Many epistemologists claim that epistemic norms differ from other domain-specific norms like the norms of etiquette or rules of a game. Epistemic norms are thought to exhibit a substantial normativity in a way that such conventional norms do not. Yet they also want to isolate the epistemic realm from other normative domains. I contend one cannot have both (a) epistemic purity and (b) substantial normativity. While I will spend some time clarifying how I am understanding these two characteristics, it is important at the outset to recognize they are distinct and can come apart. Purity is essentially about grounding. If epistemic norms carry an authoritative weight because they are grounded in moral norms, then they are not pure. Substantial normativity is essentially about weight. If epistemic norms are like sartorial or gustatory norms, they might well be pure, but not substantial.
It is common when thinking about which norms govern belief—truth, knowledge, justification, or something else—to appeal to constitutive norms, namely to norms we uncover when thinking about the nature of belief, what a belief is. An analogy frequently used to help explicate what is meant by such constitutive norms are the rules of chess. Constitutive norms, minimally, provide correctness conditions for the particular practice for which they are norms. Just as the rules of chess will state when a move is correct or incorrect, the thought is that, likewise, the constitutive norms of belief will state when a belief is correct. It has been observed, however, that the kinds of norms generated in, for example, understanding the rules of the game, are much less substantial than many epistemologists think of as the norms applied to belief (often called epistemic norms). This point was made long ago by Judith Jarvis Thomson:
the rules of a game impose no obligation of any kind on the players … Suppose you are playing chess, and it is your turn to move. You then learn that if you don’t move your bishop horizontally, hundreds will die! Are you all the same under an obligation not do so? Must you, ought you not do so? That idea is just silly. (Reference Thomson2008, p. 90)
So if epistemic norms are going to generate obligations, or any substantial normative reasons, they must differ from the norms of chess. As we will see, some argue that constitutive epistemic norms differ importantly from other constitutive norms and so are substantial while others abandon the constitutivist project. An attraction of constitutivism is that it offers the more clear way of maintaining purity.
I (Reference Schleifer McCormick2015) raised this challenge over a decade ago, saying the “standard evidentialist picture of doxastic norms is flawed in that it assumes one can make sense of a distinctly epistemic realm that has normative force”(48) and have continued to raise in the context of discussing epistemic duties, blame, and reasons (Reference Schleifer McCormick, Carter and Bondy2019, Reference Schleifer McCormick, McCain and Stapleford2020, Reference Schleifer McCormick2021). In recent years, a number of theorists have attempted to meet this challenge, namely to show that an isolated epistemic domain generates such substantial norms. Sebastian Schmidt (Reference Schmidt2024) presents the challenge as a dilemma to the theorists he terms “normativists,” namely those who wish “to preserve a purely epistemic kind of normativity” that is also “significant” where significance is understood such that a subject who “at least normally or in paradigm cases of an epistemic norm violation is blameworthy or criticizable for not complying with the epistemic norm.” (Reference Schmidt2024, p. 4) He claims that “any normativist will, it seems, end up in either of two horns: (i) Epistemic norms are purely epistemic, but they fail to be significant. (ii) Epistemic norms are significant, but they fail to be purely epistemic.” (Reference Schmidt2024, pp. 8–9) He contends that this challenge can be met; I will argue that his proposed solution, as well as two others on offer, do not succeed. The structure of the paper is as follows: I will (1) make the challenge clear, which requires saying a little more about substantial normativity and purity, then (2) lay out three recent attempts to meet this challenge (Goldberg, Reference Goldberg2020; Horst, Reference Horst2022; Schmidt, Reference Schmidt2024), and argue that they are not successful. I will end by (3) attempting to clarify what is at stake in the debate; what are the concerns if purity is lost?
1. The Challenge
1.a Substantial normativity
There are various ways to characterize the kind of “substantial” normativity that many theorists claim epistemic norms exhibit. Côté-Bouchard (Reference Côté-Bouchard2016) uses the language of normative “force” and “authority” to distinguish a kind of substantial normativity from the thinner norms found in practices like etiquette and games. Mona Simion (Reference Simion2024) refers to the “normative oomph” of epistemic norms: “epistemic norms are, in a significant way, ‘heavier’ than mere conventions)”(Reference Schmidt2024, p. 97). Sandy Goldberg (Reference Goldberg2020) distinguishes the “normative” from the “evaluative” and explains “normative” as follows:
To be normative in the sense intended here, a theory must be such that, as applied to particular cases, it yields categorical verdicts about what is permitted or required. An epistemological theory is normative, then, just in case it yields categorical verdicts about which doxastic states a subject is permitted or required to have. (364)
Epistemic norms are norms that any rational agent must follow, not just one who happens to have an interest in knowledge. Ultimately, I think their categorical nature is what unifies these ideas that epistemic norms are more substantial than, for example, the norms of etiquette. In distinguishing epistemic reasons from “chess reasons,” David Horst says the following “competence in chess is a matter of being merely conditionally responsive to considerations relevant to the correctness of one’s chess-moves… what we expect from a competent believer is that she is, not just conditionally, but unconditionally responsive to considerations relevant to the correctness of her beliefs.” (Horst, Reference Horst2022, pp. 245–246). In distinguishing the normative significance of epistemic evaluation from, for example, evaluations of chess moves, Kate Nolfi (Reference Nolfiforthcoming) says epistemic norms “seem to have a kind of normative authority for us,” that they have “normative significance.”
