1. Introduction
When we assert something, we represent ourselves as both knowing and being sure of what we say. This is why it is normally appropriate to challenge an assertion that p by asking questions like ‘What makes you sure that p?’ and ‘How do you know that p?’. This also explains why assertions with the form ‘p, but I don’t know that p’ and ‘p, but I am not sure that p’ sound infelicitous. By making such assertions, we represent ourselves as being sure and knowing that p, while simultaneously denying that we are in these states. This set of claims seems to be manifestly incoherent.
Some philosophers have argued that the best way to explain the previous data is to assume that the practice of assertion is governed by the following norms:Footnote 1
(KN) Assert that p only if you know that p
(SN) Assert that p only if you are sure that p
How do these two norms relate? According to a standard account, commonly attributed to Unger (Reference Unger1975) and Williamson (Reference Williamson2000), KN is more fundamental. If we understand being sure in an ordinary, not overly demanding sense (as when we say that we are sure of what our name is, or that Paris is in France), then it is plausible that knowing that p implies being sure that p. Once we accept this implication, SN can be straightforwardly derived from KN. If we shouldn’t assert what we don’t know, and we know something only if we are sure of it, then we shouldn’t assert something if we are not sure of it. Using technical terminology, we can say that SN is descriptively parasitic on KN.
Goodman and Holguin (Reference Goodman and Holguin2022) have recently argued for an alternative picture on which SN is more fundamental.Footnote 2 KN can be derived from SN and a further norm governing surety:
(SKN) Be sure that p only if you know that p Footnote 3
They argue that KN is normatively parasitic on the other two norms in the following way: if one should assert that p only if one is sure that p, and one should be sure that p only if one knows that p, then one should also assert that p only if one knows that p. By asserting a claim such as ‘p, but I don’t know that p’, a speaker would either be unsure of what she asserted (thus violating SN) or be sure of something she doesn’t know (thus violating SKN). Either way, they would violate a norm. This would explain why that assertion is improper.
Goodman and Holguin hold that their account is better than the orthodox one because, they argue, being sure is not a necessary condition for knowing. There are cases in which one can remember or see that p without being sure that p. Given that remembering and seeing that p imply knowing that p, in these cases the subject knows that p without being sure. Moreover, the two authors argue that in such cases it would be inappropriate to assert that p. This impropriety cannot be explained by the orthodox view.
In this paper, I defend the orthodox view from Goodman and Holguin’s challenges. More precisely, I provide objections to Goodman and Holguin’s account of how the knowledge and surety norms are related, and to their arguments against the orthodox view. In Section 1, I argue that their derivation of KN from SN and SKN is problematic. My objection is based on a general point: it is not possible for a norm to be normatively parasitic on others in the way described by Goodman and Holguin. No norm supposed to govern our everyday practices can be derived in this way. In Section 2, I argue that their arguments to the effect that one can know without being sure are not convincing. These cases are underdescribed. Once we fill in the relevant details, we can see that either they do not involve propositional seeing and remembering, or they also involve surety.
2. Against Normative Parasitism
Goodman and Holguin’s derivation of KN from SN and SKN relies on the idea that a norm can be normatively parasitic on others. This is how they describe this type of derivation:
[S]uppose that in addition to the norm “ϕ only if you are F” there is also a norm “Be F only if you are G”. Then there is a normatively parasitic norm “ϕ only if you are G”. (If one ϕs without being G, then one must either be in violation of the norm to ϕ only if one is F or in violation of the norm to be F only if one is G.) (2022, p. 635)
This method of deriving normative requirements has been regarded by many as highly questionable. For instance, Broome (Reference Broome2013, pp. 120–1) argues that from the fact that one is required to F and required to G if one Fs, it does not follow that one should G. Broome illustrates this point with the following example: Suppose that you have signed up for a marathon. It would be prudent for you to exercise vigorously every day. Moreover, if you exercise every day, you should also eat heartily. But from these two facts it does not follow that you should eat heartily. If you choose to be lazy and not exercise, ‘intuition suggests it is wrong to conclude that prudence requires you to eat heartily’ (2013, p. 120). In fact, prudence may require the opposite, as it would be worse to run a marathon without proper training and being overweight.
