The epistemic significance of non-epistemic factors: an introduction

We constantly assess each other’s epistemic positions. We attempt to distinguish valuable from worthless information, reliable from unreliable informants, etc. Without established social practices of epistemic evaluations we could not navigate the flood of information we are exposed to every day in order to perform essential selections of valiable information. Yet the way we epistemically evaluate each other, ascribe or deny knowledge, who we deem knowledgeable or ignorant, and whom we refer to as an expert or a layman also crucially shape our epistemic milieu and the structure of our society. Epistemic asymmetry often results in and reflects social asymmetry; higher epistemic appraisal often increases social standing. Also, epistemic evaluations such as knowledge ascriptions are commonly performed against the background of certain epistemic and non-epistemic (e.g., practical) concerns and interests. Consequently, epistemic and non-epistemic factors interact in guiding our epistemic practice. To advance our understanding of how they do so is not only a worthwhile project from an epistemological point of view but can be expected to have repercussions on decision making, in debates within political and social theory as well as within ethics, and help us understand and evaluate how we act and even how to act. Moreover, it might shed light on the perennial question of how theoretical and practical rationality relate to one another.

A much-discussed question in recent debates on knowledge ascriptions is the question of whether—and if so, how—epistemic standards (standards of how much it takes to count as knowing or as a knower) are influenced by, and/or contextually vary with, non-epistemic factors such as stakes, interests, aims, etc., and whether this in turn affects the truth-conditions of knowledge ascriptions or only their assertibility (or sayability) conditions. This has been a main point of contention between contextualists, invariantists and relativists (of various brands) concerning knowledge acsriptions (cf., e.g., Baumann, 2016; Blome-Tillman, 2014; Cohen, 1987; DeRose, 1992, 2009; Fantl & McGrath, 2009; Kompa, 2002; Lewis, 1996; MacFarlane, 2014; Stanley, 2005; Stei, 2016; for an overview see Ichikawa, 2017; for a Pragmatic perspective, see Kvart, 2018a, 2018b, 2020 and forthcoming.).

Various contributions to the special issue address the question of how to best model the (alleged or real) “context- or interest-sensitivity” of knowledge ascriptions and argue for or against a particular position (Dinges; Kompa, Kvart, Lossau), for example by examining which aspect of the content conveyed by means of a knowledge ascription can be canceled (Lossau). Others discuss whether a particular position is committed to pragmatic encroachment or not (Prichard; Newton; Blome-Tillman; Goldberg; Lawlor, Kvart), for example, whether anti-luck epistemology is susceptible to pragmatic encroachment (Prichard). One contribution examines the arguments for moral encroachment, i.e. the claim that moral features that do not bear on the truth of p can affect whether the belief that (or credence of) p is (epistemically) rational (Fritz/Jackson). Others focus on key notions in the debate that seem to defy (or have so far successfully defied) systematic treatment such as the notion of relevant alternatives (Lawlor) and the notion of stakes (e.g., Baumann), provide long overdue accounts of those heavy-duty notions, or discuss the role of emotions as knowledge-conducive (Dietz).

Moreover, over the last three decades, epistemologists became more and more interested in the purpose our epistemic evaluations serve in general (Henderson & Greco, 2015), and the different functions knowledge ascriptions fulfill in particular. The latter are said to serve the function of flagging good informants (Craig, 1991), certifying information on which to base one’s theoretical or practical deliberation (Williamson, 2000, Hawthorne & Stanley, 2008), keeping epistemic gate (Henderson, 2009, 2011), terminating inquiry (Kusch, 2009; Rysiew, 2012), serving as a basis for attributions of praise and blame (Beebee, 2012), and of steering addressees toward, e.g., an action (Kvart, 2012, 2015a and 2020), to name just a few.