Such language has a distinctly Kantian flavor. In the Groundwork of the Metaphysics of Morals, Kant (1785) says that hypothetical imperatives are “a means to achieving something else which one desires, “whereas categorical imperatives present “an action as of itself objectively necessary, without regard to any other end”(§39). The normative force of categorical norms is characteristically not contingent upon one having a certain goal or end. A categorical requirement is in force even when one would rather not comply. So epistemic norms are supposed to apply to everyone regardless of their particular goals or desires; one cannot escape their force or opt out of them, except by possibly ceasing to be a rational agent or one who forms and maintains beliefs. Even if I would rather not believe the truth about, for example, an unfaithful spouse or I would rather not know how a movie ends, these desires do not allow me to escape the force of epistemic norms.Footnote 1
For Kant, and others, moral norms have this categorical force. While the idea of morality applying categorically is not universally accepted, it does seem that if one recognizes and accepts a moral norm, such as “keep your promises” that the norm does not cease to apply simply because one does not feel like keeping the promise. Again, there is a striking difference here from what it means to recognize and accept a rule of chess. If one starts a game and then gets bored and decides to use the chess pieces to reconstruct a historic battle and so opts out of playing by the rules just to have a little fun, this can (assuming your opponent agrees) be perfectly fine. Putting categoricity aside for a moment, as Thomson noted, the idea of obligation or duty seems “silly” when it comes to abiding by the rules of chess. The idea that moral norms are more substantial than chess rules is something that can be taken as a datum, and the question is if epistemic norms are more like moral norms or “chess norms.”Footnote 2 As I have argued elsewhere, and will discuss further in Section 4, I think they are more like moral norms but this is because they are ultimately grounded in broadly practical considerations; they inherit their substance from this source. But if this is the case, then it seems they are not purely epistemic. Whether this is so may depend on what is meant by a “purely epistemic kind of normativity.” I will now turn to clarifying the idea of purity.
1.b Purity
One way of understanding purity concerns the content of epistemic norms. A purist of this kind would be contrasted with, for example, someone who endorses the idea of doxastic wronging. For such a purist, the content of the norm can only reference truth-related considerations, such as those related to evidence. “Do not believe without sufficient evidence” only has epistemic content while, “Do not hold a belief that is disrespectful to another” is one with non-epistemic content and so violates this kind of purity. It is not even clear that someone who endorses pragmatic or moral encroachment needs to deny purity of this kind. Even if ultimately what counts as sufficient evidence takes into account some practical or moral features of the context, the content of an evidential norm such as “only believe on sufficient evidence” does not thereby lose its purity.
Emily Tilton (Reference Tilton2025) has discussed the many different ways that epistemic purism and “impurism” are understood in the literature and argues that a certain kind of purism can still be sensitive to social and contextual factors. Georgi Gardiner (Reference Gardiner and McCain2018) makes a similar point in her arguments against anti-evidentialist views. Some (such as Rima Basu, Reference Basu2019) claim that beliefs can be morally wrong despite being epistemically impeccable but Gardiner argues one can always discover a cognitive flaw that leads to such beliefs. For example, in thinking about cases where one holds a racist or sexist belief which seems to be based on (often statistical) evidence Gardiner reminds us of how poorly people are generally at reasoning with statistics, or how in many of these cases, there is important evidence that is being ignored, or certain considerations are taken as evidence for something they do not support (such as when sexist people take the fact that women are physically weaker on average to show women are inferior to men). After an extensive discussion of the ways in which social and contextual factors can poorly affect one’s reasoning Gardiner concludes:
Advocates of moral encroachment aim to describe a person whose beliefs are epistemically impeccable—well supported by the evidence and conscientiously considered—yet morally wrong because racist or sexist. My contention is that no such belief can exist. If a belief is morally wrong then there is some corresponding prior epistemic error. The belief is not well supported by the evidence and/or it is not interpreted through a morally appropriate understanding, and that understanding is not epistemically well supported. (Reference Gardiner and McCain2018, p. 192)Footnote 3
Sometimes purity is taken to be an attribute of knowledge. On this view, knowledge is something untainted by any social context, completely objective, and neutral. As Tilton points out, even very traditional epistemologists can and do acknowledge the causal influence of social factors:
[E]ven the most extreme of purist views has more room to be socially engaged than it previously seemed. The purist can only take social, moral, and political factors to be epistemically irrelevant insofar as those factors make no difference to the evidence that one has or one’s ability and willingness to reason in light of it. Unsurprisingly, though, these factors often make quite dramatic differences to the evidence one has or one’s willingness to reason in light of it. In many contexts, it will make no sense for purists to disregard these factors, though they will explain the difference these factors make in terms of how they affect the quality of evidence or reasoning. (Reference Tilton2025, p. 4)
This is just what we saw in Gardiner’s discussion of supposedly morally problematic but epistemically impeccable beliefs.