The above point is a familiar one to those acquainted with the literature on contrary-to-duty imperatives (Chisholm, Reference Chisholm1963). These are norms telling us what we should do if another norm has been violated. For example, we should keep our promises, and we should apologize when we fail to keep them. However, from these two norms it does not follow that if we have made a promise, we should never apologize. On the contrary, when a promise is not kept, apologizing is the right thing to do.
The same idea is implied in what economists and political philosophers call the Theory of the Second Best.Footnote 4 According to this theory, if achieving an ideal state requires a specific set of conditions, and one of these conditions is not met, the remaining conditions may also lose their value, or even become undesirable. As a consequence of that, given our limitations, the ideal course of action may not be the one that we should pursue. Consider an example from DiPaolo (Reference DiPaolo2019). Dr. Brown should administer her patient one of two medicines, M1 or M2, for two consecutive days. M1 is more effective than M2. Therefore, ideally, she should give her patient medicine M1 today. Moreover, she must absolutely avoid mixing medications (for example, giving M2 today and M1 tomorrow). Does it follow from the above requirements that Dr. Brown should give M1 tomorrow? Of course not! If Dr. Brown mistakenly gives M2 today, tomorrow she should give M2 instead.
If this is true, then what Goodman and Holguin call ‘parasitic norms’ are not norms at all. In non-ideal situations, following such norms can lead to multiple normative violations. For instance, if Dr. Brown follows the parasitic norm to give M1 tomorrow, she would violate both the norm to give M1 today and the norm to avoid mixing medications.
The same point also applies to norms of assertion. Suppose that you should not assert what you are not sure of (SN) and that you should not be sure of what you don’t know (SKN). From these norms it doesn’t follow that you are always committed to a further norm according to which you should not assert what you don’t know (KN). Instead, in sub-ideal circumstances where you are sure that p but do not know that p (in which SKN is violated), given SN, you are fully permitted to assert that p even in the absence of knowledge.
There may even be cases in which violating the (alleged) parasitic norm KN would prevent multiple normative violations. Suppose that you have good prudential reasons to assert what is properly assertible about a specific proposition p – say, you will be given a reward if you assert that p if and only if your assertion will be proper.Footnote 5 You do not have strong evidence supporting the truth of p, but you are unreasonably certain that p. According to SN it may be appropriate for you to assert p, and you should do so given the prospect of the reward. In this circumstance, if you were to follow KN, you would not assert p, and you would thereby violate two norms: You would be sure of what you don’t know (violating SKN), and you wouldn’t assert what is proper (violating prudential requirements).
It is worth observing that, if my argument above is correct, from SN and SKN, not only can we not derive a knowledge norm, but not even any other norm of assertion defended in the recent literature. For example, we cannot derive norms to assert only what is true or what we warrantedly believe (e.g., Weiner, Reference Weiner2005; Gerken, Reference Gerken2011; Whiting, Reference Whiting2013). If there are circumstances where one is permitted to assert what one is unreasonably sure of, then one would also be allowed to assert what is unwarrantedly believed or what is false. This could be seen as a further counterintuitive consequence of Goodman and Holguin’s account, or at least as a disadvantage when compared to the orthodox view. The latter can straightforwardly explain why it is inappropriate to assert something false or unwarrantedly believed. If knowledge entails truth and warrant, these other norms would be descriptively parasitic on KN.
Goodman and Holguin may reply that, although it isn’t inherently inappropriate to assert things that one doesn’t know (i.e., nothing is wrong with the assertion qua assertion), someone who does so would be presenting herself as being sure of something she doesn’t know, which would go against SKN and reveal an inconsistent state of mind.Footnote 6 Notice, however, that this response does not affect my above point, namely, that no knowledge norm of assertion can follow from SN and SKN. Moreover, it is important to keep the propriety of an assertion separated from the rational status of an asserter. When we assess the former, we may not be concerned with the latter. Consider again the example of someone who has independent reasons to assert properly with regard to p (she will receive a huge reward if she does so). Assuming SN, it’s difficult to see anything wrong or inappropriate if this person is sure that p without knowing it, and yet she asserts that p. In this circumstance, not only would it not be improper for that person to assert that p, but refraining from asserting that p would seem inappropriate.
