just) over the truth of a specific epistemic principle, or even a set of specific epistemic principles. Thus, Fogelin-style deep disagreement can also be called level-3 deep disagreement, or framework-level deep disagreement.Footnote 13

To drive home this distinction, consider how this contrasts with Lynch-style disagreements. According to Lynch, unless there is a common basis for shared meaning among the parties to the dispute, it’s not clear that there is a genuine disagreement as opposed to a difference in meaning. While it is unclear how much agreement (or lack thereof) is required for a Fogelin-style deep disagreement to occur, in a Lynch-style deep disagreement, the disputants must, at least, possess shared epistemic goals and a similar understanding of the FESP under dispute (Lynch 2010, p. 265). For instance, a skeptic and a proponent of the principle that induction is reliable must at least have a similar grasp of the meaning of such a principle in order to disagree over the principle’s reliability. Thus, one reason to think that Fogelin and Lynch have something different in mind is the different ways that the two emphasize a shared background understanding among the parties to the disagreement. In Lynch’s case, it is crucial that there is a shared meaning over the key propositions and principles involved, whereas Fogelin tends to emphasize the lack of shared meaning in contexts of deep disagreement. Thus, it is entirely conceivable that a Lynch-style deep disagreement could occur between those who share overarching conceptual frameworks in Fogelin’s sense. Arguably, for instance, this is the sort of dispute that occurs when two empiricist philosophers, with similar educational and cultural backgrounds, have an argument on the reliability of some inductive principle.

Similar points distinguish Fogelin-style deep disagreement from Goldman-style deep disagreement. Goldman-style disagreements concern epistemic position—that is, how to interpret or assign constants to the variables in norm schemata. Fogelin-style deep disagreements may involve this as well—since differences with respect to epistemic position may be part of conflicting frameworks. But a Fogelin-style disagreement is not just over how to make assignments for variables in a shared schemata, but over how to understand the non-varying elements of the schema too—that is, over networks of concepts such as, e.g. “belief,” “reason,” or “evidence.” In other words, Fogelin-style disagreement concerns how to even understand what lurks in the background of other, less holistic, kinds of deep disagreement.

To state the differences between these three kinds of disagreement sharply; Lynch-style disagreements are disagreements over the truth of epistemic principles, Goldman-style disagreements are disagreements about how to assign content to schematic epistemic principles and Fogelin-style disagreements are holistic disagreements in meaning not truth, they are disagreements over how to understand networks of epistemic concepts and the beliefs those concepts compose.

It is also clearly the case that within Fogelin-style deep disagreements the prospect of one party rationally persuading the other through epistemic reasons is undercut. Indeed, Fogelin takes this a step further, claiming that his notion of deep disagreement undercuts the very possibility of argument all together. Fogelin stresses that he is not just endorsing,

The weak claim that in such contexts arguments cannot be settled. It is the stronger claim that the conditions for arguing do not exist. The language of the argument may persist, but it becomes pointless since it makes an appeal to something that does not exist: a shared background of beliefs and preferences. (Fogelin 2005, p. 9)

It is worth noting that Seigel (2013) and Patterson (2013) have expressed doubts as to whether there are genuine cases of Fogelin-style deep disagreement. One thought is that Fogelin-style disagreements are not genuine disagreements at all. Rather, they are merely disagreements in meaning.

We agree Fogelin-style disagreements involve disagreements in meaning, although, for familiar Quinean reasons, we are suspicious about whether there is anything “mere” about such disagreements. Moreover, when disagreements in meaning are holistic enough, as Fogelin-style disagreements purport to be, they will have the consequence the parties won’t be able to appeal to epistemically cogent reasons in order to persuade one another. That in turn ensures that they count as deep in the sense of the term we’ve articulated.

That said, answering the larger question of whether there are any genuine Fogelin-style deep disagreements is not our purpose here. But we do note one promising thing about them should they exist—it may be possible that these disputes could move to a more normal setting once the participants pin down the meaning of key terms involved in the dispute. At that point there may be no issue in the participants offering epistemically cogent arguments to the other. That’s an optimistic thought, and one that informs the speculative point with which we’ll conclude the paper.

4 Rationally Navigating Deep Disagreement

As we have seen, in each of the varieties of deep disagreement considered, it is not possible for the parties to the dispute to rationally persuade one another by only offering epistemic reasons for their conflicting points of view. Why? The reason is that the epistemic reasons that one party to the dispute should recognize as offering sufficient support for their standpoint should be not recognized by the other party to the dispute as offering sufficient support for that standpoint. One response to this situation is to “throw your hands up” and abandon the prospect of fruitful argumentation.

However, various approaches have been proposed for rationally navigating, if not resolving, these kinds of difficult disputes.Footnote 14 Someone may wonder whether the tripartite distinction between three “levels” of deep disagreement means that the solutions that have been proposed could only apply to one kind of disagreement. We think not. At least one strategy may be applicable to any of the varieties of deep disagreement considered here.