But the kind of purity that is sought after by Schmidt’s “normativist” is not primarily concerned with the purity of the content of the norms, but with their grounding. Tilton identifies five purist theses that can be separated. The one which closely matches what Schmidt, and others who seek to isolate the epistemic from other normative domains, are seeking is what she terms “The Delineation Thesis” which she describes as follows:
We can sharply distinguish between theoretical rationality and practical rationality and, relatedly, between epistemic factors and non-epistemic factors. Theoretical rationality has to do with truth and knowledge, and so is sensitive only to epistemic factors. Practical rationality concerns how agents ought to act, and so is sensitive to both epistemic and non-epistemic factors. (Reference Tilton2025, p. 4)
What is central to this kind of purity is that the force of epistemic norms does not rely on the force of moral or practical norms. Simion (Reference Simion2024) aims for such purity and worries that many theorists who endorse the idea that we have genuine epistemic obligations fail to keep epistemic norms insulated from other norms: “just because a norm has epistemic content, it does not follow that it is an epistemic norm.” (Reference Schmidt2024, p. 76) Her criticisms of Sandford Goldberg’s “should have known” view is that it ties epistemic norms to social expectations. His view “runs into trouble due to its underlying strong normative co-variance claim for the epistemic and the practical.” (Reference Schmidt2024, p. 47) Similarly, one core problem she focuses on for virtue responsibilist theorists is that of “normative ambiguation” where it is unclear that the virtues being appealed to are “genuinely epistemic character traits.” (Reference Schmidt2024, p. 689).
In a recent paper defending an instrumentalist view of epistemic normativity, Hannon and Woodard (Reference Hannon and Woodard2026) describe many non-instrumentalists as viewing it as “a brutely epistemic normative truth that one ought to believe in accordance with the evidence. The source of normativity, according to many non-instrumentalists, is some kind of sui generis normativity.” (Reference Hannon and Woodard2026, p. 2) That there was nothing more to say on the matter was the default view until very recently.Footnote 4 But this view cannot explain how or why epistemic norms have the status that many take them to have, namely that they apply categorically, are universal and inescapable. Indeed, a demand for such explanation would be considered out of place by those endorsing such a view. Of course, this is not a problem if one accepts that this sui generis normativity is on par with sui generis normativity found in other normative domains such as etiquette. I take it that, despite their vast differences on matters of first-order epistemology, this is both Ernest Sosa and Richard Feldman’s views when it comes to epistemic normativity.
Sosa allows that it can be rational, all things considered, to believe for practical reasons. Still he argues that “a belief can be epistemically irrational though rational all things considered” (Reference Sosa2010, p. 34), and epistemology is interested in epistemic rationality. He suggests that one can identify numerous “critical domains.” Each domain has fundamental values, and we can appraise or assess (hence criticize) the derivative value of other items in the domain in terms of how well they promote the domain’s fundamental values. The fundamental values within a given domain therefore serve as the goal around which the critical domain is structured. Of course, in a certain sense, one can carve out a particular, specified normative domain in all kinds of areas where the values and norms are specific to that area. Sartorial norms can tell one how one ought to dress, gustatory norms can say that one crème brulée is better than another. Or any game one plays will dictate both whether a move is acceptable, as well as whether it was a good one. But epistemic value and norms, many contend, are supposed to differ from such “conventional” norms. They are norms that any rational agent should follow, not just one who happens to have an interest in knowledge.
Feldman (Reference Feldman2000) questions whether it makes any sense to ask the question what one ought to believe all things considered: In his view, it is important that the epistemic occupies a distinct normative realm; that the category of “epistemic value” is independent of other value. One of Feldman’s main conclusions is that the epistemic “ought” is entirely distinct from the moral or practical. He contends that for each “ought” there is an associated value and we “ought, in the relevant sense, do the thing that maximizes, or perhaps something that does well enough in achieving that kind of value.” (694) So we morally ought to do what produces enough moral value, prudentially ought to do what produces enough prudential value, and epistemically ought to do what produces enough epistemic value. But if these “oughts” conflict, there is no way to adjudicate between them, no meaningful question about what I ought to do or believe all things considered. He says: “We’ve disambiguated ‘ought’ and we can’t put the various senses back together again.” (694).
Some of Sosa’s more recent work has endeavored to distinguish between intellectual ethics and epistemology so as to make sense of what many think should be a stronger kind of normativity in the way we manage our doxastic lives, and Feldman tries to account for the seeming universality of epistemic norms, by arguing that the epistemic ought is a “role ought” the norms are dictated by the particular role one plays. For example, given one’s role as teacher, there are certain ways one ought to behave that would not apply to one who is not playing that role. But all of us play the “believer role.” Feldman realizes this:
[F]orming beliefs is something people do. That is, we form beliefs in response to our experiences in the world. Anyone engaged in this activity ought to do it right…it is plausible to say that the role of a believer is not one that we have any real choice about taking on. It is our plight to be believers. We ought to do it right. (Reference Feldman2000, p. 676)
But ultimately, if one thinks, as Quine (Reference Quine and Van Orman Quine1969, pp. 82–83) stated it, that epistemology is a branch of empirical psychology, primarily aiming to understand how human subjects come to have such a “torrential output” of knowledge from the “meager input” they are given, having nothing to do with questions of value or obligation, then this Sosa/Feldman approach may give one all one wants. But if one thinks that the normativity found in epistemology should speak to our fundamental human predicament then it seems more substantial norms are needed.
Before turning to the attempts to meet the challenge of how to maintain both the purity and substantial normativity of epistemic norms, I would like to point out a striking feature of the dialectic surrounding the source of epistemic normativity. In Hannon and Woodard’s discussion, it seems like one of the reasons for maintaining purity is that the force of epistemic norms will be diluted with a lack of purity, that one cannot have categoricity without purity. But much of the motivation for seeking a source of epistemic normativity beyond the epistemic is that doing so can explain why these norms have a force that differs from the norms of etiquette and the rules of chess. The accounts I will now turn to accept that this is something that needs explaining, that it is not just a brute fact, but argue that this explanation does not require appeals to any other normative domain.