In general, we should not expect that a violation of a norm regulating rational mental attitudes (like SKN) would be reflected in a corresponding impropriety in an assertion that represents one as committing the former violation. Just as it may appear fully proper for Dr. Brown to give medicament M2 to his patient given her previous mistake, it may also be fully appropriate for a subject to assert something that is in conformity with the norm of assertion, even though this assertion conveys that the asserter has violated some other norm. For instance, admitting that sometimes we are unreasonable does not make our admission improper. The propriety of an assertion should be judged given the norms regulating that practice, without conflating this judgment with other evaluations that pertain to different conditions.
Before proceeding further, let me address a potential objection to my argument. One may object that if we assume Standard Deontic Logic (SDL) and understand norms of assertion as taking wide scope over material conditionals, then KN logically follows from SN and SKN.Footnote 7 In SDL, the inference from Ought(A → B) and Ought(B → C) to Ought(A → C) is valid. Thus, if we accept this logical framework, the following inference from SKN WS and SN WS to KN WS is valid:
(SKN WS) It ought to be that [S doesn’t know that p → S is not sure that p]
(SN WS) It ought to be that [S is not sure that p → S doesn’t assert that p]
(KN WS) It ought to be that [S doesn’t know that p → S doesn’t assert that p]
I don’t find this objection convincing. The issues posed by contrary-to-duty imperatives and the Theory of Second Best discussed in this section cannot be simply dismissed by pointing to what inferences are validated in a specific logical framework. On the contrary, such issues, as well as related puzzles such as Chisholm’s and the Samaritan paradoxes, point to serious flaws in SDL’s interpretation of conditional norms. Nowadays, most deontic logicians acknowledge that deontic conditionals cannot be faithfully represented by deontic operators taking wide scope over material conditionals (McNamara and Van De Putte Reference McNamara, Van De Putte and Zalta2022, §4). Standard semantics for deontic modals such as Kratzer (Reference Kratzer2012) also reject wide-scope formalizations of deontic conditionals based on similar considerations.
Assuming SDL and the above wide scope formulations of SKN and SN leads to absurd consequences. Consider the following norm:
(OP) It ought to be that [one keeps one’s promises].
Assuming SDL, from OP and the factivity of knowledge (Kϕ → ϕ), it follows that:
(O¬K¬P) It ought to be that [one doesn’t know that one didn’t keep one’s promises].
From (O¬K¬P), SKN WS and SN WS it would then follow that:
(O¬A¬P) It ought to be that [one does not assert that one didn’t keep one’s promises].
From the claim that one should keep their promises and the wide scope formulations of SKN and SN, we derived the absurd conclusion that one should never admit that they didn’t keep their promises, even if they breached a promise and want to apologize for it. This argument can be generalized. By the same token, one should never acknowledge their wrongdoings and mistakes or apologize for them. Yet sometimes such admissions and apologies seem not only not inappropriate, but required.Footnote 8
Such absurd consequences are just one symptom of a larger problem affecting over-idealized models such as SDL. These models provide abstract representations of what would be the case in ideal circumstances in which all norms and duties are fulfilled. While helpful for specific technical purposes, such models provide an inadequate representation of what norms require from actual agents in real circumstances. For instance, in a SDL framework, the above inference from SKN WS and SN WS to KN WS expresses the almost trivial idea that in ideal circumstances in which an agent satisfies the first two conditions, the latter is also satisfied. Such abstract considerations about what agents would do in ideal circumstances are often irrelevant to address the most pressing issue of what an actual person is required to do in real (possibly sub-ideal) situations – for instance, whether someone may assert p given that she is sure of p, but doesn’t know p. As I have explained above, such considerations could even lead to multiple normative violations if used as normative standards. To the extent that norms of assertion regulate real-world practices, they concern what real agents are permitted to assert given their actual shortcomings. Over-idealized frameworks such as SDL can’t provide plausible models of the norms that bind such practices.
