1 Introduction

One of the most vibrant developments in current epistemology is the exploration of epistemic ends and values other than knowledge, most notably understanding. However, exactly what understanding consists in and what it requires, are still highly controversial. An original approach is Catherine Elgin’s proposal that so-called objectual understanding (specifically, of a given topic or domain) is a matter of creating or improving reflective equilibrium in one’s system of thought (Elgin 1996, p. 118, 2017, pp. 3–4). On Elgin’s holistic account, such a system consists not only of particular factual judgments or beliefs, but also of theoretical commitments, acceptances,Footnote 1 generalizations, methods of inquiry, standards of justification, and epistemic goals and values (Elgin 1996, p. 106, 2017, p. 102). Call such a system a noetic network. In what follows, we discuss Elgin’s equilibrium analysis of understanding and argue that, among other merits, it has the potential to provide a seminal perspective on the social-epistemic function of epistemic authority: appropriately reconstructed, the account highlights the way in which epistemic authorities can foster understanding in their epistemic communities. This conclusion departs significantly from current mainstream approaches to epistemic authority, which locate its key function in the social transmission of true beliefs or knowledge. Our argument is this.

Many if not most of an agent’s attempts to straighten out noetic dissonance have an essentially social dimension. As Elgin herself notes, when we try to achieve reflective equilibrium we often profit from the division of cognitive labor (see, e.g., Elgin 1996, p. 118, 2017, pp. 4, 59, 108, 112). Of course, we do not always need to consult others,Footnote 2 but when complex theories come under pressure from new data, for instance, or when customary standards, methods, or principles of reasoning appear to lose their reliability, cognitive reformation rarely proceeds as a solipsistic enterprise. When noetic disruptions like these befall us, we are often at a loss to decide on our own which elements of our networks to abandon or adjust (and how to do this rationally), and which ones to preserve. What should we do in such cases?

What typically comes to the rescue is engagement with epistemic authorities. Epistemic authority, as we construe the notion, is a relation between an epistemic agent (or group of agents) S, a given topic or subject matter A, and some other epistemic agent (or group of agents) S* who enjoys a sufficient degree of epistemic superiority over S with respect to A. This may involve S*’s possessing more evidence or more (first-order) factual information than S. However, epistemic authorities are also more conscientious or reliable (or both) in handling the evidence: they possess superior methodological skills and more advanced higher-level competences to pursue certain sorts of epistemic aim. An agent may be an epistemic authority for you vis-à-vis topic A but not topic A*, and she may be an epistemic authority regarding A for me, but not for you. Moreover, S may catch up with, or even outperform, S* at some point, be it because S* loses competence or information or because S gains it over time. This suggests that epistemic authority should also be relativized to times: S* is an authority for S, regarding A, at or during a given (period of) time t. Note that this characterization focuses on an “objective” sense of the notion. Individuals and communities may (and, alas, often do) regard certain people as epistemic authorities who actually fail to possess the competences that would qualify them for this role. Such agents serve as mere de facto or mere reputational authorities, but fail to constitute “real” authorities.Footnote 3

The fact that we often (rationally) turn to epistemic authorities in order to create or reinstate reflective equilibrium in our noetic networks has the following interesting consequence. Suppose we agree with Elgin that creating reflective equilibrium produces or deepens understanding. Then epistemic authorities often serve as sources of understanding for their interlocutors. This conclusion differs significantly from traditional approaches. For example, in her influential book Epistemic Authority: A Theory of Trust, Authority, and Autonomy in Belief (2012), Linda Zagzebski says that she is “interested in the special case in which I have reason to believe that another person is more likely to get the truth about some question than I would be if I attempted to get the answer to the question by a direct use of my faculties” (2012, p. 105). “In cases of these kinds”, she continues,

the conscientious thing to do is to let the other person stand in for me in my attempt to get the truth in that domain and to adopt his belief. This in broad outline is what I mean by epistemic authority (Zagzebski 2012, p. 105).

Zagzebski’s approach is inspired by Joseph Raz’s theory of political authority and his application of it to “theoretical authority”. Raz argues that:

[j]ust as with any practical authority, the point of theoretical authority is to enable me to conform to reason, this time reason for belief, better than I would otherwise be able to do (Raz 2009, p. 155; see also 1986, p. 68).

Similarly, arguing that “practical authority tells you what to do”, Edna Ullmann-Margalit maintains that “[t]heoretical authority tells you what to believe” (Ullmann-Margalit 2007, pp. 54, 57). The examples could be multiplied.Footnote 4 The epistemic goal that these and many other authors consider is not understanding, but the (rational, conscientiously guided) acquisition and transmission of true belief. Yet, many have argued in recent years that understanding is not reducible to acquiring or sustaining true beliefs or even to knowledge.Footnote 5 If such anti-reductionism about understanding is correct, then belief-transmission models of epistemic authority cannot account for the fact that epistemic authorities often produce understanding in their interlocutors. In any case, even if one accepts reductionism, our argument shows that epistemic authority is good for far more than transmitting individual beliefs: it can also promote the richer and more ambitious epistemic goal of improving complex noetic networks in terms of understanding.