2. Three Attempts to Meet the Challenge
I will now present three attempts to meet the challenge set out in the previous section and argue that none succeed. Horst’s view fails to account for the substantial normativity of epistemic norms, and for Goldberg and Schmidt, the purity of the epistemic realm is violated, and something beyond the epistemic domain is appealed to
2.a Horst
As mentioned above, Horst argues that we can vindicate the substantial normativity of epistemic norms by appeals to competence. Horst frames the challenge as a challenge to epistemic constitutivism (EC). He notes that other “standard governed activities” such as playing chess, dancing tango or cooking risotto do not “give rise to reasons,” and what he means by “reasons” here are something like categorical reason. And so “proponents of EC owe us an explanation of what’s special about the activity of believing such that, unlike other human activities, it does give rise to reasons” (Reference Horst2022, p. 233) and that “the challenge [is] to identify a distinctive feature of believing which can explain why considerations that bear on the correctness of particular beliefs necessarily amount to reasons for these beliefs.” (243).
He argues that we can see the difference between believing and chess playing if we attend to the different dimensions of evaluation of performing well qua believing and performing well qua chess player. Borrowing from Sosa’s view of the evaluation of performances, Horst reminds us that we can separate three dimensions of evaluation: correctness, competence, and aptness. The crucial difference that Horst points to is that in the context of chess (and other conventional normative domains) one can recognize what the correctness conditions call for but choose to do something different for a variety of reasons. Such choices, Horst claims, have no bearing on the person’s competence. If I choose to move my chess pieces around in nonstandard ways to have fun in my pretend battle, this does not reveal that I am an incompetent chess player. But one cannot make such choices, Horst claims, in the epistemic realm without it impinging on your competence as a believer. To be a competent believer, according to Horst, just is to respond unconditionally to epistemic reasons: “it’s only because of the distinctive character of competent epistemic performance—that is, the fact that such performance is unconditionally, not just conditionally, responsive to its own correctness—that we can extract conclusions about what there is reason to believe from competent belief-formation.” (Reference Horst2022, p. 254) If these were not different, then epistemic reasons would be just like chess reasons, but—says Horst—we know that these reasons apply to all of us qua believers. So, according to Horst, any flouting of epistemic norms diminishes competence, while this is not the case in other normative domains like chess or etiquette.
Horst locates the difference between epistemic norms and merely conventional norms like those found in chess or tango dancing by appealing to judgments of competence. One can be a competent chess player but choose to make a bad move (or even an illegal one), but supposedly cannot be a competent believer while violating epistemic norms. It is certainly the case that if person has many false or unjustified beliefs, this would be good evidence that they lack doxastic competence. Yet, I assume this would be so in the other realms as well. So, we need to have some evidence to decree someone a competent chess player or tango dancer before we would be willing to accept that someone with such competence could flout the norms for other reasons. So imagine we have evidence that someone is generally competent at attending to the evidence, does not draw hasty conclusions, and generally manifests the virtue of an epistemically competent agent. What then would we say if we observed that, in a certain context, they formed what appears to be an unjustified belief of some kind?
It would depend on the kind of case, but we would likely find various excuses for them. Maybe they are lacking evidence that I assume they have, maybe they are really tired or stressed, and so were not as careful as they usually are. If their epistemic errors multiply, I might start to revise my view of their competence (as I would in other domains as well). But this seems somewhat beside Horst’s main point. The distinction he is pointing to is that, he claims, a competent believer cannot choose to flout epistemic norms in a way that it seems a competent chess player can. It can be helpful here to think about cases of self-deception. When the evidence that your spouse has been unfaithful, or that your child is a drug dealer, mounts high enough, we may call someone who continues to believe against the evidence “self- deceived.” There is a sense in which the self-deceived agent is flouting epistemic norms for reasons, practical reasons that relate to the stability of their relationships, or love of their son. Horst is denying it is possible, and here is joined by many others, that one can make the choice to violate epistemic norms for these other reasons, that one cannot believe against the evidence while recognizing one is doing so. We now see that Horst’s way of distinguishing between epistemic norms and others relies on the assumption that only epistemic reasons for believing are possible, not about the nature of epistemic competence. But Horst offers no argument for the view that says it is impossible to believe for “non-epistemic” reasons, and there have been many who have argued against this view.Footnote 5
Now here is not the place to adjudicate between so-called pragmatists who think believing for nonevidential reasons is possible and the evidentialists who deny this. My point is rather to show that Horst’s argument requires him to reject the idea that there can be practical reasons for belief. Interestingly, when he lays out the commitments of epistemic constitutivism, he says it does not “imply that there can’t be practical reasons for belief: that is, considerations that count in favor of a belief, not by bearing on its correctness, but by bearing on (say) the benefits of so believing” (235) and references some people who think there are such reasons. But if he allows that one can believe for such reasons, this undermines the argument he provided. Yet once the assumption is granted that one cannot believe for such reasons, then it is simple to see the difference between epistemic norms and more conventional ones. If it were impossible for me to fail to follow epistemic norms when I recognize them, this would point to an important difference. But this difference has more to do with the powers and limitations of agency then with the force or substance of the norms.