It is worth observing that the problems affecting normative parasitism discussed in this section do not also affect descriptive parasitism, and in particular they do not affect a derivation of the surety norm as descriptively parasitic on the knowledge norm. The problem with normative parasitism relies on the derivation of a requirement from two other ‘primary’ norms. This derivation presupposes that one conforms to both such norms. As this may not always be the case, the derivation posits requirements where they are not supposed to be. By contrast, in derivations involving descriptive parasitism, we only have one primary norm. We then derive further requirements from the constitutive conditions of what that norm requires. This derivation may seem straightforward, as conforming to the derivative requirements is a metaphysically necessary condition to comply with the primary norm. A well-known example from Goble (Reference Goble2009, fn. 41) illustrates this point. Goble observes that if a law requires that there shall be no camping at any time on public streets, it seems to obviously follow that there should be no camping on Thursday night on those streets. ‘Given the general law, the particular is implied’ (ibid.). Similarly, if being sure is necessary for knowing, in order to assert only what one knows one must also assert only what one is sure of. By violating the latter requirement, one would ipso facto violate the former.
While descriptive parasitism is exempted from the problems discussed in this section, this way of deriving norms is affected by another type of issue. This is a restricted version of the well-known Ross paradox, which affects what is known as the Principle of Inheritance (McNamara and Van De Putte, Reference McNamara, Van De Putte and Zalta2022, §6.3; Broome, Reference Broome2013, pp. 121–2). The derivation of descriptively parasitic norms requires a restricted and qualified version of this principle:
(PI) If a condition C1 is required by a norm, and a condition C2 is constitutive of C1, then also C2 is required
As illustrated by Goble’s example, in most cases this type of derivation is very plausible and intuitive. However, sometimes it can lead to counterintuitive consequences. Consider a specific example: one must be alive and breathe in order to help a friend. Now, if Jane has an obligation to help her friend, from (PI) it follows that Jane also has a derivative obligation to be alive and breathe. But this sounds absurd.
How serious is this problem? It is important to observe that, while this derivation may lead to the above counterintuitive consequences, other times it delivers highly plausible results. It is sometimes pretty obvious that one is required to do things that are constitutive of other things one should do. Thus, the lesson we should learn from this problem is not that we cannot derive norms via some form of descriptive parasitism. Rather, we should not think that this sort of derivation is a straightforward consequence of general principles like (PI). We should instead assume some sort of restriction of the Inheritance Principle, allowing us to derive that we should not camp on Thursday from the requirement not to camp at any time, but blocking the derivation from a norm to help a friend to norms to be alive and breathe. This restriction should clarify which kind of constitutive conditions are derivatively required and which aren’t. For instance, one may suggest that the constitutive conditions that are normatively required are ones over which an agent bears some form of control and responsibility or that are about attitudes one can hold for reasons. Whatever the case may be, there seem to be plausible ways in which this restriction could be worked out. Whatever one will say about Goble’s camping example, one could also say about the descriptive derivation of a surety norm from a knowledge norm. In this respect, the current problem seems to be less serious than the one affecting normative parasitism.
3. Knowing without being sure?
Goodman and Holguin’s argument against the orthodox account relies on examples in which a subject (i) knows that p, (ii) is not sure that p, and (iii) it would be inappropriate for her to assert that p. These cases suggest two things. First, knowing does not imply being sure. If so, the derivation of SN from KN advocated by the orthodox view is not possible: from the assumption that one should not assert what one doesn’t know, it does not follow that one should not assert what one is not sure of. Second, it may be inappropriate for someone to assert that p even though she knows that p. This suggests that the primary norm of assertion is surety, not knowledge.
The authors discuss two cases, one involving memory and the other perception. Let’s first look at the memory case:
Suppose Jones reads, and thereby comes to know, that the Battle of Hastings was fought in 1066. We may suppose that, at that moment, he is also sure that the battle was fought then. Years later, he has long since forgotten reading the book. Someone asks him if he knows when the Battle of Hastings was fought. He replies “I’m not sure, but I think it was in 1066”. Although Jones is no longer sure that the battle was fought in 1066, he still thinks that it was, and his thinking this is the result of a memory trace preserved from the time he originally learned it. In such a case, we think Jones knows: he hasn’t forgotten that the battle was fought then, so he still remembers that it was fought then, so he knows that it was fought then. (2022, p. 638)
According to Goodman and Holguin, in this case Jones is not sure that the Battle of Hastings was fought in 1066, but he still remembers it. Since remembering that p implies knowing that p (Williamson, Reference Williamson2000; Moon, Reference Moon2013), Jones knows that the battle was fought in 1066. Moreover, intuitively, it is not appropriate for Jones to flatly assert that the Battle of Hastings was fought in 1066.
