Chapter 4 - The Evidence View

Summary

In what precedes, we have explored two major reductionist approaches to normative reasons. And the implicit assumption with which we have been working was that all existing reductionist views could be reduced to one or the other of two camps: the Reasoning view camp that appeals to good/fitting patterns of reasoning from fitting premise-responses to fitting conclusion-responses in order to define reasons; or the Explanation view camp that appeals to the role of reasons in explaining some normative facts (ought-facts or value-facts). However, one might reasonably protest at this way of partitioning the debate, for a very prominent view within the reasons, literature doesn’t seem on the face of it to fit into this picture, namely the Evidence view of reasons, which doesn’t seem on the face of it to fall into either the Reasoning view camp or the explanationist camp. According to this view, as defended most notably by Stephen Kearns and Daniel Star in a number of publications (Kearns and Star 2008, 2009, 2013, 2015), normative reasons to F are evidence that one ought to F.

4.1 Introduction

In what precedes, we have explored two major reductionist approaches to normative reasons. And the implicit assumption with which we have been working was that all existing reductionist views could be reduced to one or the other of two camps: the Reasoning view camp that appeals to good/fitting patterns of reasoning from fitting premise-responses to fitting conclusion-responses in order to define reasons; or the Explanation view camp that appeals to the role of reasons in explaining some normative facts (ought-facts or value-facts). However, one might reasonably protest at this way of partitioning the debate, for a very prominent view within the reasons, literature doesn’t seem on the face of it to fit into this picture, namely the Evidence view of reasons, which doesn’t seem on the face of it to fall into either the Reasoning view camp or the explanationist camp. According to this view, as defended most notably by Stephen Kearns and Daniel Star in a number of publications (Kearns and Star 2008, 2009, 2013, 2015), normative reasons to F are evidence that one ought to F.

The focus of the present chapter is on the Evidence view of reasons. In what follows, we explore the details of the view along with arguments in its favour that appear to make it quite promising. But we also consider some of the most popular objections to the view. We will conclude, however, that contrary to what one might initially think, the Evidence view, when fully developed, reduces to a version of the Reasoning view. And thus the view also suffers from some of the same problems that we have observed with respect to that approach. In this way, it will miss out some of the aspects of normative reasons as we commonly understand them, the aspects for which reasons actually matter for us. Nevertheless, working through the details of this view will also help us to see what exactly a theory of reasons should be able to do. A new proposal that builds on the conclusions from this and discussions in the previous chapters will be elaborated in the chapter after this.

4.2 The View

According to a prominent version of the Evidence view of normative reasons, a normative reason for one to F just is evidence that one ought to F. This version of the view has been notably defended by Stephen Kearns and Daniel Star in a number of publications (see Kearns and Star 2008, 2009, 2013, 2015). Before assessing the merits of the view, let us first see in a bit more detail what exactly the view suggests.

The official version of the view that we will focus on in what follows reads as follows:

R Necessarily, a fact F is a reason for an agent A to φ iff F is evidence that A ought to φ (where φ is either a belief or an action).

(Kearns and Star 2009: 216)

The first thing to note about the view is that it is presented as a simple bi-conditional, which, of course, need not be taken as presenting an explanatory analysis/definition of reasons. Kearns and Star are clear that their arguments in favour of the Evidence view support, strictly speaking, only this bi-conditional. However, they do commit themselves to a stronger position: ‘we also believe that the best explanation of the truth of all these principles [e.g. the genus principle R and more specific reasons as evidence principles applied to reasons for beliefs and reasons for action] is that the property of being a reason and the property of being evidence of an ought are identical’ (Kearns and Star 2009: 219).

Another thing to note is how Kearns and Star characterise evidence. An important element in that characterisation is being a reliable indicator. But they don’t endorse the somewhat strong and unqualified claim that evidence that p just is a reliable indicator that p. They suggest that the possibility of there being misleading evidence for necessary truths is one reason why such a view should not be accepted, and that that a generally reliable source of information (e.g. a phone book) can be wrong about some specific p (e.g. the number of someone in particular) is another reason why evidence is not just a reliable indicator of p. Some facts can be evidence that p even without it reliably indicating that p (see Kearns and Star 2009: 229–230). Nonetheless, they maintain that something similar enough to the ‘reliable indicator’ conception of evidence should hold. For instance, they write:

We may therefore conclude that a fact is evidence for a proposition if and only if relevantly similar facts reliably indicate relevantly similar propositions. In the normative case, then, we can say that a fact F is evidence that one ought to φ if and only if facts relevantly similar to F reliably indicate propositions relevantly similar to the proposition that one ought to φ.