A related issue is the role of testimony in general. Various authors have explicitly argued that, whereas testimony can be a distinctive source of true beliefs and knowledge, it cannot be a source of understanding (see, e.g., Zagzebski 2008, pp. 145–146; Hills 2009; Pritchard 2010, p. 81). If our argument is on target, we must demur. Epistemic authorities of the sort in question can, by entering into testimonial relations, produce understanding in their interlocutors. Or so we shall argue.Footnote 6 One interesting consequence is that, if understanding is an achievement (as many argue), then, in the cases of epistemic improvement discussed here, significant credit for this achievement goes not just to the agent, but to the epistemic or intellectual authorities he or she consults. We conclude, therefore, that there are processes we can aptly call “understanding transmission” and that such processes involve a distinctive distribution of epistemic responsibility.

Before developing this argument in detail, some clarifications and disclaimers are in order. First, we said that Elgin is concerned with so-called objectual understanding of a given topic or domain. Examples include understanding climate change, understanding US politics, understanding libertarian theories of free will, and so forth. This kind of understanding is commonly distinguished, for example, from factual understanding of a single event or phenomenon (e.g., the election of Donald Trump); propositional understanding (that this mushroom is toxic); explanatory understanding (why the house burnt down); practical understanding (how to operate a chain saw); or symbolic or semiotic understanding (what a text, a diagram, a work of art means). That said, it should be noted that—as Elgin would agree—these distinctions turn out on closer inspection to be fuzzier than they may at first appear. For example, understanding a single event is not a piecemeal matter. It involves understanding why the event occurred as well as how its explanation is systematically connected with other aspects of the domain. Similarly, understanding a phenomenon such as current global warming requires symbolic understanding of theories about climate change and thus of the propositional and non-propositional representations they involve.Footnote 7

Second, we will not make heavy weather of the claim that epistemic agents strive for balance in their noetic networks. Granted, some people seem better able to tolerate, and less apt to suffer from, noetic distortions and incoherencies than others. What matters most for present purposes is the normative claim that agents should aspire to reinstate cognitive balance when things have gotten out of whack—that there is rational pressure to do so. Note that this does not in itself presuppose ambitious holistic interpretations. For example, even epistemological foundationalists (of knowledge or justification, for example) agree that agents ought to organize their minds systematically and that there is rational pressure to straighten out cognitive dissonance. Here we will not venture to defend this assumption but will simply accept it.

Third, note that we deliberately talk about “producing or deepening/approaching understanding”. Understanding is gradable. Reflective equilibrium should thus turn out to be gradable too. This makes sense; after all, balance need not be perfect. The tightrope walker may struggle but still keep her balance—even though she would be better off, safer, more stable, if her pole were not tilted. From a holistic point of view, a noetic network approaches equilibrium when epistemically relevant relations such as support relations in some (ideally non-isolated) subsystem, or connections between different subsystems, are multiplied or strengthened. This is typically not an all-or-nothing affair, but a matter of approximation.

In the next section, we clarify the notion of reflective equilibrium and identify ways in which agents may encounter trouble in their noetic networks. Section 3 explains why epistemic agents must often consult epistemic authorities to preserve, achieve, or approach reflective equilibrium; Sect. 4 distinguishes experts from authorities. Section 5 looks at the competences or epistemic virtues that such authorities need to fulfil this role. Recent discussions of epistemic authority have largely focused on laypersons or novices,Footnote 8 asking how they should rationally react to the views of epistemic authorities.Footnote 9 But what about authorities themselves? Which skills and virtues will they possess and exercise when they promote understanding in their epistemic communities? We argue that an important intellectual virtue in this context is what we call epistemic empathy.

2 Approaching reflective equilibrium

The notion of reflective equilibrium figures in two ways in the literature. On the one hand, it denotes a state of cognitive balance; on the other, it refers to the procedures or cognitive dynamics that yield such states (see e.g. Kelly and McGrath 2010; Walden 2013; Brun 2014; Cath 2016; de Maagt 2017; Baumberger and Brun 2016). Although the term was originally introduced in logic and ethics (Goodman 1953; Rawls 1971),Footnote 10 in recent literature reflective equilibrium is portrayed broadly as a “procedure for investigating … [a given] domain” (Kelly and McGrath 2010), “a deliberative process in which we reflect on and revise our beliefs about an area of inquiry, moral or non-moral” (Daniels 2016, p. 1). Some see it as nothing less than the “the status quo” of all philosophical reasoning (Ellis and McBrayer 2019) or even “the … rational method of inquiry” in general (DePaul 2005).

Where do such processes start, and how do they proceed? According to Elgin, we have no choice but to commence in medias res, i.e., begin with those “opinions, values, methods, and standards that we consider relevant and … that we have some inclination to accept” (Elgin 2017, p. 64). Such contents enjoy “initial tenability” (Elgin 1996, p. 102). She also reminds us that “our convictions form the basis for our actions” and that “confidence in a given judgment indicates that we have not yet found it an impediment to action” (ibid.).Footnote 11 The best way to interpret Elgin here, we suggest, is as saying that we have an initial reason to accept a certain content, as long as accepting it does not frustrate our success in action.Footnote 12