But then, the two debates seem to be intimately intertwined. Which functions knowledge ascriptions (are taken to) fulfill seems to depend on the ascribers’ interests, aims, etc. Recently, Henderson (2009, 2011), Lawlor (2013) and McGrath (2015) have appealed to certain social and practical functions of knowledge ascriptions to motivate specific kinds of contextualist approaches to knowledge. Some of the contributions to the Special Issue thus acknowledge a connection between the two debates and exploit it to defend one position or another (Henderson; McKenna/Hannon; Kompa).

Furthermore, by serving such various functions, knowledge ascriptions (are designed to) provide the addressee with epistemic as well as practical reasons, i.e. reasons for beliefs and reasons for actions (cf. Robitzsch, 2016, 2019). Providing reasons is one way in which epistemic evaluations (or what seem to be such) may steer people into action and belief states with related mechanisms (such as non-propositional Steering Thrusts) that might come into play too (Kvart, 2018a, 2020). And while knowledge ascriptions (and epistemic evaluations more generally) may provide agents with practical reasons, agents may, in turn, also have practical reasons to conduct inquiry in the first place. While this might lead us to support pragmatic encroachment (or in supposedly practical reasons to inquire further that might seem to reflect a rise in the threshold for justified belief and knowledge), in one of the contributions to the Special Issue it is argued that practical reasons to inquire (further) can be seen as being merely practical and in no way affecting the degree of justification required for justified belief in the context. (Goldberg).

More generally, the connection between epistemic and practical normativity is an upcoming topic in epistemology (cf., e.g., McHugh et al., 2018; Meylan, 2013; Peels, 2017; Star, 2018). Within certain quarters of philosophy of science, it has been acknowledged for a while now that good scientific practice is governed by epistemic as well as non-epistemic norms (cf., e.g., Kelly, 2003; Kitcher, 1993; Laudan, 1990; Siegel, 1990). Yet a fuller account of how these norms interact and how they are to be weighed against each other is still pending. Accordingly, one of the contributions is devoted to examining the way in which our epistemic concepts are sensitive to epistemic norms (and needs) and how epistemic norms in turn function as social norms, thus regulating epistemic practice (Henderson). Another contribution is concerned with explicating the manner in which knowledge ascriptions are sensitive to a standard of reasonableness and what reasonableness amounts to (Lawlor).

Relatedly, knowledge (as such) has been said to be intimately tied to action and assertion (cf., e.g., Brown, 2008; Hawthorne, 2004; Williamson, 2000). Knowledge is said to be necessary (or sufficient, or both) for action and assertion. For example, some hold that one ought to assert only what one knows or that one ought to act only upon what one knows. Yet, contextualist or prespectivalist positions more generally are thought to jeopardize (or at least question) the close link between knowledge and (norms of) assertion or action. Accordingly, something’s got to give, it seems. Two contributions tackle this issue and either defend perspectivalism by appealing to the function knowledge ascriptions fulfill (McKenna/Hannon) or taking it as a starting point for constructing an argument against impurism (Blome-Tillman).

Newton examines the issue of whether non-epistemic factors such as modal factors that pertain to characteristics of non-actual worlds and accordingly don’t affect the probability of p (in the actual world) make an epistemic difference, specifically to whether a true belief is also knowledge. Newton argues that prima facie, safety and sensitivity allow that non-epistemic changes in the context can make a difference to the ordering of possible worlds and thus to a true belief’s being knowledge. (The critical remarks in the discussion of Newton, Dietz, Pritchard and Dinges below as well as in the Appendix commit only Kvart.)¹

However, to commit to Newton's last thesis without the ‘prima facie’ caveat requires being more specific about the notion of probability employed. Using a mere probability distribution over a set of possible worlds in construing modality, unless chancy (which would require a common root), is, it seems, hard to understand non-formally and philosophically. (It might be worthwhile to contrast such a conception of safety,  formulated as in ‘not easily can a given true belief be false’, with Kvart’s (2006) conception of safety (as a necessary condition for Knowledge) in terms of chances, and more specifically, as a condition for Knowledge, in terms of requiring a very high conditional chance of p given the belief that p plus carefully selected factual features roughly ‘up to’ the time of belief that p.² As such, it reflects to a large extent a constraint on the belief that p being Highly Token Indicative of the fact that p.)