2.b Goldberg
Goldberg (Reference Goldberg2020) argues we can account for the categorical nature of epistemic norms and explain how they differ from “the standards of fashion, dance, effective criminality, landscape architecture, and so forth” by grounding them in normative expectations, “of what we owe one another as social creatures.” (366) We are entitled to expect others to conform to epistemic norms whenever we are entitled to expect them to cooperate. Golberg forms this principle “COOP” as a conditional: “If we are entitled to expect cooperation, then we are entitled to expect conformity to the evaluative standards of epistemology by those with whom we are cooperating.” (372) It turns out this entitlement covers an extremely wide range of human activities. Indeed, quoting Karen Jones, Goldberg says the contexts in which this expectation is operative applies to “most of the activities of our daily lives,” including, for example, every time we have a conversation.
To understand why this expectation is so expansive, Goldberg points out that it applies not only when we are entitled to expectations concerning an existing project that requires cooperation but “also in circumstances in which we are entitled to expect another’s cooperativity. By the latter I have in mind an expectation that we might have in advance of any particular cooperative action we have already agreed to initiate with them or in which we are already engaged with them.” (373) This applies to “any belief whose truth is potentially relevant to any potential cooperation…I do not know whether this is every belief whatsoever; but I do know that it is a good many of our beliefs.” (374) To fail to meet these standards is to risk doing harm to others, and the harm in question is often epistemic. To show how this is so Goldberg introduces the idea of “epistemic externalities.” He argues that whenever a subject fails to meet epistemic standards one puts others at risk. The risks involve potential practical harms “arising out of potential cooperative activity whose success potentially depends on true beliefs about the relevant matters” as well as epistemic ones “e.g. arising out of the testimony one would give on the matter in question.” (378) Goldberg spends some time discussing one of the primary negative epistemic externalities, namely “the testimonial spread of false or otherwise unwarranted beliefs.” (379) Given that we are social creatures who rely on each other, we owe it to each other to do what we can to avoid this spread, which means satisfying the evaluative standards of epistemology and so Goldberg concludes: “in a great range of cases, epistemically normative verdicts—epistemology-driven verdicts regarding what subjects are categorically permitted or required to believe—can be seen to reflect what we owe one another as social creatures whose practical and epistemic dependence on each other is both systematic and ineliminable.” (382).
And so, according to Golberg, we can account for the force of epistemic norms by considering our obligations to and expectations of one another: “we owe it to one another to adjust our doxastic attitudes in a manner that conforms to the evaluative standards of epistemology.” (365) Why do we have these expectations, and see those who disappoint them as revealing a flaw? To explain this, I contend, we must provide a kind of Cliffordian justification. We rely on one another’s beliefs; the way we form and maintain beliefs affects one another, and to do so poorly can have disastrous societal effects. As Clifford puts it:
Belief, that sacred faculty which prompts the decisions of our will, and knits into harmonious working all the compacted energies of our being, is ours not for ourselves, but for humanity…Every hard-worked wife of an artisan may transmit to her children beliefs which shall knit society together, or rend it in pieces. No simplicity of mind, no obscurity of station, can escape the universal duty of questioning all that we believe. (Reference Clifford and McCarthy1987, p. 24)
Once epistemic expectations are tied to social ones, the purity and exclusivity of the epistemic realm are called into question. Goldberg even says his view is Cliffordian so one may wonder if he is even trying to maintain purity. It seems he would like to if possible in that he distinguishes his view from Clifford by saying:
whereas Clifford held that the sort of normativity at issue reflected our ethical values, I will be neutral as between whether the basis of epistemic permissions/ requirements is exhausted by our ethical values or also include the value of (cooperative) rational agency itself (where this is conceived to be a value independent of our ethical values. (36)
Also, as we saw, in identifying the potential harms that can arise in violating epistemic norms, he separates the practical harms from the purely epistemic ones. But such separation is problematic. The primary epistemic harm is the danger of the spread of false or otherwise unwarranted beliefs. But why is such a spread worrisome? Presumably because of the potential disastrous practical costs. I sometimes think that our current epistemic landscape, littered with deep fakes and reposting of debunked conspiracy theories, is an instantiation of a Cliffordian nightmare.
Goldberg concludes his article by considering to what degree his arguments support the categorical nature of epistemic norms. This partly depends on whether there is a categorical requirement to be cooperative. And even if such a demand is not categorical “it is plausible to think that this demand does not rest on what one happens to want or value”(382) and so the norms are more substantial than those found in chess and tango dancing, and so forth As a final thought on how deeply substantial they might be Goldberg offers the “Kantians” a final speculation:
Perhaps we should think of our disposition to cooperate as reflecting our social nature, part of who we are as a species—in which case grounding epistemic normativity in cooperation is no more to render it hypothetical than grounding moral normativity in our rational nature rendered it hypothetical for Kant. (382)
2.c Schmidt
Schmidt (Reference Schmidt2024) appeals to our practices of reactive attitudes to show how epistemic normativity is substantial in a way that conventional norms are not. In forming the challenge and response to it, Schmidt accepts that if a violation of an epistemic norm where there is nothing practical at stake does not sometimes warrant a blame reaction, then this would show that the norms are not substantially normative. More explicitly, he accepts this argument, “The Argument from Doxastic Blameworthiness” which closely parallels an argument I (Schleifer McCormick Reference Schleifer McCormick2021) defended:
-
(1) Evidence provides us with epistemic reasons for belief only if we can be blameworthy
(or criticizable) for violating purely epistemic norms.