I don’t find this case convincing. Simply having an accurate memory trace of p is not enough for knowing (and remembering) that p. Other properties are necessary. For instance, the subject should have a sufficently high degree of credence in p and her attitude toward p should be based on reliable grounds. These properties do not seem present in Jones’s situation. It seems clear from the description of the case that Jones’s memory is very weak. Jones’s admission of uncertainty clearly conveys that he is only marginally confident of the information. Moreover, forming opinions based on shaky memories such as those of Jones cannot guarantee a level of reliability sufficient for knowing (and remembering).
Goodman and Holguin claim that, based on his fading memories, Jones ‘still remembers that [the battle] was fought then’. To my ears, this claim sounds implausible. When we consider someone having a distant and fading memory of a past event, we typically do not claim that she remembers what happened. Instead, we tend to hedge our statement with expressions like ‘She seems to remember…’ or ‘it seems to her that…’.Footnote 9 This suggests that, while Jones may have a memory trace of the information, we don’t consider his attitude as a paradigmatic instance of propositional remembering. As it is unclear whether Jones remembers the relevant fact, it is also unclear whether he really knows it.
The above case is underdescribed. Goodman and Holguin may reply that this is not how they intend the case to be interpreted. They could redescribe the case in such a way that Jones is actually very confident that the battle was fought in 1066 (p) and that his memory trace provides a sufficiently reliable basis for knowledge. However, if this were the case, Jones would not claim to be unsure and merely think that p. Once we fill in the relevant details, it seems that either the case does not involve propositional remembering (and knowing) or it also involves being sure (in the ordinary sense of being sure specified above). To make a compelling case for their view, the authors need to find an example where it is relatively uncontroversial that the agent meets normal confidence and reliability conditions to qualify as knowing, but not to qualify as being sure. Since proponents of KN deny that such cases are possible, the burden of proof falls on Goodman and Holguin to find uncontroversial cases of this sort. Without such cases, their argumentative strategy seems dialectically ineffective.Footnote 10
Notice also that, by claiming ‘I’m not sure, but I think it was in 1066’, Jones implies not only that he is unsure of the information, but also that he does not know (remember) it. As argued by Benton and Van Elswyk (Reference Benton, van Elswyk and Goldberg2020), hedged assertions such as ‘I think that p’ and ‘I believe that p’ weaken the speaker’s commitment to what is asserted by specifying that the speaker has a weak attitude towards p. Such claims are typically used to express a partial lack of confidence in p or a weak epistemic position towards p, suggesting that that the speaker doesn’t take herself as knowing that p.Footnote 11 These considerations do not show that Jones doesn’t know that the Battle of Hastings was fought in 1066 (p). However, they strongly suggest that Jones doesn’t take himself as knowing it. Even assuming that Jones knows that p, he is not aware of it. This leads us to two important points about the above case.
First, if Jones is mistaken or lacks awareness about what he knows, he may also be mistaken or lack awareness about what he is sure of. As we can interpret the case as one in which Jones unwittingly knows that p, we can seemingly interpret it as one in which he is unwittingly very confident that p, even sure of it. This puts pressure on Goodman and Holguin’s claim that Jones knows but isn’t sure that p. On the one hand, Jones doesn’t take himself to know or to be sure that p. On the other hand, for all that we are told, he could be mistaken on both counts: he could both know and be sure that p. Nothing in the description of the case suggests that we should treat Jones’s knowledge and surety states as diverging regarding p. Whatever one could say about Jones’s knowledge, it could also be said about his surety. So the case does not support the surety norm over the knowledge norm.