The problem, however, is that people are often inclined to accept cognitive products of bias, idiosyncrasy, delusion, or other manifestations of irrationality. (And indeed, unfortunately we not only do this initially.) For example, “superstitions, delusions, and prejudices begin their careers as initially tenable beliefs” (Elgin 1996, p. 102). So how should the method proceed from what appears “initially tenable”? Elgin says that “[c]onsistency is mandatory”, (1996, p. 103). But this requirement is not as easy to fulfill as it might appear. Not every contradiction is explicit, and some, whether explicit or implicit, are harder to detect than others. Moreover, it is often unclear how contradictions, once identified, should be resolved. This will typically not be a matter of simply rejecting certain contents and accepting others. Instead, we might have to restrict, or redraw, the conflicting claims’ scope or domain of application. And consistency does not suffice for equilibrium anyway: beliefs or acceptances might be consistent yet mutually irrelevant. For a system to be in reflective equilibrium, its elements must also “hang together”, be “mutually supportive”, and be “reasonable in light of one another” (Elgin 1996, p. 99, 2017, pp. 72, 98). Moreover, none of this ensures, or even makes it probable, that the network correctly represents the phenomena it deals with. Still, lifting it into equilibrium is “the best we can do in the epistemic circumstances” (Elgin 2017, p. 64).

Suppose, then, that you do your best to apply the method properly. You adjust your initially tenable beliefs and commitments (via appropriate revisions and expansions of the network) until all requirements of reflective equilibrium, as far as you can tell, are fulfilled, at which point you cease your noetic rearrangement. The resulting network may nonetheless fail to achieve equilibrium. For things could easily go wrong in the “reflective” part of the method: our meta-level assessments may go awry. “Appearances can be deceiving. We may believe, with reason, that a system is in equilibrium when in fact it is not” (Elgin 1996, p. 14).

In order to see where epistemic intervention from others can help, it will be useful to distinguish the various ways in which a noetic system may fail to be in reflective equilibrium. Two basic considerations are whether or not the system is in balance and whether or not the agent is reflectively aware of this. As far as we can tell, theorists of reflective equilibrium have not offered in-depth analyses of the “reflective” part of the notion. Daniels (2016), in the passage quoted above, characterizes the procedure as a “deliberative process in which we reflect on and revise our beliefs about an area of inquiry”. However, the objects of such deliberation are not exhausted by beliefs. Recall that in Elgin’s approach, the targets also consist of acceptances, commitments, methods, standards, epistemic goals, and so forth. Moreover, deliberation itself may be construed as a cognitive process that yields beliefs, states of acceptance, commitments, and the like. It also yields higher-order knowledge, in which case we shall say that the agent “is aware” of the system’s condition. For simplicity’s sake, let us look at such cases and cases in which agents lack such awareness.

First, consider an imbalanced system. The agent may become aware of some or all of the imbalances—call this a state of reflective imbalance. She may nevertheless have no idea how to reform the network. Secondly, she may also fail to become aware of the imbalances, for example because she never engages in any in-depth assessment of her noetic network in the first place (call this unreflective imbalance). Third, often people not only lack awareness of the poor status of their systems, but even positively believe that the latter are in balance when they are not. In such cases, agents suffer from illusions of balance. Next, consider a system that is in balance. Once more, the agent may be aware of this, a state we may call reflective balance (so far the only epistemically good case); or she may be unaware of the balance (unreflective balance), or even falsely believe that the system is imbalanced (illusion of imbalance).

More distinctions could be drawn. We have noted already that noetic balance is not an all-or-nothing affair. Moreover, the reflective part too is gradable. Sometimes our self-assessments produce outright beliefs, but we also harbor partial or graded beliefs about the (degrees of) balance that our noetic profiles enjoy. However, we shall set aside further complications here; what matters is that the method of reflective equilibrium aims at delivering epistemic results that are, as Elgin puts it (2017, pp. 67, 11), reasonable in the overall “epistemic circumstances” and not just relative to the best of an agent’s subjective ability to assess her cognitive network. What, then, could agents do to meet this condition?

3 Equilibrium, authority, and understanding

Rawls recommended striving for what he calls wide reflective equilibrium. To achieve wide equilibrium agents must compare their commitments and beliefs with alternative viewpoints.Footnote 13 Elgin agrees: “An acceptable cognitive system must answer not just to my initially tenable commitments, but to ours” (Elgin 1996, p. 116, our emphasis). Of course, not every voice is worth hearing (certainly not on everything). There is “little to be gained by entertaining the economic views of someone convinced that extraterrestrial forces control the stock market” (cf. Elgin 1996, p. 117, 2017, p. 114). A rational move is to socialize one’s epistemic endeavors, but not randomly. While we can learn from testifiers with many different kinds and degrees of competence, it will often be particularly effective to consult those who are more advanced regarding some relevant epistemic aim. In other words, a rational move is to turn to epistemic authorities. We have already given a structural characterization of epistemic authority. But how exactly can epistemic authorities help agents achieve reflective equilibrium, and on which competences and intellectual virtues do they rely when they do so?