-
(2) We cannot be blameworthy (or criticizable) for violating purely epistemic norms.
-
(3) Thus, evidence does not provide us with epistemic reasons for belief.
I was not primarily concerned with the normativity of evidence but rather with the question of whether blame is ever appropriate simply because an epistemic norm is violated and nothing else was at stake. As Schmidt says in a footnote, “McCormick argues that all blame for belief is ultimately prudential or moral. She therefore rejects that there is a normative domain of the purely epistemic” (Reference Schmidt2024, p. 10), and by normative here, he means the substantial normativity we have been discussing.
Schmidt argues, however, that there is a distinct kind of epistemic blame which is part of our practice of holding each other answerable to epistemic norms. “If we allow for a broad concept of blame in terms of impaired relationships, then we might be able to make room for something like purely epistemic blameworthiness.” (Reference Schmidt2024, p. 18) Further, he claims that “[o]ur actual practice of holding each other answerable to purely epistemic norms presupposes their normative significance insofar as we express this significance in our reactive attitudes towards norm violators.” (19) When it comes to cases where one non-culpably violates epistemic norms, or does so on entirely trivial matters, where it might seem that blame is not appropriate, Schmidt says even such violations “might indicate a general flaw in the epistemic character of the person” (Reference Schmidt2024, p. 20) and that it can be appropriate to reduce one’s epistemic trust in such a person. And so, he claims, blame will often be appropriate, which allows him to reject the premise (2) of the argument that claims to show that pure epistemic normativity lacks the significance it is often thought to have.
To make his case that blame can be an appropriate response even for purely epistemic violations, Schmidt appeals to answerability views of responsibility. On such views, one is answerable for a state if it is rationally evaluable, where it is appropriate to ask you for reasons or justifications. One is answerable for one’s beliefs, and often one’s emotions, in a way one is not answerable for one’s headaches. But these kinds of views often divorce responsibility from blameworthiness, given that we are often answerable for “non-culpable” violations. Schmidt considers the case of a nonculpable denier of human-induced climate change, someone who perhaps grew up in a community of such deniers where consideration or even exposure of counterevidence never comes up. Schmidt says of such a person that “resenting them or being indignant does not seem to be appropriate.” (15) Yet, if they are answerable for such beliefs, that is if they are “active responses to their reason-giving environment” they can still “legitimately give rise to serious forms of criticism or blame.” Now, of course, a certain kind of criticism is apt, the kind that one finds in all standard-governed activities where one fails to meet the standard. What Schmidt is searching for is a kind of criticism that has more “oomph” than this, one that we can call “blame” but without resentment or indignation. I remain skeptical that such a watered-down version of “blame” can ground the substantial normativity that is sought in the epistemic domain.
Recall that Schmidt thinks that in all cases of epistemic norm violation (even on the most trivial matters) it is appropriate to negatively modify our relationships with such transgressors and reduce our epistemic trust in them, and in so doing, we blame them:
we might be able to make room for something like purely epistemic blameworthiness. In an initial attempt, we might state that if we are blameworthy morally as soon as our relationship to our moral community is impaired, then we are blameworthy epistemically as soon as our relationship to our epistemic community impaired. (Reference Schmidt2024,18)
What we need to better understand to see if Schmidt succeeds in maintaining purity is what he means by the epistemic community and epistemic trust?
Schmidt does not expand on the notion of an epistemic community, but it is very hard for me to see how one can disentangle the epistemic from the moral here. As Goldberg pointed out, we are naturally social creatures; we live in communities and form relationships with members of these communities. To designate some of these relationships epistemic (as Schmidt and others do) strikes me as an idle designation. Do I have some relationships that are moral but not epistemic; what would that mean? Schmidt says that when someone reveals epistemic vices such as “dogmatism, gullibility, or a tendency for wishful thinking” (19), it can be appropriate for me to blame them (epistemically) because it is appropriate for me to reduce my epistemic trust in them, which signals a negative change in our relationship. But of course, discovering someone is lacking such care in the judgments they form is a discovery that will affect our relationship tout court. To isolate a purely epistemic notion of trust would be to say something like now, I cannot view them as reliable in conveying true or justified beliefs. But once such “trust” is isolated from other moral expectations, it seems one cannot distinguish it from “mere reliance.” Schmidt embraces this distinction: “Negative relationship modifications in response to vices can plausibly count as blaming responses because they are only legitimate towards responsible beings…Neither computer nor children or animals can display vices that give rise to [these] negative reactions.” (17).