(Kearns and Star 2009: 230)

The phone book example illustrates this idea. The normally reliable phone book is wrong about John’s number. Thus, Kearns and Star concede that the book indicates that such and such is the person’s number (a fact) doesn’t reliably indicate that such and such is the person’s number. But that fact is evidence that such and such is the person’s number. It is so because, according to Kearns and Star (2009: 230), ‘there are many relevantly similar facts that do reliably indicate relevantly similar propositions (e.g. the book does reliably give the correct numbers for Mary and Henry and many others).’

Another aspect of their view of evidence is that they endorse a probabilistic conception of evidential support. This is apparent from their treatment of the strength of evidence. They write: ‘The strength of a piece of evidence E for a proposition P depends on the degree to which E increases the probability of P’ (Kearns and Star 2009: 231–232). And more specifically, they tell us that appeal to increase of probability also captures the ways in which one piece of evidence is stronger than another, outweighs other evidence, and can combine with further evidence:

The more probable P is given E, the stronger evidence E is that P is true. E is stronger evidence than another piece of evidence E* for P if and only if E makes P more probable than E* makes P. E outweighs a piece of evidence E* if and only if E is evidence for P, E* is evidence for ∼P and E makes P more probable than E* makes ∼P. Two pieces of evidence, E and E* can combine to form stronger evidence if the probability of P given the conjunction of E and E* is greater than both the probability of P given E and the probability of P given E*.

(Kearns and Star 2009: 232)

The kind of probability that they have in mind here is ‘evidential or epistemic probability’ (cf. Kearns and Star 2009: 232, fn 10). They refer to Williamson (2000, chapter 10) for an exploration of this kind of probability.¹ A question that one might have at this point is how exactly the quasi-indicator understanding of evidence is supposed to fit with the increase of evidential probability understanding of evidential support. For one thing, one might wonder whether it is always the case that when relevantly similar facts to F reliably indicate relevantly similar propositions to p, it is always the case that the probability of p given F is higher than the probably of p without F (say, given some other facts). Maybe the aforementioned example of the possibility of misleading evidence for a necessary truth (that Kearns and Star themselves provide as a genuine instance of evidence) is one case in point. Given that it is a necessary truth, nothing can increase nor decrease its probability (which is 1), be it quasi-reliable indicator or not. See also Logins (2016b) for a related objection from necessary truths to a version of the probabilistic conception of evidence. But let us not dwell on this specific potential worry. Arguably, there might be ways to fix the problem of necessary truths (for instance, by giving up the idea that there is evidence for necessary truths at all or to alter the probabilistic conception of evidential support, or to add a clause about special treatment for necessary truths).

4.3 Arguments in Favour

Kearns and Star offer a battery of considerations in favour of their view. I suggest we focus here on what appear to be the four strongest arguments. The first line of thought that, I think, provides a good prima facie case in favour of the evidence is based on the simple observation that, typically, in standard cases it does make sense to explain normative reasons in terms of evidence for oughts. Often we can simply paraphrase the reasons statements with evidence for ought states, which, of course, gets a straightforward explanation if something like the Evidence view is on the right track. Kearns and Star suggest that considerations of this sort offer a good inductive argument in favour of their view.² In support of the claim that standard situations where one has reasons to F are also situations where one has evidence that one ought to F, Kearns and Star provide two examples. But they also suggest that there are many more standard cases, indeed a large number of situations that would provide fitting examples to illustrate the point (see Kearns and Star 2009: 222–223). The first example is a case where a subject sees a friend who is in terrible pain; his foot is stuck under the wheel of a car. The suggestion here is that the fact that the friend is in pain is a reason to help the friend, and also it is a reason to believe that the subject ought to help his friend. Crucially, the same fact is also evidence that the subject ought to help the friend. As the subject is helping the friend, he remembers that he has promised to meet another friend in two minutes’ time for a philosophical conversation. That he has promised to meet the other friend is a reason to leave and go to meet the other friend. It is suggested that it is also evidence that the subject ought to rush to meet that other friend. Of course, the first friend’s pain provides much stronger evidence that the subject ought to stay and help than the evidence that the subject ought to rush to meet the other friend. Finally, it is suggested that the reason to stay and help (pain) is much stronger than the reason to rush away (promise).