Someone may challenge the reasoning above by claiming that, intuitively, it seems undeniable that in this case Jones is not sure that p. In response, it seems equally undeniable that, intuitively, Jones doesn’t know that p (see my arguments above). If in Goodman and Holguin’s argument it is legitimate to bracket our intuitions about whether Jones knows, it should also be deemed legitimate to bracket intuitions about whether Jones is sure. Moreover, as it could be argued that Jones’ memory entails that he (unwittingly) knows that p, it can be argued that the same memory entails that he is (unwittingly) sure that p. Again, whatever one says about knowledge can be said about surety.Footnote 12
Second, that Jones doesn’t take himself as knowing that p gives to upholders of the knowledge norm a straightforward way to explain the intuition that it is inappropriate for Jones to flatly assert that p. If Jones were to assert that p without being aware that he knows that p, he could be criticised for being reckless and insensitive to normative considerations, and for showing disrespect towards KN (Williamson, Reference Williamson2005). This situation would be similar to one in which an agent conforms to a law by sheer luck, without being aware that they are conforming to it. This type of behaviour deserves to be criticized as reckless and disrespectful of the law. Similarly, even if we assume that Jones knows the relevant fact, the orthodox account can still explain the intuition that Jones would be criticizable for asserting that p by pointing to his lack of awareness that he knows that p. While Jones’ assertion that p may conform to KN, asserting that p without being aware of knowing that p would deserve criticism for conforming to KN in a reckless and disrespectful way. Thus, an upholder of KN can explain the intuition that for Jones it would be inappropriate to assert that p without appealing to a surety norm.
Goodman and Holguin also consider a second case involving perception:
Smith is looking at a red wall in ordinary light. But she has some suspicion that it might be a white wall in red light. As a result, she is not sure that the wall is red. Still, her visual system is functioning normally. It is therefore plausible that Smith sees that the wall is red. Moreover, it is plausible that seeing that something is the case is a way of knowing that it is the case. (2022, p. 638)
This case is supposed to describe a situation in which someone sees that p and thus knows that p, without being sure that p. As in the previous case, it seems that Smith cannot properly assert that the wall is red. Based on this, the authors conclude that the norm of assertion is not knowledge, but surety.
The case is underdescribed. How confident is Smith that the wall is red? Why does Smith suspect that the wall could be in red light? These (and other) details are essential in determining whether Smith really sees that the wall is red, and thus whether she knows that. For instance, if Smith has reasons to believe that the wall is being illuminated by a red light, this would count as a defeater of her justification that will prevent her, not only from being reasonably sure, but also from knowing that the wall is red. If rational, Smith would also have low confidence that the wall is red, not enough to know it. In this situation, it would be inappropriate to describe Smith as seeing that the wall is red. Instead, she would be merely perceiving a red colour on the wall. In other terms, if Smith has some serious and reasonable doubts that the wall is red, we would describe her mental state as an appearance rather than an instance of propositional seeing (and knowing).Footnote 13
By contrast, if Smith is merely entertaining the far-fetched possibility that there could be red light projected on a white wall, without any reason to think this is true, then it seems that Smith should not have any suspicion that the wall is red. However, in this case she would see (and know) that the wall is red; but intuitively she would also be sure of it and entitled to assert that the wall is red.
In conclusion, this case, like the previous one, does not show that there can be cases in which (i) one knows that p, (ii) one is not sure that p, and (iii) one’s assertion that p is improper. Rather, depending on how we interpret such cases, they may involve a perfect alignment of knowledge, surety and proper assertibility.
4. Conclusion
In this paper I have argued that Goodman and Holguin’s picture of the relationship between knowledge and surety norms of assertion is problematic. I have shown that a knowledge norm cannot be derived from a surety norm and the assumption that we should be sure only of things we know. I have also argued that Goodman and Holguin’s arguments against the orthodox view fail. Once properly clarified, the cases they discuss can be easly accommodated within the standard framework.
The orthodox view also has its own problems. It relies on the contentious assumption that knowing implies being sure. Moreover, this view assumes that a surety norm can be derived from a knowledge norm in a descriptively parasitic way. As I explained in Section 2, this type of derivation is not straightforward either. Despite these concerns, in light of the issues discussed in this paper, the orthodox view still seems preferable to the one advocated by Goodman and Holguin.
Acknowledgements
I would like to thank two anonymous reviewers for helpful comments on earlier drafts of this article. An earlier version of this article was presented at the School of Philosophy WIPS Meeting at Zhejiang University. Thanks to the audience for their helpful feedback. In particular, I would like to thank Dong An, Bruno Bentzen, Xiaoyu Ke, Zhongwei Li, Luqi Tang, Ke Zhang, and Yili Zhou. Special thanks to Bruno Bentzen for helpful comments on an earlier version of the article. Part of the research work that led to this article was supported by the National Social Science Fund of China (No. 21CZX016).