Dietz, in her paper, makes two main claims: That emotions can be conducive to knowledge, while responding to objections, and that emotions can enhance safety. Dietz, by and large, follows Williamson’s picture. Dietz claims that having an emotion like fear can be evidence of danger as the subject takes into account her relevant past performance. (From the perspective of Kvart’s High Token Indicativity account (2006, 2018b), it seems that this can be captured to a significant extent by saying that sensing a token fear-emotion may indeed token-indicate there  being a ‘sensed ‘ danger out there.) This issue seems to reflect notorious chicken-sexer cases, varying with how much pertinent record the subject has had, to what extents she is in a position to ‘take them into account ‘, and with what degree of awareness.

Dietz construes safety (following Williamson) in terms of ‘close’ worlds. She says that for the safety theorist, “perhaps the only kinds of cases that will count as ‘close’ are those cases which involve very small tweaks to the recent history or current environment.”³ In that, she follows the general outlook of the tradition (starting with Stalnaker’s (1968) ‘minimal changes’) in being relatively unspecific about what ‘small tweaks’ (or ‘minimal changes’, in the tradition) boil down to. But what ‘minimal changes’ or ‘small tweaks’ boil down to remains mysterious, and presumably context-dependent and interest-relative. When it comes to safety, resorting to notions related to minimal changes suffers from the lack of a needed strong constraint on them, as is the case in the counteractual literature (where they first appeared): For counterfactuals, very roughly, the corresponding picture is that they are true if there are minimal changes that restore (for a counterfactual p > q) a p-world state (out of the -p-actual-world, or a -p-world-history), albeit a modified one, with q ending up holding, but no such changes counting as ‘minimal’ can do the work for -q. For this to take off the ground, the requisite ‘minimal changes’ that will ‘incorporate -p’ must be a function of p but not of q. So vis-a-vis safety (in relation to knowledge), the notion of ‘close’ is relatively (in this sense and to such an extent) unconstrainted , and is usually taken to be context-dependent and interest-relative.⁴ (See below, regarding Pritchard, the claim that such a conception of modal ‘closeness ‘ seems to force stakes-dependence.)

Pritchard‘s anti-luck epistemology⁵ tackles the main issues head-on with a bold theory with consequential repercussions—a kind of theory many of us welcome. He anchors his anti-luck condition on knowledge (and in epistemology, more generally) in his modal conception—which employs the ‘close’ relation between worlds,⁶ and concludes that this knowledge account of his is not committed to Pragmatic Encroachment. This issue boils down to whether the ‘close’ relation is interest-relative (and thus stakes-dependent). Yet it doesn’t seem that one can ensure that a ‘minimal change’ (lurking back to Stalnaker’s theory of counterfactuals) not be interest-relative (once it’s context-dependent): Differences in interest are prone to confer differences in context, and thus in what counts as ‘minimal’ in a context⁷—see also the above comments on Dietz. (For Pritchard, it seems, the appeal is also to ‘relevantly similar’.⁸ But what is ‘relevant’ can also vary with interests.)

Thus, ‘minimal changes’ are very much context-dependent and stakes-dependent: What is minimal for one subject in one case need not be so in another for another subject, or even for the same subject but with, say, different stakes. Stakes thus very much enter the pot: A change that underpins considerable stakes for the subject in one context need not be minimal, whereas it can be minimal in a context in which it doesn’t underpin much of a change for the subject’s stakes there. So the repercussions to the subject in view of these stakes don’t matter much in the second context , thus allowing for that change to count as ‘minimal’, but it need not be minimal in a context where it does matter, or matters very much—where the change does raise/lower the stakes considerably. So Pritchard’s modal conception of luck, anchored in his above construal, would seem to commit him to truth-conditions of knowledge-ascriptions that are stakes-dependent, contrary to his claim that his account doesn’t commit him to Pragmatic Encroachment.