Suppose you are in a state of what we called reflective imbalance (i.e., your system fails to be in equilibrium, and you are aware of this). Add that, although you are trying your best to achieve equilibrium, you are struggling. Maybe you have noticed a contradiction but find it hard to decide which content to preserve and which to discard. Suitable epistemic authorities can help you figure out how best to resolve the tension. Or imagine that you have observed a phenomenon P under normal perceptual conditions, but the occurrence of P doesn’t fit your expectations. For example, suppose that in December you are walking along the Swiss bank of Lake Constance and observe what appears to be a flock of wild flamingos. However, you also have the well-grounded, rational belief that there are no wild flamingos in Switzerland, or indeed anywhere else so far north in Europe. You call the local ornithological society and they confirm that a flock of wild flamingos has indeed recently settled at Lake Constance. They hypothesize that climate change has altered the conditions at Lake Constance in recent years and enabled flamingos to survive there. In general, an epistemic authority will be able to help you answer such questions as: Why am I observing P? Why does this phenomenon appear in the way it does? The authority might explain how to update—expand, contract, or rearrange—your system to integrate the information that P is occurring. Moreover, if you had been expecting P not to occur, the authority will be able to help you figure out which assumption or cluster of assumptions is responsible for your misguided expectations (e.g., “It is too cold for flamingos to survive at Lake Constance, and they don’t have the right kind of food there”). Similarly, if your system is contaminated with bad or false explanations (“The flamingos must have escaped from a nearby zoo!”), the authority can show you what is wrong with these explanations, given currently acknowledged background information, epistemic standards, and available alternatives. (“The flamingos we are observing at Lake Constance have no tags, and there is no zoo nearby!”) When laypeople interact with epistemic authorities in such ways, the latter serve as sources of learning, and the laypeople’s cognitive progress amounts to progress in understanding.Footnote 14

Note that the relevant procedures might also involve highly destructive steps. Consider cases in which you suffer from what we called an illusion of balance. You think that things fit together in your network, but are mistaken. An epistemic authority is in a position to make you aware that something went wrong in your self-assessment. She will show you the flaws in your views, e.g., by pointing to implausible consequences of your beliefs and commitments that you failed to notice, or by uncovering contradictions or other noetic dissonances. This is how Plato’s Socrates often conducts philosophical discourse. Typically, an important first step is to shatter his interlocutors’ hubris by demonstrating that their deep conviction that they know what they are talking about is false, or unwarranted, or in tension with other things they believe. Remember for example Euthyphro, who—initially very self-confidently—confronts Socrates with his strong conviction that he knows how to characterize the pious (hosion) and the impious (ahosion). Socrates rapidly corners Euthyphro and gets him to concede that he doesn’t really know how to say what he means, and that his arguments, on whatever premises he tries to build them, constantly “turn round and walk away from us” (Plato, Euthyphro, 11b). Eventually Euthyphro gives up and walks away himself (following his arguments, so to speak, perhaps trying to catch and improve them), promising to face Socrates in continued conversation another time.Footnote 15 In the Apology (23b), the general aim of the Socratic elenchus (his art of dialectic refutation, as we may translate) is said to be to explore whether anyone is wise (sophos) at all, and if any given person turns out not to be, to persuade them of this fact. Unfortunately every candidate Socrates meets, at least among politicians, poets, and craftsmen, turns out not to be wise.Footnote 16

All this should not be misread as the claim that novices, or other interlocutors who understand less than an authority, cannot play important roles in furthering the latter’s understanding. The layperson too can inspire the more experienced inquirer to think more deeply, to reconsider the presuppositions of a given claim, to consider things from different angles, or to support her views at points she had regarded as self-evident.Footnote 17 All this will deepen her understanding. That said, it would nonetheless be misguided to maintain that less advanced interlocutors can assist epistemic agents in equally systematic, effective, or reliable ways as epistemic authorities can.

4 Experts and authorities

As noted earlier, Elgin herself discusses some social-epistemic dimensions of reflective equilibrium. “Rather than relying exclusively on considerations in my ken”, she says, “I draw on the expertise of others, and they in turn draw on mine” (Elgin 1996, p. 114; cf. 2017, p. 112); we rely on the views of “competent peers with a suitable level of expertise” (2017, p. 115). Those with a suitable level of expertise are experts in a given discipline, so Elgin’s suggestion is that we might attain adequate, socially integrated forms of reflective equilibrium by consulting disciplinary experts. However, though experts may serve as epistemic authorities, the latter have a skill-set that is broader than mere competence in their discipline. In fact, being an expert is not even necessary for being an authority.

First, there appear to be domains for which agents receive authoritative epistemic support from others, but for which there are no experts. For example, Williams (1995) and more recently Hills (2009) have influentially argued that there are no moral experts.Footnote 18 Yet we consult others all the time when facing moral decisions, especially hard ones. Among those we consult are exemplars or role models whom we clearly treat, and refer to, as moral authorities. What we receive from them is moral understanding or minimally, justified moral belief, hence their moral authority involves epistemic authority. Like most philosophical arguments, the Williams–Hills argument is controversial.Footnote 19 Since space prohibits entering this debate, our conclusion is thus tentative and conditional. There is a case to be made that justified moral beliefs and moral understanding can be obtained from morally advanced interlocutors even though there are no moral experts. If so, some authorities don’t constitute experts.

A second argument is straightforward. Epistemic authority, as our above characterization made clear, is an essentially social notion. Croce (2018a) puts it this way: whether S* is an epistemic authority for S is subject-dependent; it depends on who S is. Being an expert, however, is subject-independent.Footnote 20 Consequently, someone’s position as an authority may vary with respect to different persons while her position as an expert does not. Consider a child who seeks advice from her elementary school teacher on some elementary math question. The teacher is in a position to help and constitutes an epistemic authority for the child. But elementary school teachers are not typically expert mathematicians—they do not need to be. Being an expert, then, is not necessary for being an epistemic authority. Elgin might reply that on her view (as we saw), experts are “those who are competent, conscientious, and suitably sensitive to the level of confidence that their evidence supplies” (2017, p. 116), and that this will usually apply to math teachers. This may be correct, but since one would not classify the elementary school algebra teacher as a disciplinary expert in algebra,Footnote 21 Elgin’s idea requires fine-tuning.