When I say I do not “trust” the dog will not bite me, this is simply to say the dog is not reliable. In distinguishing between trust and reliance, it is common to point out that trust requires making yourself vulnerable, where feelings of being let down and hurt can be appropriate. Given these observations, Richard Holton concludes that “trust is reliance from the participant stance.” (Reference Holton1994, p. 67) For example, if my colleague always prints out the handouts, I may rely on the fact that he will in a way that does not constitute trusting that he will. If I actually trust him to do it, this comes along with reactive attitudes that are not present in mere reliance. If I am merely relying on him to make the copies because this is the usual state of affairs, I would not be betrayed if he failed to do so. And if I find myself irritated for his not doing it, this is a flaw in me, not him.Footnote 6 It is because failures in cases of trust do constitute grounds for attitudes like resentment and anger that one can only trust people and not objects. If the blame is going to go beyond the disappointment I might have in my car or my dog, then something about moral expectations needs to come in. And thus purity is violated. Of course, one might have good evidence that one’s friend lacks objectivity on a certain issue and so does not treat them as a trustworthy source on the matter, and perhaps this matter is not of any moral importance. Is this not an example of lacking epistemic trust in them? I think in such a case, one treating one’s friend as an unreliable source about the matter, just as one would treat a broken clock or thermometer as an unreliable source about the time or weather. Of course, we might well say “don’t trust that clock” but to talk about blaming the clock makes no sense.Footnote 7
3. Why Purity?
I will close by saying a little more about what is at stake in this debate. What is the motivation for purity? What is the contrast between accounts of pure epistemic normativity and those that are not pure? Throughout this discussion, we have seen that those who seek to establish the substantial nature of purely epistemic norms are sometimes contrasted with instrumentalism and sometimes with pragmatism. In many of these discussions, it is pointed out that the way we think and talk about epistemic norms reveals that they have a kind of non-instrumental force. If I have evidence that supports the truth of a proposition, it seems I have reason to believe it regardless of my desires, inclinations, or motivations. And if you find yourself in identical epistemic circumstances, you have the same epistemic reasons that I do, even if our desires and interests vastly differ. I know that you have a reason to believe the earth revolves around the sun without knowing any of your specific aims or goals. As Nofli (forthcoming) argues, our ordinary practices reveal that we accept criticism when we do not live up to epistemic norms and expect correction from others to experience pressure for correction when we make a mistake. Further, “ordinary thought and talk seems to presume that evaluation or assessment with respect to epistemic norms, standards, and prescriptions has global or universal normative significance.”
But that we talk this way about some normative domain is not enough to show that its norms are indeed categorical. Kelly (Reference Kelly2003), for example, recognizes that we talk and act as if reasons tied to the general desire of self-preservation are categorical. I say, for example, that you have a reason not to consume a poisonous substance without considering whether you have the goal of living longer or not. Kelly argues that when it comes to beliefs, we cannot make similar assumptions: “there is simply no cognitive goal or goals, which is plausible to attribute to people generally…Individuals do not typically have this goal: believing the truth.” (Reference Kelly2003, p. 623).
I agree that if one tries to find a specific desire or goal like the goal of acquiring true beliefs, one will not find that in all agents. But what if the system of epistemic normativity that provides us with reasons to believe is ultimately dependent on its extreme practical value? In a footnote, Kelly considers a wider conception of instrumental rationality based on Williams’s (Reference Williams and Williams1981) discussion of internal and external reasons which counts a reason as instrumental if it will “advance not only goals which I actually hold but also goals which I might reach by a process of sound deliberation from my present ‘subjective motivational set.’” (2003, p. 633) While it seems clear that I can have a reason to believe a particular proposition even if I have no desire to believe (and perhaps even a desire not to believe) that particular proposition, it may well be that in my subjective motivational set, as well as every other agent, one can find a commitment to practices that contribute to human flourishing. I view our system of epistemic norms and reasons as one of those practices, and so these norms inherit their force from their practical value. What is lost if we accept this view? Before answering this question, I will say a little more about the view I favor and have previously defended. It shares much in common with Hannon and Woodard’s (Reference Hannon and Woodard2026) instrumentalist view though, for reasons I will try to make clear, I have never embraced that label.
A way of framing the debate we have been discussing is to ask “why should one follow epistemic norms?” If one is committed to their categorical nature, perhaps such a question appears misguided. But this cannot be right since Kant devotes a lot of time to answering that question in the moral realm. If one fails to follow moral norms, he argues, one’s autonomy is undermined, one ceases to operate as rational agent with dignity and becomes more like a “thing.” A natural place to start to answer the question about epistemic norms is that following them will help you get true beliefs and knowledge. Now what if someone says “I don’t care about true belief and knowledge” do they thereby escape the force of these norms? No, because even if one cannot introspect to see that they have this particular care or desire, just by caring at all about anything, including survival, reveals that they do. I take it this was Hilary Kornblith’s point in addressing something like Kelly’s concern:
I have argued that epistemic evaluation finds its natural ground in our desires in a way which makes truth something we should care about whatever else we may value. This provides us with a pragmatic account of the source of normativity, but an account which is universal and allows truth to play a central role. (Reference Kornblith1993, p. 373)
Kornblith expresses some dissatisfaction with his pragmatic characterization of epistemic value. If someone could give substance to an account that made the value of truth noninstrumental and the injunction to seek truth categorical, he “would not be hostile to such an account,” (Reference Kornblith1993, p. 374) but cannot see how it can be done. What we have been surveying are some attempts to meet this challenge.
Unlike Kornblith, I do not think anything is lost in an account that acknowledges that the value of epistemic norms and the reason to pursue them is grounded in the broadly practical domain. This is, for example, what we find in Goldberg’s discussion, and he even insists they remain categorical in an important sense. In considering this kind of approach Nolfi (Reference Nolfiforthcoming) worries that it will not be able to vindicate the apparently “non-hypothetical character of the normative significance of epistemic evaluation” and this is because any variant of instrumentalism will allow that the norms are contingent on their ability to satisfy certain desires or attain certain goals.