The second example concerns someone who likes to spend evenings either reading some scientific book to better understand the world or by relaxing and watching TV. It is suggested that the fact that reading a book would lead him to better understanding of world is a reason for him to read the book, and it is also evidence that he ought to read the book. And the fact that watching TV would provide the subject with some pleasure is a reason for him to watch TV, and it is also evidence that he ought to watch TV. In this example, it is then suggested that the subject weighs his reasons and determines that he ought to read the book, which is taken to be just the same as ‘saying that he weighs the evidence that he ought to read a book against the evidence that he ought to watch television, and judges that he ought to read a book on this basis’ (Kearns and Star 2009: 223). These are only two examples of a much larger number of standard cases (i.e. cases of transparent facts and no misleading evidence; cf. Kearns and Star 2009: 223) where a fact is both a reason to F and evidence that one ought to F. On the basis of induction, one seems to be warranted, then, to conclude that being a reason to F just is being evidence that one ought to F (or at least to hold the corresponding bi-conditional about being a reason and being evidence that one ought to F).

The second argument that I would like to rehearse here appeals to considerations about deliberation. The official version of the argument can be summed up as follows: reasons to F have to be able to play a role in reliable reasoning about whether one ought to F or not; only evidence that one ought to F can play that sort of role in reliable practical reasoning. Thus, normative reasons to F are evidence that one ought to F (cf. Kearns and Star 2009: 224). As such, this argument might be a bit of an overstatement. One might think that things other than evidence can play a role in reliable reasoning. For instance, unless you think that all premises in reliable/good reasoning about whether one ought to F are pieces of evidence, you might think that a non-evidentialist Reasoning view might account for the reasons–reasoning connection just as well as, if not better than, the Evidence view. We will come back to this, or something close enough to this, thought when examining the worries faced by the Evidence view of reasons. For now, we can agree that a toned-down version of the argument that appeals to reasoning can be taken to provide some support for the Evidence view, namely the argument that the Evidence view is well suited to account for most of the reasons–reasoning connection, given some further assumptions about the role of evidence in reasoning. One way of filling out the details here would just be to say that by definition playing some crucial role in reasoning just is being evidence, and being evidence just is playing a role in reasoning. If so, then of course, by definition, the Evidence view would provide straightforward explanation of the most relevant aspects of the reasons–reasoning connection and its simplicity alone would constitute a basis for preferring it to potential attempts by rival non-reasoning-centred views to account for the observed connection. But let us not anticipate our discussion on the Evidence view as a version of the Reasoning view.

The third argument concerns another aspect in which the Evidence view seems to have an advantage at least over some of the rival views. The Evidence view has the advantage (over some views) in terms of the simplicity of the explanation that it can provide of why reasons appear to have various degrees of ‘strength’ or ‘weight’. On the Evidence view, the explanation is straightforward and hence theoretically powerful: it’s just because evidence (for a proposition) by definition comes in various degrees of strength, given the probabilistic gloss over the strength of evidence. Given Kearns and Star’s understanding of evidence, it is only a consequence of their view that reasons (to F) are gradable; their degree of ‘strength’ is inherited from the degree of increase of probabilistic support for the proposition that one ought to F provided by the corresponding piece of evidence. Again, the official version of the argument might be a bit too ambitious, since, strictly speaking, Kearns and Star claim that the Evidence view is ‘the best explanation’ of the gradability of reasons,³ which might be contested, for instance, by proponents of some alternative views that we have seen earlier. But even if one tones down the argument slightly, and only claims that the Evidence view of reasons provides a straightforward and theoretically powerful explanation of the gradability of reasons, it is already a good prima facie reason for taking the Evidence view seriously. For it is at least among the best possible potential explanations of the gradability of reasons, and certainly it does better in this respect than some of the alternatives. Note, however, that unfortunately Kearns and Star only compare their view in this respect to reasons-first and Broome’s version of deontic explanationism. But, of course, a fuller picture would also need to compare it to other versions of explanationism (e.g. value-based accounts and Reasoning views), in which case it is not clear that the Evidence view would constitute the best (e.g. the simplest) possible explanation of the gradability of reasons. Nevertheless, that the view can easily account for the apparent gradability of the ‘strength’ of reasons is certainly a point in its favour.