Perhaps the most prominent account of expertise is Alvin Goldman’s, on which experts must (1) “possess a substantial body of truths in the target domain” and (2) have “a capacity or disposition to deploy or exploit this fund of information to form beliefs in true answers to new questions that may be posed in the domain”. Importantly for our purposes, Goldman emphasizes that “some non-comparative threshold of veritistic attainment must be reached” (2001, p. 145). On Goldman’s view, if the teacher fails to meet this threshold or lacks this disposition, then she is not an expert. Footnote 22 In more recent work Goldman suggests that:

S is an expert about domain D if and only if (A) S has more true beliefs (or high credences) in propositions concerning D than most people do, and fewer false beliefs; and (B) the absolute number of true beliefs S has about propositions in D is very substantial (Goldman 2018, p. 5).

This may also rule out the average elementary school teacher as an expert in algebra. Thus on this account too being an expert is not necessary for constituting or acting as an epistemic authority.

Nor is a high level of expertise—one that is high enough for being an expert—sufficient for epistemic authority; even an objectively high level of expertise may not be enough to render S* an authority for S. As noted, epistemic authority is subject-dependent. So suppose that S* is an expert, but S is approximately on a par with S*. Then S*, though being an expert—and, as it may be, an epistemic authority for many others in the relevant epistemic community—is not an authority for S.

Of course, none of this is to deny that there are—perhaps many—epistemic authorities who are also experts; the two classes overlap. We may call epistemic authorities who are also experts expert authorities.Footnote 23 Our arguments just show that the two notions should be kept apart, and that it is the notion of epistemic authority, rather than expertise, that an account of the social dimensions of reflective equilibrium should invoke.Footnote 24

The literature also contains a functional account of expertise the spirit of which is akin to our characterization of epistemic authority. For example, David Coady argues that an expert is “someone laypeople can go to in order to receive accurate answers to their questions” (2012, p. 39). Goldman (2018) also flirts with this approach (which is sometimes called the “service conception”), except that he is not concerned with understanding:

S is an expert in domain D if and only if S has the capacity to help others (especially laypersons) solve a variety of problems in D or execute an assortment of tasks in D which the latter would not be able to solve or execute on their own. S can provide such help by imparting to the layperson (or other client) his/her distinctive knowledge or skills (Goldman 2018, p. 4).Footnote 25

Now Goldman also urges—we believe rightly—that it is important to know the ground or basis of the capacities in question. His answer invokes the “truth-linked” or “veritistic” conditions quoted above: experts must have more true beliefs and fewer false ones in their domain than most members of their epistemic community, and the absolute number of their true beliefs must be substantial. On this account, the elementary-school teacher does not generally come out as an expert, which is the right result. However, the teacher clearly has the capacity to “help laypersons”: he serves as an epistemic authority for his pupils. But what explains this ability? It is not the amount of his true mathematical beliefs, for that amount is not sufficiently substantial. We are left, then, with a crucial question: what other factors beyond or instead of possessing a substantial body of true disciplinary beliefs can account for an agent’s ability to (systematically and reliablyFootnote 26) guide less experienced inquirers in their attempts to achieve reflective equilibrium or understanding?

5 Epistemic empathy

Recent debates on epistemic authority have largely focused on which epistemic skills and virtues agents must possess and activate in order to identify and rationally respond to epistemic authorities.Footnote 27 But the foregoing discussion also puts a different question on the agenda: if certain authorities have the ability to guide interlocutors in their quest for understanding, then which skills and competences does this require?

We propose that certain intellectual virtues will play a role here, including traditional ones such as intellectual thoroughness, fair-mindedness, or perseverance. There is also another epistemic virtue however, which, though largely neglected in the literature, plays a crucial role in this context. We call it epistemic empathy.Footnote 28 In a first approximation, epistemic empathy is the ability—or, if exercised, the mental state—of putting oneself in someone else’s epistemic position and of inspecting certain areas of it, thereby learning how the world looks from that perspective. To avoid misunderstandings: Our claim is not that epistemic authorities can always fully achieve this. Typically, there will be limits, be it because the epistemic distance between the interlocutors is too great, be it for principled reasons concerning particular experiences or perspectives of the agent that are hard or impossible for her to share with those who lack such perspectives. So, depending on the interlocutors, attempts to empathize epistemically may succeed to a greater or lesser extent, and certain perspectives may even in principle be inaccessible to outsiders. All this does not undermine the claim, however, that epistemic empathy exists, and that attempts to engage in it will often constitute a desirable social-epistemic goal.