The strategy we have in mind is articulated in Case Sunstein’s (2018) discussion of “incompletely theorized agreements” in his Legal Reasoning and Political Conflict. Incompletely theorized agreements occur, “when people agree on something important, even though they disagree on a lot” (Sunstein 2018, p. xi). Even in the midst of disagreement motivated by differences in epistemic position, FESPs and overarching conceptual frameworks, there may be some claims, and some principles, to which the parties to the dispute can agree. For instance, two parties disagreeing over which of two fundamental epistemic principles is true may agree on other epistemic principles; two parties disagreeing over how to assign for the variables in a schematic E-norm may nonetheless share other E-norms or a common understanding of their non-variable terms. And parties with distinct epistemic conceptual frameworks may still find that those frameworks partly overlap. Of course, some elements of all of the above may be involved in the messy details of actual cases, as when two parties adopt completely different understandings of evolutionary theory and on the virtues of scientific experts on that subject but still appear to agree, in some general sense, that both religious and scientific perspectives are worthwhile teaching in schools.Footnote 15 The specifics of how, and to what extent, these different perspectives are taught is left unspecified and not fully theorized.

One way to flesh this suggestion out is to claim that the concepts employed in propositions like,

S: Scientific perspectives should be taught in schools

And,

R: Religious perspectives should be taught in schools

are “open-textured” (Waismann 1945) “unsettled” (Shapiro 2014) or “fluid” (Lynch 1998). Concepts like “religious,” “scientific perspectives” and “taught” are usefully vague in a particular way. That is, while all applications of the concept may share certain minimal characteristics, it is possible to enrich or further specify those concepts in distinct and non-overlapping ways. If so, then there may be room for decision makers to unpack and argue for specific interpretations of these principles in such a way that they can endorse both principles—precisely because they’ve enriched the content of the relevant concepts in distinct ways. The practice of how the principle gets acted on evolves in order to address the needs of teachers, students and communities over time, thereby impacting which principles they take as true, which assignments they make to shared schematic norms, and the entire shape of their conceptual framework. In this way, it is similar to legal rules that are interpreted, conceptualized and implemented by judicial-decision making bodies over time to address the needs emerging from unique contexts, and which in turn impact the broader legal frameworks employed by jurists.

Through identifying some point of agreement and drawing on those points to construct a framework for acting on the points of agreement, new practices emerge that enable deeply disagreeing parties to collectively navigate their different views in order to address pressing social needs that emerge when people with diverse perspectives live together. To reiterate, this does not mean that agreement over the contested issue emerges. For what constitutes agreement is now itself be a complex matter (see, e.g. Lynch 1998; Shapiro 2014). Relative to their distinct specifications of the relevant concepts, they neither agree nor disagree—they are talking past one another. Relative to the shared overlapping content of the concepts, however, their disagreement persists. And these two facts suggest that there is conceptual and political space for the parties to emphasize, and work from, common understanding rather than differences.

The above suggestion is speculative, of course. But it has proven to be surprisingly durable in accounts of disagreement in metaphysics (Lynch 1998) and the philosophy of logic (Shapiro 2014). Moreover, note that the possibility of applying this approach to rationally navigating deep disagreement can be employed if the disagreement is at the level of epistemic principles, or whether it emerges because of differences in epistemic position or framework. In each case, the idea is to look to principles that the parties can agree to and work from the varying principles, positions or frameworks to interpret those principles over time.

It is also worth sounding a cautionary note, and not being overly optimistic. In many circumstances people may form an incompletely theorized agreement, but go on to stack the deck. One way that the deck can be stacked is by assigning people to interpret and enforce the agreement so as to favor one side of the issue. This kind of stacking of the deck can occur, for example, when some parties to the dispute are not sincere in an effort to move forward in good faith to resolve the disagreement. Moreover, it is possible that the relevant disagreement may be so deep (in all three kinds of ways) that there is not enough significant overlap to employ this strategy. This could happen, for example, when the relevant conceptual frameworks involved are simply alien to one another. In such circumstances, the incompletely theorized agreement strategy will be limited in its effectiveness. However, this kind of approach, we submit, is worth investigating further as a way of trying to address important political and social issues in the context of deep disagreements, whether those disagreements are at the principle, sub-principle or framework level.

5 Conclusion

In the epistemic and argumentation theoretic literature the term ‘deep disagreement’ has been used to characterize a variety of different kinds of intractable differences of opinion. In this paper, we discussed three different kinds of disagreement that could be reasonably characterized as “deep.” In each case, the disagreement is deep because the participants to the disagreement are unable to provide an epistemically cogent argument to one another in support of their varying points of view. However, in spite of the inability to offer epistemically cogent arguments, it is nevertheless possible, in some circumstances, for deeply disagreeing parties to rationally navigate their deep disagreements. We discussed one strategy for navigating deep disagreements that has been raised in the literature, highlighting its versatility in responding to the varieties of deep disagreement considered here.