The fourth consideration in favour of the Evidence view is that it does respect the theoretical unity constraint on a theory of reasons. More precisely, the account applies perfectly both to normative reasons for actions and normative reasons for attitudes. Strictly speaking, Kearns and Star only explain how it applies to beliefs, but they stipulate that the view can be easily extended to other attitudes as well (cf. Kearns and Star 2009: 216, fn2).⁴ Kearns and Star assume that the default position should be that normative reasons for attitudes and normative reasons for action are of the same kind. They do provide considerations in favour of taking this to be the default position, namely, the consideration that we can refer to one and the same fact as being a reason both to act in a way and to believe a proposition, something that becomes difficult to explain if reasons for action and attitudes are of a totally different sort, and the consideration that apparently reasons to believe and reasons to act can be weighed against each other (see Kearns and Star 2009: 220–221). The suggestion, then, is that the Evidence view has a straightforward explanation of this. Again, let us focus on this interpretation of the argument, rather than the bold claim that the Evidence view provides the best possible explanation of the apparent fact that reasons for attitudes and reasons for actions are of the same kind. On the Evidence view of reasons, normative reasons just are evidence, in the belief case, reasons to believe that p are evidence that p and also/or evidence that one ought to believe p (depending on whether one wants to leave the pragmatism option about reasons to believe open), and in the case of action, reasons to F are evidence that one ought to F. The unifying principle in reasons to believe and reasons to act is that both are evidence for some proposition. Given this common element, the Evidence view is perfectly placed to explain why reasons to believe and reasons to act are of the same kind. This explanatory power in complying with a natural constraint on a theory of reasons is then another prima facie consideration in favour of the Evidence view of reasons.

4.4 Worries

Despite its prima facie plausibility, the Evidence view has attracted a number of critics. A non-exhaustive list of publications that contain objections to the Evidence view of reasons includes Broome (2008), Brunero (2009), McNaughton and Rawling (2011), McKeever and Ridge (2012), Fletcher (2013), McBride (2013), Setiya (2014), Littlejohn (2016), Schmidt (2017), and Hawthorne and Magidor (2018). In what follows, we will not cover all the existing objections, however. We will focus only on some of the most problematic aspects of the view.

Testimonial evidence. To begin with, let us consider what seems to be the most popular line of objection against Kearns and Star’s Evidence view of reasons, namely the line of thought that some examples clearly demonstrate that not all pieces of evidence can be normative reasons. Consider the example that Kearns and Star (2009: 233) actually discuss themselves as a possible counterexample to their view (the example is attributed to James Morauta). A reliable newspaper states that there are people starving in Africa. That this reliable newspaper says it, is clearly evidence that one ought to give money to a charity, say, Oxfam (presumably, the assumption here is that this would somewhat help to alleviate the suffering of starving). Now, the suggestion is that while that the reliable newspaper repeats that there are people starving in Africa is evidence that one ought to send money to Oxfam, it is not by itself a reason for one to send money to Oxfam. That people are starving is a reason. Not that someone, reliable or not, says so. However, the Evidence view of reasons predicts that the mere report from the newspaper is a reason for one to send money to Oxfam. Thus, the case seems to present a counterexample to the Evidence view of reasons.

The initial reply that Kearns and Star offer to this sort of potential counterexample contains two possible ways of dealing with the objection. One line of reply is to bite the bullet and to claim that in the normative domain, when something is evidence that one ought to F, it has a potential to create effectively an obligation, the moral obligation in the case of Oxfam, to give money. The mere fact that the newspaper reports the starvation can create a reason to give money to Oxfam. They write: ‘The fact that the newspaper says that people are starving in Africa may in itself be enough to create an obligation to send money to Oxfam’ (Kearns and Star 2009: 233). This is, of course, a bit too quick. For this is exactly what is at issue here (assuming that, at least typically, having an obligation to F entails having a reason to F, and assuming that those who propose this counterexample would also put forward the intuition that a mere report by a newspaper doesn’t yet create by itself a moral obligation). To say that the newspaper’s report creates an obligation without providing further theoretical grounds for why this is so would amount to begging the question against those who offer the newspaper example against the Evidence view of reasons.