Characterizations of empathy vary widely. A major divide separates views that focus on affective phenomena and those that construe empathy more broadly in terms of emulating other peoples’ cognitive (but not necessarily emotional) perspectives. For example, Karsten Stueber (2006, p. 28) argues that empathy is “a form of inner or mental imitation for the purpose of gaining knowledge of other minds”. More recently Justin Steinberg (2014, p. 50) says, similarly, that “empathy involves taking up the perspective of another, which means understanding something about the content of another’s mind.” Our proposal is in line with such more general approaches that allow for non-affective forms of empathy. Moreover, our specific notion of epistemic empathy must be understood broadly. In an Elgin-style framework, for example, the relevant areas of a noetic network (and thus possible targets of epistemic empathy) include not only knowledge and belief, but also commitments, acceptances, various kinds of experience, methods of inquiry, cognitive goals and values, as well as relations among these items, such as relations of support, conflict, and so forth. Such items too are in the target set of (acts of) epistemic empathy.Footnote 29

Importantly, the epistemic empathizer need not endorse the view that she empathizes with. Instead, she preserves cognitive and evaluative distance, remains aware of (and typically stands by) her own view, which she is in a position to compare with the view she empathizes with. For example, if an agent believes that p whereas an authority believes that not-p, the latter’s epistemic empathy does not involve her switching to the belief that p.Footnote 30 Instead, in such cases the empathizer identifies, but does not adopt, the agent’s belief and sees how it is embedded in the agent’s system. Along similar lines, we can understand what a false theory states despite the theory’s being false. Here, too, we realize how things look from the perspective of the system, but do not endorse it. This does not rule out, of course, that the epistemic empathizer also shares attitudes with her interlocutor. For example, while disagreeing with him about whether p, she may nonetheless share his epistemic goals, the belief that q, or his acceptance of r. Moreover, like such objects of epistemic empathy, nor are the attitudes that constitute the authority’s epistemic empathy restricted to beliefs. They too can be of various kinds. They may consist of (justified, or reliable, rational, apt…) beliefs, partial beliefs or credences, attitudes of mere acceptance, knowledge and, of course, understanding.

Empathy is standardly taken to be a success concept, and epistemic empathy is no exception. Compare affective empathy: If you feel bereft on your neighbor’s behalf because her partner has left her, yet rather than being unhappy about this herself she is in very good spirits, you do not empathize with her. Your mental state has missed its target, which is her mental state. That said, empathy is not an all-or-nothing affair. An epistemic empathizer may correctly represent some aspects of her interlocutor’s epistemic profile but misconstrue others.Footnote 31

This picture can be refined. Epistemic empathy can be local or extended and shallow versus deep. It is local insofar as it concerns only certain regions of an epistemic system. The more subsystems the authority empathizes with, the farther her empathetic insight extends. Moreover, if the authority’s empathy encompasses relatively few aspects of a given agent’s noetic system or subsystem, her empathy is shallow; her empathy is deep to the extent that she correctly represents more aspects. This observation also sheds light on the fact that some epistemic authorities are better than others, and on the fact that some are better in some respects, whereas others are better in other respects. An important comparative dimension is how local or extended, and how shallow or deep, a given authority’s epistemic empathy is. If you can choose between consulting someone who appears to have shallow and local epistemic empathy with you, thereby understanding you only in a shallow and local fashion, and someone who appears to bear deep and extended epistemic empathy towards you, it might be wise to go for the latter.

Here are some illustrations of how epistemic empathy works. Take our flamingos-in-December-at-Lake-Constance example (which is inspired by a true incident). You call the local ornithological society and inform them that you observed what appeared to you to be a flock of flamingos on the Swiss shore of Lake Constance. “However,” you remark, “I am pretty sure that wild flamingos don’t live on Lake Constance, so might they have escaped from a nearby zoo?” The expert replies that, yes, Lake Constance was previously not a flamingo habitat and that, indeed, wild flamingos have never before lived in Switzerland or this far north in Europe. “Yet,” she explains, “this winter a flock of wild flamingos has settled at Lake Constance; we know that they did not escape from a zoo. Our experts believe that their presence can be explained by climate change. Nowadays the winters at the Lake are much milder than in the past.” What the authority has done is extrapolate several features of the questioner’s epistemic system from what the latter explicitly reveals, including his reasons for disbelieving that the flamingos at Lake Constance came from the wild. For example, the authority premises her response on the hypothesis that the interlocutor believes that wild flamingos still don’t live that far north in Europe and that they are fit for areas with a much milder climate than today’s Switzerland. Yet, the authority herself neither shares these reasons nor the belief or conjecture they support (namely that “the flamingos at Lake Constance have escaped from a nearby zoo”). She has superior knowledge and understanding to her interlocutor, and in light of what she gathers about his noetic system (including that he is aware of the basics of climate change, which she does not bother explaining to him), she provides him with information he needs to update it and to improve his understanding of the phenomenon at issue. In this case, the interlocutor’s update may involve connecting two previously unrelated elements of his noetic network, namely the perceptual data and beliefs he has about the flamingos, on the one hand, and what he knows about climate change on the other.

Next consider some examples from fiction. A famous epistemic authority is Conan Doyle’s Sherlock Holmes, who is regularly a few steps ahead of his friend Dr. Watson (let alone other characters). Sometimes Holmes and Watson share the same evidence, and yet Holmes interprets and explains it with more skill and explanatory creativity. Other times, Holmes uses better methods of inquiry or is more observant in collecting evidence in the first place. Holmes often explains to Watson which clues suggest which hypothesis, and in so doing he projects himself into Watson’s less sophisticated thought patterns and guides him, by anticipating how Watson will reason, through noetic processes that eventually lead Watson to form the correct view about the case at hand. (Sometimes Watson remains clueless in spite of Holmes’s prompting and Holmes just has to give him the answer. In such cases, Holmes fails to serve as a source of understanding.) Such examples are not confined to lighter fiction. Consider the intellectual interactions between the philosopher and Franciscan friar William of Baskerville and his intellectual and spiritual student, the Benedictine novice Adso of Melk, in Umberto Eco’s The Name of the Rose. Step by step, the two discover what lies behind the mysteries and murders that haunt the Benedictine abbey they are visiting. (It turns out that the disturbing events are connected with the fact that the abbey’s labyrinthine library contains the only surviving copy of the lost second book of Aristotle’s Poetics.) Adso often comes up with various incomplete or erroneous hypotheses, yet even when he doesn’t explicitly lay them out, William tends to know or anticipate them. What is going on is that William, who is a great intellectual authority for Adso and more generally for the relevant epistemic community (of 14th century monks and theologians), enjoys great empathetic insight into Adso’s overall noetic system. William guides Adso through processes of correcting and updating false assumptions, and leads him to see the relevance of certain data that he misinterpreted or overlooked and to draw the right conclusions.Footnote 32