So the general lesson is that the very basic modal orientation that Pritchard commits to, as well as other accounts conceived in such terms, are left vulnerable to Pragmatic Encroachment. Thus, in particular, a Safety condition, formulated in such terms, is also subject to Pragmatic-Encroachment. And the more general lesson is that Lewis’ and Stalnaker’s modal conceptions, as construed by them, separately, commit them to contextualism for counterfactuals as well as for any other locution or construction conceived in terms of such semantics under these construals.⁹ In conclusion, if one wants to resist Pragmatic Encroachment, then it looks like it would require a very fundamental change in orientation, so that the modal notions appealed to not be cashed out in terms of the classical modal construal of Stalnaker and Lewis, but rather in terms of a chance structure (such as the chance structures I offered for counterfactuals, causation, and knowledge, and which, I have noted elsewhere,¹⁰ are pretty much objective, and accordingly pretty much context/interest-independent).

Dinges considers a strategy for dealing with stakes-effects on knowledge ascriptions which he calls Doxasticism. On this approach, knowledge ascriptions depend on stakes since so do belief ascriptions. But Dinges lumps together salience and stakes as instigating belief formation or its inhibition which may lead to retraction. Regarding inhibition, Dinges, it seems to me, needs to distinguish four very different types of retraction: First, there is a purely epistemic retraction, which is instigated just by a change in epistemic pressures (e.g., more evidence, or a mere new error possibility). Second, there is purely Practical Retraction, such as one that is instigated by a change in risk/expected-gain information (and may or may not be communicated). Third, there is Cognitive Retraction, largely due to loss of confidence. And fourth, there is Pragmatic retraction (in the sense of Pragmatics), where the retraction is communicated by conveying implicit (Pragmatic) reversed Steering Thrust content. The first three are characterized in terms of what instigates the retraction, whereas the latter—in terms of how it’s communicated. Any combination of such retractions types may be co-instantiated in a given retraction token (thereby giving rise to their superpositions in such mixed cases). These four different retractions exhibit very different types of cognitive phenomena and cognitive mechanisms, such as epistemic processes; Action-Directed decision-dynamics; Pragmatic (in the sense of Pragmatics) processing, such as in conveying Steering Thrusts, as in a Steering towards retraction; and purely cognitive processing mechanisms in a ‘general cognitive system’ as with temporary partial ‘resets’ or just ‘time off’ calls for the purpose of re-processing or ‘cognitive recovery’.

Pragmatic retraction (as I use the term, in the common Action-Directed case) is the use of Pragmatic (communicational) tools in order to steer (an audience) away from an action (usually canceling thereby a prior steering towards it—a Steering retraction). This phenomenon is distinct from Action-Directed cognitive retraction in processing that may well lead to a retraction of an action-decision and which may or may not be communicated in Pragmatic terms (or not communicated or even signaled at all). (The two are often co-present and interact.) Belief retractions are usually just epistemic adjustments involving withdrawals from a prior belief. But a retraction signaled by negating a knowledge-ascription (as in: Now I don’t know that p) can be either purely epistemic, or purely Action-Directed -- steering away from an action; or indicate a temporary confusion; or be a superposition of any two or more of those. It’s important to realize that a case of such a superposition is to be analyzed as such—in terms of the component retractions.