One might also want to know what exactly is ‘enough’ in the example ‘to create an obligation to send money to Oxfam’. Is it the fact that people are starving in Africa, or is it that the newspaper reports it? If the former, then the creation of obligation has nothing to do with there being evidence that one ought to do the relevant thing and so this is still a counterexample to the Evidence view. But if the mere fact that the reliable newspaper publishes something is enough to create an obligation, then, arguably, we have many more moral obligations than we could ever suspect that we have (e.g. today’s newspaper published all sorts of things, some bad, some not; is there a moral obligation for me corresponding to each of the assertions? If not, why not? What’s so special about the starvation case?). Anyway, we need more theoretical considerations if we are to take the first response to be anything more than just begging the question against the objection from the newspaper counterexample.

The second line of response is based on the idea that implausible consequences will follow if the newspaper’s report is not a reason to send money to Oxfam. On the face of it, this line appears to be more promising. Kearns and Star’s published version of this response is relatively short. I attempt here to work out the implicit details of the argument. The suggestion seems to be that we only have the relevant pre-theoretical judgments in the newspaper case (i.e. that the mere report is not a reason to send money) because the fact that even a reliable newspaper can be mistaken is somehow made salient. And it is made salient mainly because of the focus on the non-conclusive aspect of the newspaper’s report. But denying that the newspaper’s report is a reason on the basis that it is non-conclusive (or non-entailing) leads to a slippery slope, according to this line of objection. If one denies that the fact that the newspaper says that people are starving is a reason to F, because of the non-entailing character of the report (i.e. that a reliable newspaper says that p doesn’t entail that p), then one will also have to deny that the fact that people are starving is a reason to donate to a charity that can help them. For even the fact that people are starving doesn’t on its own entail that these people are in a terrible condition. Even starving is only a non-entailing condition with respect to suffering. They write: ‘After all, the fact that people are starving need not mean that they are badly off. In some distant possible worlds, starving might be extremely pleasant and not life threatening at all’ (Kearns and Star 2009: 234). And they add: ‘Indeed, most of the facts we cite as reasons merely indicate what we ought to do. Therefore, unless one wishes to deny that most facts that we think of as reasons really are reasons, the fact that the newspaper says there are people starving in Africa is really a reason to send money to Oxfam’ (Kearns and Star 2009: 234).

So, this line of reply seems to have two elements. The first is the error-theory element – that is, we only have this common-sense pre-theoretical judgment because we focus on the non-entailing aspect of the newspaper report, not because it is not a reason. And the second is the counterargument element – the reductio argument according to which if the newspaper report is not a reason, then the fact that people are starving in Africa isn’t one either, but, of course, that people are starving in Africa is a reason to send money to Oxfam; hence the newspaper report is a reason to send money to Oxfam too.

Despite being somewhat more promising than the first reply, this line of objection is nonetheless unsatisfactory. For one thing, as John Hawthorne and Ofra Magidor (2018) have observed, the judgment about testimonies being mere evidence that one ought to F without being a normative reason doesn’t seem to be tied to the non-entailing aspect of the testimony in this sort of case. Indeed, as they have showed, the same sort of pre-theoretical judgment seems to be elicited even when we focus on cases of testimony that cannot be wrong (in the relevant sense of ‘cannot’). The case that Hawthorne and Magidor offer to illustrate the point focuses on a far-away hermit who knows the relevant facts. Knowing that p entails p. Moreover, given that knowledge is safe, it couldn’t easily be the case that p is false (i.e. the worlds where p is false are not among close possible worlds to the world where the hermit knows that p). Here are the details of the case from Hawthorne and Magidor (2018: 131).

To our ears, if a hermit in a cave on the other side of the world knows that the apple is poisonous, it remains hard to recover a context where it seems true to say ‘That someone knows that the apple is poisonous is a reason for the agent not to eat it’ (though of course the fact about knowledge is evidence that the agent ought not to eat it).