A fourth example comes from philosophy itself. The ways in which William relates to Adso have Socratic features. In Sect. 3 we mentioned Socrates, whom Plato portrays as someone who, though often claiming ignorance about whatever philosophical question arises, has the intellectual and dialectical skills to show his interlocutors where their reasoning goes astray. We said above that these include the skills to destroy the agents’ (especially the sophist’s) hubris and their conviction that they know what they are talking about. We may add here that Socrates’ effectiveness in his elenchus—his art of refuting his respondents’ poorly grounded claims and getting them to concede the incoherence of their views—often involves anticipating the cognitive dynamics in which they will engage. This in turn requires the ability to project himself into their minds. After pinning down some claim p that his interlocutor endorses, Socrates often anticipates additional beliefs and further commitments on the latter’s part, only to demonstrate in a next step, when the interlocutor articulates these views, that they are inconsistent with p. Although (and because) this process is to some extent destructive, Socrates’ insights into his interlocutors’ noetic system, and his ability to convey them to his interlocutors, enable the latter to make progress in understanding.

Socrates also exercises the intellectual skill of what Watson, J.C (2018) and Watson, L (2018) calls the skill of good questioning. However, whereas Watson considers this skill from the perspective of agents who wish to gather information and use it for themselves (an important topic), in a Socratic dialogue the epistemic authority uses it to foster understanding in others. And an important condition for this application of the skill of good questioning, it seems, is the ability to exercise epistemic empathy. Only if you can project yourself into another agent’s thinking in the ways we have outlined—only if you can construct a sufficiently detailed picture of the interlocutor’s noetic network, with both its good parts and its problematic areas—will you be able to ask questions that prompt her to improve that system and bring it into equilibrium.

In Sect. 4 we distinguished experts from authorities, but conceded that the two classes overlap.Footnote 33 We can add now that an (expert) authority who is able to guide her interlocutors in their attempts to achieve or improve their understanding, will be epistemically empathetic. Note that we are not claiming that epistemic empathy is necessary for producing understanding in one way or another. We maintain instead that epistemic empathy is necessary for having the ability (deliberately and reliably) to produce understanding in others.Footnote 34 Surely an authority could produce understanding by happy accident and without epistemic empathy. Yet in such cases, this result would not be attributable to the authority’s ability to guide or teach her interlocutor. (Abilities are standardly construed as dispositions or potentialities which, in the right circumstances, reliably produce certain results.)

The above examples illustrate this point. William of Baskerville may be an expert in philosophy, theology, church dogma, and so on. But he would not be a systematic, reliable source of understanding for Adso if he could not epistemically empathize with him. Similar observations apply to our Holmes and Socrates examples. Arguably, these agents too have special expertise. Yet without epistemic empathy, they could not teach and guide their interlocutors in improving their understanding. Consider the following analogy. Imagine a vast network of water reservoirs, akin, e.g., to a system of Nepalese rice terraces, most of which are connected through a complex channel network. Some of the reservoirs are filled, others contain some water but are not full, and other areas are entirely dry—either the higher-level reservoirs and the channels and ditches that would fill them are dry or closed off. Suppose that the system is so large and complex that the farmer who is in charge, though he tries hard to fill as many lakes and ponds as possible so as to keep as many areas as possible fertile, can only progress slowly and laboriously on his own. Many fields remain dry, many remain with little if any water, various channels are blocked, and watering other fields would require building new channels from scratch. One day, an experienced fellow-farmer visits the site and offers help. She brings maps of the whole area and realizes where things go wrong. Importantly, she has the competence to identify areas of minor or major significance and to judge which operations would be most effective. Some actions will have an “explosive” effect: when certain channels are opened, large and important areas of the whole territory become flooded or imbued simultaneously. We may add that, since the area is vast but time is short, during each visit the visitor can only inspect and improve limited parts of the territory. This picture illustrates the holistic way in which epistemic authorities operate when exercising epistemic empathy. The lakes, we may suppose, correspond to significant beliefs, important commitments, central acceptances, or pivotal epistemic goals of their interlocutors; ponds correspond to less significant subsystems, and the channels and ditches illustrate support relations. An authority may initially gain insight only into limited parts of an interlocutor’s system, but is in a position to extrapolate where the most effective work can be done in order to pursue the cognitive goal of filling and connecting cognitively “dry” areas.