I have  recognized Pragmatic Encroachment regarding beliefs, though in my view it comes in a lesser degree than as it appears to be in knowledge ascriptions. But, I submit, it’s a mistake to consider Pragmatic Encroachment about belief as underpinning (and securing) Pragmatic Encroachment about knowledge¹¹: This mistake is not due to misgivings about the seemingly obvious point that knowledge requires belief as a constituent. Knowledge that p can be had even with less than a belief that p, such as even with a mere (sufficient) warranted disposition to believe that p.¹² But obviously knowledge requires warranted beliefs (or the like), not mere beliefs; and importantly, warranted beliefs, unlike mere beliefs, are not subject to Pragmatic Encroachment: Whereas one can indeed modify one’s degree of belief that p purely due to (say) a new risk, being warranted to believe is not subject to such a change—such a ‘diminution ‘—due to such a new risk (‘warranted’ here is used epistemically).¹³ That is, the warranted status to so believe isn’t amenable to ‘retraction’: Being warranted to belief is invariant under new risks.¹⁴ That is, being warranted to believe (that p) doesn’t ‘get retracted’ (doesn’t vanish or diminish) just due to the appearance of such a new risk. So even though belief is subject to Pragmatic Encroachment, being warranted to believe is not, and consequently knowing needn’t be (and, as I have argued, isn‘t).¹⁵ Therefore, it seems, Doxasticism, even though allowing for Pragmatic Encroachment about belief, needn’t hold—and doesn't—for knowledge-ascriptions: Pragmatic Encroachment about belief doesn't impinge on the Pragmatic Encroachment debate regarding knowledge.¹⁶

In summary, this Special Issue is concerned with the different ways in which epistemic and non-epistemic norms and interests might interact in governing our epistemic practice, and specifically knowledge ascriptions. It aims at a better understanding of (i) the functions knowledge ascriptions fulfill and the purpose epistemic evaluations serve more generally; and more specifically, (ii) how they fit into communication—which communicational roles they fulfill; (iii) the way they are sensitive to epistemic and non-epistemic interests, concerns, reasons, or aims and how these are to be best modeled (in contextualist, or invariantist, or Pragmatic terms, etc.); (iv) how knowledge relates to (norms of) action or assertion, (v) which norms underpin our epistemic practice—its individuation, evaluation and reports; and (vi) what role they play in, and how they fit into, our Practical Reasoning, covering different types of, e.g., heuristics.

The contributions to the Special Issue address these topics and provide a broad and multifaceted perspective on the phenomenon of knowledge ascriptions; the epistemic significance of non-epistemic interests, reasons, or norms for our practice of epistemic evaluation; how we report such knowledge ascriptions; and what functions such reporting serve. For although there are many interesting connections between these topics, there has as of yet not yet been that much discussion between those working on the different topics. This special issue is meant to foster and invite cross-topic discussion.

The contributions in the Special Issue connect debates within epistemology and philosophy of language but also touch upon questions within ethics, action and decision theory and philosophy of science. How to answer the question of how epistemic and non-epistemic factors interrelate in shaping our epistemic practice in general, and how that is reflected in the way we ascribe knowledge in particular and in the various functions such reporting serves, has important consequences for many other philosophical disciplines and for our epistemic life in general.

Appendix

In conclusion, a few comments on my Pragmatic perspective and my meta-heuristic orientation, which might lay bare orientational biases reflected my role in the conference that gave rise to this special issue (and as an editor of this issue).

The Pragmatic Encroachment literature, in all its variants (e.g., epistemic contextualism, SSI, Shifty Epistemology) has brought to light hitherto unnoticed phenomena and riddles, which tax our traditional conception, present intriguing challenges, and suggest alternative conceptions. But for all the very valuable contributions that emanated regarding knowledge, language, practical reasoning and more, I consider (and have considered) the main move as unwarranted, despite the theoretical depth it has managed to generate.