Thus, neither the error-theoretical aspect nor the reductio part of Kearns and Star’s second response to the objection from the newspaper case seems to be well motivated. It’s simply not the case that our pre-theoretical judgments in this sort of case stem specifically from the focus on the non-entailing aspect of the relevant considerations, since the hermit case doesn’t have this aspect. Some facts that constitute entailing evidence that one ought to F don’t constitute a reason to F, contrary to what the Evidence view states.⁵

Elsewhere, Kearns and Star (2008: 48–49) discuss a somewhat similar objection suggested by John Broome in correspondence (Broome 2008 repeats and modifies the objection) that appeals to a case of normative testimony in particular. The objection is similar to, and yet also crucially different from, the newspaper example objection. It focuses on a putative counterexample, discussed earlier, where a reliable book says that one ought to eat cabbages. The suggestion is that that the book states that one ought to eat cabbage is evidence that one ought to eat cabbage, yet on its own it is not a reason to eat cabbage. And, of course, for Broome, it is not a reason to eat cabbage, because it is not [part of] an explanation of why one ought to eat cabbage (but see also our discussion of constitutive deontic explanationist view in Section 3.2, where a similar observation about meta-ethical theory saying that one ought to F was taken to speak against Broome’s own deontic explanationism). The objection is similar, since it is also a case of testimonial evidence, yet it is different because it is a case of evidence for an ought fact and, crucially, the objection assumes that reasons are right-makers – that is, that they have to play a role in explaining a normative fact. And Kearns and Star make it explicit that they are not committed to reasons being centrally right-makers (e.g. explanans of normative facts): ‘as we have mentioned, we are not committed to the idea that all reasons are right-makers. Thus even if the fact that a reliable book says one ought to eat cabbage is not part of an explanation of why one ought to eat cabbage, we are still open to its being a reason’ (Kearns and Star 2013: 73). But they also provide further considerations against Broome’s argument by granting for the sake of the argument that reasons are right-makers. The line of reply there is to put forward a dilemma. They write: ‘Either facts such as the fact that the book says you ought to eat cabbage are parts of explanations of what one ought to do, or many other facts (such as the fact that cabbage helps the digestive system) are not parts of explanations of what one ought to do’ (Kearns and Star 2008: 48). This is so because explanations, according to Kearns and Star, can be fundamental or non-fundamental. Fundamental explanations of normative facts would appeal to correct normative theories (e.g. utilitarianism, deontologism). Many things that we offer as reasons don’t appeal directly to fundamental normative theories (e.g. ‘that cabbages are good for the digestive system’; Kearns and Star 2008: 49). So, again, the line of objection is that if the normative testimony of the book is not a reason (because it is not a part of an explanation of why one ought to eat cabbage), then many other considerations that we typically take to be reasons will not count as reasons after all. For they too will not count as fundamental explanations of why one ought to eat cabbages. But, of course, all these considerations (e.g. that cabbages are good for the digestive system) are reasons, so, by reductio, the normative testimony from the book is one too. If we focus on non-fundamental explanation as the mark of reasons, then normative testimony too, exactly like many other non-fundamental considerations, should count as reasons for one to eat cabbage according to this line of thought.⁶ I would like to suggest that this line of reply is, however, problematic. The main problem is that Kearns and Star themselves recognise that they are not committed to reasons being right-makers. And indeed it seems they shouldn’t, given that on the face of it evidence that p and explanans of p/p-maker can commonly come apart. But would Kearns and Star even avoid the counterexample by endorsing the claim that reasons are always right-makers? Actually, we might doubt that. For one thing, as we saw earlier, when discussing explanationist views, not all normative reasons are parts of explanation of deontic facts. It is also not the case that evidence that one ought to F is always a part of an explanation why one ought to F. We have seen some of the relevant cases already in our discussion of the Reasoning view and will return to this shortly (e.g. undercutting defeaters, self-undermining considerations, Moore-paradoxical considerations). Thus, it is still not clear why we should accept the idea that cases of testimony, like the newspaper case or the reliable cabbage book case, don’t constitute straightforward refutation of the Evidence view of reasons. Some considerations are evidence that one ought to F but don’t seem to constitute reasons for one to F.

If the reader is not convinced by the aforementioned simple examples of testimony (of the descriptive or normative variety), here is another, somewhat more sophisticated counterexample. It constitutes a basis for an objection against the right-to-left direction of the Evidence view of reasons (i.e. the claim that if e is evidence that one ought to F, then e is a reason for one to F). Consider the following example reconstructed from a recent counterexample to the Evidence view proposed by Eva Schmidt (2017). A student ought to do his best in his upcoming biology exam. In order to do his best, he has to study all Sunday. As it happens, the student goes to a party on Saturday night and gets totally drunk. As a result, he remembers absolutely nothing about the exam, studying, or biology upon waking up on Sunday morning. He notices, however, a note from his dad that enjoins him to study for his biology exam. Despite the heavy hangover, he nevertheless manages to infer from this note (correctly!) that he ought to study for his biology exam. Now comes the crux. From this he then further infers by an inference to the best explanation that he ought to do his best in the exam (we assume here that indeed there is no other equally good or better explanation for him at this point). This case is a problem for the Evidence view, since the Evidence view predicts that the fact that the student ought to study for the exam is a reason for him to do his best in the exam. But this just seems wrong. Schmidt (2017: 712) writes:

But the fact that Henry [e.g. the student] ought to study for the exam is clearly not a normative reason for him to do his best in it: It does not count in favor of his doing his best in the exam. If anything, that he ought to do his best in the exam is a fact that counts in favor of his studying for the exam, or a normative reason for him to study for the exam

Schmidt further suggests that the fundamental problem that this sort of example reveals is that the Evidence view misconstrues the symmetry relations. According to Schmidt, the evidence that one ought to F is symmetrical in the case of evidence in favour of fundamental ought facts and in the case of evidence in favour of derivative ought facts. However, roughly, the normative ‘favouring relation’ between derivative and fundamental ought facts is asymmetrical (cf. Schmidt 2017: 716). This verdict seems to speak in favour of views that define reasons as right-makers of some sort. Without entering into details of the proposed verdict, we can nonetheless agree that in the exam case, the derivative ought fact (i.e. that the student ought to study) can be evidence in favour of the [more] fundamental ought fact that the student ought to do his best in the exam (the hangover amnesia case illustrates this), while the derivative ought fact cannot be a reason for one to do the thing ordered by the fundamental fact (i.e. doing one’s best in the exam). Thus, it is another case where the right-to-left direction of the Evidence view seems to fail.⁷

Enabling conditions (again). Another substantial set of worries for the Evidence view comes from considerations about enabling conditions (and the worry is similar to a worry we observed earlier for the Reasoning approach; see Section 2.5). The Evidence view seems to be unable to distinguish on theoretically well-motivated grounds mere enabling conditions for appropriate F-ing (or enabling conditions for some other facts to be reasons to F) from genuine normative reasons for one to F. For in certain situations, enabling conditions may constitute evidence that one ought to F (especially assuming the increase in probability conception of evidential support) without constituting reasons to F. A notable illustration of an argument against the Evidence view of this form within the recent literature can be found in Brunero (2009) (but see also Fletcher 2013 for an argument that appeals to a different sort of enablers; see also Brunero 2018). Brunero relies on Jonathan Dancy’s insightful discussion on enablers versus reasons (see Chapter 2 for details). In particular, Brunero asks us to consider a case of a promise. Let’s say that I have promised to F. My promise to F is a normative reason for me to F. But consider the fact that there is no reason for me not to F. From a pre-theoretical point of view, this second consideration is not a reason for me to F. However, this consideration matters in a normative sense. It enables me to move, in an instance of deliberation, from the fact that I have promised to F to F-ing (i.e. to move to F-ing in an appropriate/fitting way); alternatively, it allows me to conclude [appropriately] that I ought to F (see Brunero 2009: 544). The worry is that the Evidence view predicts that this second consideration, the enabling condition, is a genuine normative reason for me to F. For, presumably, the probability that I ought to F, given that there is no reason for me not to F and given that I have promised to F, is higher than the probability that I ought to F, given merely that I have promised to F. This is problematic for the Evidence view, since, as we observed earlier, that there is no reason for me not to F doesn’t seem to be a genuine reason for me to F.

In more recent writings, Kearns and Star reply to Brunero’s argument from enabling conditions. Their response is to try to put some pressure on the assumption in Brunero’s argument that the fact that there is no reason for me not to F is not a genuine normative reason to F. Their argument proceeds by considering a different case, where they claim it is natural to admit that an enabling condition can constitute a genuine reason to F, in virtue of there being some further facts. Once this is accepted, they argue, it is difficult to see why the same sort of move cannot be appealed to in the case of the promise. That is, it is difficult to see, according to Kearns and Star, why the fact that there is no reason for me not to F cannot count as a reason to F, given that some further facts about F-ing are in place, in particular, given that I have promised to F. Consider the following passage that captures the central aspect of their response to the objection from ‘no reason not to F’ enablers:

Bob gets depressed by various facts. In particular, Bob gets very depressed by the fact that he has no reason not to take anti-depressants. Given that Bob gets very depressed by the fact (when it is a fact) that he has no reason not to take anti-depressants, the fact that Bob has no reason not to take anti-depressants (when it is a fact) is a reason for Bob to take anti-depressants.