To summarize what we have argued so far: epistemic empathy is a social-epistemic virtue, because it helps improve people’s epistemic situation (as opposed e.g. to their emotional situation). Moreover, it is a social epistemic virtue, because its primary or immediate target is other peoples’ epistemic states and episodes.Footnote 35

Epistemic empathy is a social epistemic virtue because it constitutes an epistemic excellence. Corresponding to different traditions in virtue epistemology, this may be spelled out in different ways. According to virtue reliabilists such as Ernest Sosa or John Greco, epistemic virtues are cognitive faculties or abilities the exercise of which reliably promotes epistemic ends. Traditional examples are vision, memory, introspection, rational intuition, etc. Whereas such faculties have individualistic epistemic targets, however, epistemic empathy is essentially a social virtue, since it is concerned with other people’s intellectual condition. As for responsibilist traditions in virtue epistemology, as influentially developed, e.g., by Linda Zagzebski, conceive of epistemic virtues not only as reliable faculties, but also as intellectual character traits that agents can deliberately influence. Whereas reliabilist virtues can operate sub-personally, character traits are so-called personal excellences. However, as Jaeson Baehr (VE) has argued, both kinds of virtues contribute to our overall intellectual wellbeing. If this is so, the contrast between the two traditions is less dramatic than meets the eye. What matters for present purposes is that epistemic empathy can also combine both sorts of virtue. It is a competence which, if operating under the right conditions, reliably furthers the corresponding epistemic goals (in some standard reliabilist sense); but it is also a personal excellence that agents can deliberately train and cultivate in their attempts to form their intellectual character.

6 Conclusion: authorities as sources of understanding

We have explored some social-epistemic dimensions of the conceptual connections between reflective equilibrium and objectual understanding. We started with Catherine Elgin’s proposal that an agent’s achieving or approaching reflective equilibrium in her noetic profile provides her with understanding and argued that epistemic authorities often help agents make this sort of progress and thus often serve as sources of understanding. This thesis departs from mainstream accounts of epistemic authority, on which epistemic authorities transmit (individual) beliefs and knowledge to their interlocutors. After discussing various ways in which authorities help interlocutors to achieve reflective equilibrium, thereby acting as testimonial sources of understanding, we have looked more closely at the competences involved in such processes. An important yet hitherto underexplored virtue in this context is epistemic empathy. We have considered some core features of this notion, distinguished various of it’s forms and aspects, and illustrated it with examples.

Yet, it may be asked, exactly how should we understand talk about serving as a testimonial “source” of understanding? Is the kind of noetic guidance epistemic authorities provide akin to the ways in which knowledge and belief can be transmitted via testimony?Footnote 36, Footnote 37

We have expressed sympathy for non-reductionism about understanding (the view that understanding does not reduce to true beliefs or knowledge) and have followed Elgin and others in construing the epistemic items involved in understanding as not only beliefs, but also acceptances, epistemic commitments, epistemic goals, standards of inquiry, and so forth. In this framework, there are generally two ways in which we might construe an authority’s help to the layperson in achieving reflective equilibrium, and corresponding to these are two ways in which an authority can serve as a source of understanding. The items she feeds into her interlocutor’s network may or may not correspond to her own respective attitude. Suppose they don’t. Then, even though the authority will help the interlocutor improve the equilibrium of his network, we should not characterize the situation as one of transmitting understanding. As Greco (2016, p. 483) remarks concerning knowledge transmission, “one cannot transmit what one does not have”. Note that situations of this type may include cases in which, although the authority does not endorse what she offers the agent, the latter’s network will be objectively improved. An example inspired by Elgin: Suppose that a young child tells you that humans evolved from hairy creatures who otherwise look a bit like humans but can climb and play games in trees. In an animal atlas, the child then shows you pictures of chimpanzees, gorillas, and orang-utans. “You mean that humans evolved from great apes?” “Yes, great apes!”, the child replies. “Very well, something like that!”, you respond and leave it at that. The conversation has arguably improved the child’s understanding of the topic since it has learned a general name for what is an important categorization in this context. But you neither believe nor accept nor are implicitly committed to the claim that humans evolved from great apes.Footnote 38

In other cases, by contrast, the epistemic authority does provide the interlocutor, via testimony, with parts of the authority’s system, and in such cases talk of “transmitting understanding” seems appropriate. For example, consider a professor of evolutionary theory who explains to her 14-year-old child in some detail (and with epistemic empathy) how homo sapiens and anthropoid apes such as gibbons, gorillas, orang-utans, bonobos, and chimpanzees evolved from a common ancestor. Anticipating, in acts of epistemic empathy, where her interlocutor might feel puzzled, ask for evidence, and so forth, she distinguishes the claims that merit fairly firm beliefs from those meriting weaker forms of credence, different again from those which, given the current state of scientific inquiry, deserve no more than acceptance. The biology professor enjoys (a high degree of) understanding of the topic, and by handing over central parts of it, transmits understanding to her child. All this is not to say, of course, that when such “understanding transmission” takes place, then no epistemic work will be left to the agent. However, our arguments show that a considerable amount of epistemic responsibility for whether or not, or to what extent, agents achieve understanding should be attributed to the epistemic authorities they consult.

Both the distinction between knowledge and understanding and the idea that pursuing epistemic goals is an essentially social enterprise have a long and venerable tradition. Indeed, they appear to be as old as philosophy itself. “Knowledge of many things teacheth not understanding”, Heraclitus declared.Footnote 39 He also argued that the logos is common and took issue with those who live “as if they had a wisdom of their own”.Footnote 40 It looks as if Heraclitus was right.Footnote 41