An obvious answer to why we should follow epistemic norms is that true beliefs and knowledge are valuable because they are extremely useful. One’s life may depend on knowing which drink is the poison. Less dramatically, our ability to meet friends for coffee, drive to the grocery store, and cook dinner, all require us to know a lot. More lofty knowledge has led to the eradication of disease, building of cars and planes, computers, the internet, all of which are extremely useful. As we have seen, a worry about the pragmatic explanation of epistemic value is that the value of knowledge would be relativized to what people want and value. It seems no sense could be given to knowledge being a universal value. I may value knowing something that you have no reason to care about at all. And if the value of knowledge is grounded in the care and desires of humans, then it seems that epistemic value will be as fleeting and fickle as those desires. But once we stop focusing on the way following these norms benefits individuals, and instead, recognize that they are social norms that need to be followed for society to function well, this worry loses its force. One may worry that shifting the focus from the individual to society does not do justice to the idea that an epistemic norm is violated when, for example, one has decisive evidence for a particular belief but fails to hold it, though holding it would do nothing to benefit either the individual or society. But the point of this shift is to recognize that these norms are best thought of as guiding belief-forming and maintenance practices that need to be inculcated regardless of whether the content of an individual belief might be of value. What is valuable is having individual cognitive systems guided by norms that lead to having true beliefs given that we collectively depend on and rely on each other in our navigating the world.
This is Clifford and Goldberg’s idea, and one recently defended by Hannon and Woodard (Reference Hannon and Woodard2026): “The functioning of nearly every human endeavor depends on having reasonably accurate representations of the world.” (5) And so according to their view “epistemic normativity is a cultural achievement. Our epistemic duties are rooted in human concerns and values.” (6) I said something similar when defending my view of epistemic normativity:
the value of believing what is true is dependent on the value of both individual and collective flourishing. Having true beliefs is good because true beliefs are likely to be life-enhancing. If a world existed where having true beliefs and knowledge were not required for flourishing then they would cease to have value. (Reference Schleifer McCormick2015, p. 45)
Hannon and Woodard note that “many find this idea unpalatable.” But why? Again, what is lost? Given that these norms are not contingent on one’s particular desires, and remain in force even if one would prefer not to comply, they are categorical in the sense articulated by Horst and Goldberg, among others: “An upshot is that we cannot escape our epistemic obligations even when it is in our own personal practical interest to believe irresponsibly. The normativity applies regardless of our desires or goals.” (Hannon & Woodard, Reference Hannon and Woodard2026, p. 7)
What such an account cannot provide is that these norms are reason-giving in a way that transcends the human world: “Epistemic normativity does not consist in obedience to eternal, ahistorical standards.” (Hannon & Woodard, Reference Hannon and Woodard2026, p. 6) Ultimately, I contend, it is the desire for this transcendence that motivates purity.Footnote 8 The concern is that instrumental accounts of epistemic value and norms fail to capture their import and significance, and that the dictates to believe in a way to promote them will lose their force. I think part of the concern stems from a common way of thinking about instrumental value. Michael Lynch (Reference Lynch2004) begins his book in which he argues for the noninstrumental value of truth by pointing to the dangers of a “pragmatist” view. He uses Stanley Fish as the representative of the view against which he is arguing, calling it “antitruth.” Lynch seems to equate the view “what makes a belief a good one is that it is useful” with the view that “truth has no value.” Perhaps Fish holds both of those views, but it is possible to separate them. For, as we have seen, one can think truth, and the epistemic norms that help achieve it matter enormously, while still thinking its value is, in some sense, instrumental. But to Lynch, and others, this picture fails to capture something deep and important about truth and knowledge. He says “most of us value truth as more than just a means to an end. We care about having true beliefs” and we care about them “as such.” (Reference Lynch2004, p. 2).
What Lynch says here points to a common way of thinking about instrumental value, and why I have rejected the label of instrumentalism for my view. On the views we have been considering that ground epistemic normativity in its broadly practical value, to say such norms are “instrumental” is misleading if we take that to designate a means-end relationship such that epistemic norms are “instruments” used for attaining the good of flourishing. But we also say of something or someone that it is “instrumental” when we are emphasizing its importance, the key part that it plays in achievement or success, e.g. “that defender was instrumental in protecting the 4–3 lead.” This is not to say that the defender was a means to the end of winning the game; I am not even sure that makes sense. But the value of the defender’s actions is certainly dependent on the value of winning the game. Similarly, epistemic value is dependent on nonepistemic value, and the relationship is asymmetrical. Epistemic value cannot stand on its own; it cannot occupy a distinct normative realm while retaining its force. The upshot of this discussion is that to maintain substantial normativity requires rejecting purity.
4. Conclusion
I will conclude by acknowledging the limits of my discussion. I have only surveyed three recent attempts to meet the challenge of showing how pure epistemic norms can capture the substantial normativity they have, which other normative domains lack. And so others may offer ways of meeting this challenge that escape the criticisms I have discussed here. But, I also hope to have shown that the motivation for this purity is misguided. We can capture the import, universality, and inescapability of epistemic norms by realizing they are grounded in practical value.
Professor Schleifer McCormick is the Lewis T Booker Professor at the University of Richmond. Her research focuses on the nature and norms of belief. She also thinks and writes about reasons, skepticism agency, hope, trust, imagination, and emotions.
Open access