I approach the issues of Pragmatic Encroachment on the one hand and the interplay between epistemic/non-epistemic Rationality (as, e.g., regarding inferences to ‘act on p’) on the other by emphasizing that Pragmatics underpins many of the issues in this area, and that Pragmatic contents (i.e., implicit contents) are in play, especially once non-epistemic pressures weigh in. (I use Pragmatic, with capital ‘P’, in contrast to (primarily) Semantic— not in the sense of ‘practical’.) To construe communication with conveyed Pragmatic contents only literally is to ignore, and thereby make a mockery of, the presence and force of Pragmatics. Yet, I acknowledge that a new adequate Pragmatics needs to be resorted to in view of the limited applicability and usefulness of traditional Pragmatics, especially (in our type of cases) in (deliberative) Action-Directed contexts. I appeal to a non-Gricean Pragmatics that I proposed whose main implicit content is Steering-Thrust, backed up by a Formal Pragmatics, where the implicit contents conveyed are not propositional and thus without truth values.¹⁷ In particular, the locution of ‘I don’t know that p’ may be used in Action-Directed contexts to convey Steering Thrust against a certain action rather than to convey a semantic content.  Likewise, the locution 'I know that p' may be used in such contexts to convey Steering Thrust for a certain action (even though not instead of conveying something close to the literal sense). Accordingly, in such cases, the seeming inconsistency (strictly speaking) between the use of these locutions (and their kin) in different contexts of the same type is off the table, and a main riddle-type dissipates. Yet there is still room for the seeming tension between two ‘incongruent’ Steering-Thrusts, invoking incompatible polarities regarding a certain action (for or against performing it). However, this tension may be innocuous since there may be new pertinent information about risk in the two different contexts which doesn’t figure, strictly speaking, in the literal (semantic) consistency issue (because not evidentially relevant) but does come into play vis-à-vis Action-Directed Rationality (since it may well be relevant to action-decisions). Consequently intriguing phenomena in bank-type puzzles (where the question may arise as to whether an actor is ‘incongruent’) can occur when a new relevant non-epistemic pressure arises (whether ethical, instrumental, or any other normative pressures): This is a theoretically built-in possibility, and the appropriate pressure is channeled towards the issue of whether the right decision has been made (rather than whether semantic, literal contents are consistent), and accordingly, which heuristics have in fact been in play, or should have been, as well as how such a decision is Pragmatically communicated. To insist on construing a case with bona fide Pragmatic contents as a case with only literal contents is consequently a methodological mistake underpinning an insistence on a Semantic construal, characteristic of Pragmatic Encroachment treatments of its all stripes. By employing Pragmatics properly, a consistency puzzle, as such, doesn’t arise.

Further: The common appeal to Practical Inferences as properly invoking Action-Directed Rationality requires an awareness of the limits of the domain of applicability of an employed heuristic such as this ( and, more broadly, of heuristics in general). Simple Practical Inferences are applicable tour court when there is no (significant) pertinent risk, but are no longer applicable when there is a new considerable pertinent risk. In such a case, they must yield to more sophisticated heuristics (e.g., invoking one of the various types of 2 × 2 decision-matrices—differing, e.g., in precision, and, as such, falling short of e.g., Expected Utility—rather than just a 2 × 1 matrix). The use of a heuristic in a case when it’s not applicable is an outright meta-Rationality mistake: With the onset of a new considerable pertinent risk, the Simple Practical Inference is no longer applicable—is outside its domain of applicability, and its Action-Directed conclusion no longer carries an injunctive force. When the heuristic in use is rendered inapplicable due to a new pertinent risk, an Inferential Ascent is called for—a meta-rationality move to a more suitable heuristic (albeit more costly). Inferential Ascent appeals to a gradable stock of richer, more refined and more demanding heuristics (when available). A retraction of a previously steered-to action is thus called for (at least prima facie, or temporarily) when a previously appropriate simple Practical Inference is rendered inapplicable given new pertinent normative pressures. Such a retraction (in deliberative contexts) is Action-Directed—it’s a retraction of previously conveyed Steering Thrust (towards an action), which as such is not propositional since in such cases new pertinent pressures invoke non-literal Pragmatic implicit contents (which are Steering Thrusts). Action-Directed retraction can be (and often is) conveyed by Steering Thrusts (away from the previously Steered-to action, or towards an incompatible action) rather than explicitly.  It needn’t represent a change in an epistemic position (towards a certain proposition) nor constitute a new epistemic reason not to act (since it needn't indicate a change of a pertinent epistemic position)—the retraction of a knowledge-ascription in such a case merely steers the hearer not to perform that action: It needn't indicate a (new, factual) epistemic position, and accordingly not a reason rooted in a change of an epistemic position—a change of previously possessed factual information. A merely Semantic (literal) construal of such exchanges has thus led to pseudo-issues about how retraction provides factual reasons (for no longer steering towards that action).