(Kearns and Star 2013: 84)

The line of thought here seems to go as follows: the fact of the form ‘there is/S has no reason not to F’ is a reason for S to F in the depression case; the depression case is a standard case (and is sufficiently similar to the promise case); hence, the fact that one has no reason not to F (e.g. the promised thing) may be a reason for one to F. Kearns and Star (2013: 84) write: ‘Similarly, we would suggest that the fact that I have no reason not to Φ can be a reason to Φ in virtue of the fact that I have promised to Φ.’ Thus, according to this response, as I understand it, it can be plausibly maintained that the fact that I have no reason not to F is not a reason for me to F, only if it can be shown that this latter fact (i.e. that I have no reason not to F) is relevantly different from the fact that Bob has no reason not to take anti-depressants. However, according to this line of thought, it hasn’t been shown that the two are relevantly different. The parallel here is supposed to go as follows: as that Bob gets depressed when he has no reason not to take anti-depressants explains that the fact that Bob doesn’t have a reason not to take anti-depressants is a reason for Bob to take anti-depressants, the fact that I have promised to F also explains that that I have no reason not to F is a reason for me to F (cf. Kearns and Star 2013).

However, this line of response is unsatisfactory. The major problem here just is that the rejection of the parallel between the anti-depressants case and the promise case is theoretically well motivated. Indeed, the two cases are relevantly different. The fact that Bob gets depressed when he realises that he has no reason not to take anti-depressants (ceteris paribus) explains why Bob ought to take anti-depressants. And this (i.e. the fact about this deontic explanation) is fundamentally why that Bob has no reason not to take anti-depressants is a reason (in a sense) for Bob to take anti-depressants. Nothing similar happens in the promise case. It is simply not a fact that I promise to F when I realise that I have no reason not to F. Contrary to the case of being depressed and realising that one has no reason not to take anti-depressants, there is no connection whatsoever in this latter case between me realising that I have no reason not to F and me promising to F. I may well realise that I have no reason not to F but never even think about promising to F. So, it simply cannot be the case that the fact that I promise to F when I realise that I have no reason not to F (ceteris paribus) explains why I ought to F. It cannot be the case because the left-hand side of the purported explanation is plainly false: it is not a fact that I promise to F when I realise that I have no reason not to F. That I have no reason not to F in the promise case is not a normative reason for me to F. It is crucially different from the fact that Bob has no reason not to take anti-depressants. Despite having the ‘there is/S has no reason not to F’ surface form, the fact that Bob has no reason not to take anti-depressants is not a mere enabling condition. It does speak in favour of Bob taking anti-depressants. Plausibly it does so in virtue of being part of an explanation of why Bob ought to take anti-depressants (together with the fact that Bob gets depressed when he realises that he has no reason not to take anti-depressants). No such deontic explanation is available in the promise case. That I have no reason not to F doesn’t speak in favour of me F-ing in the promise case. Plausibly, this is so because it is not part of an explanation of why I ought to F. Thus we can conclude that the response from Kearns and Star to the objection from enabling conditions is misguided. The Evidence view does seem unable to respect the plausible distinction between mere enabling conditions and normative reasons.

Moreover, note that the discussion here has been centred only on one sort of enabling conditions. As we have learned from Dancy, there are various other sorts of enabling conditions. And these may constitute further potential counterexamples to the Evidence view. Think for instance of the fact that a promise was not made under duress. In certain contexts this fact may well constitute evidence that I ought to respect the promised thing. However, it would be odd to accept that the mere fact that the promise was not made under duress is a reason for me to respect the promised thing. Or think of ability considerations. That I am able to F may well constitute an enabling condition for a fitting F-ing, and it may also increase the probability of the claim that I ought to F. Yet it is a considerable cost for a theory that entails that the mere fact that one is able to F is a normative reason for one to F.

Undercutting defeaters. Until now our critical discussion has been focused on the objections to the right-to-left direction of the bi-conditional of the Evidence view (i.e. the claim that if X is evidence that S ought to F, then X is a reason for S to F). Objections to this part of the Evidence view have constituted the majority of objections to the Evidence view in the literature and we have presented only some of the most promising ones. Let us now turn briefly to the worries that arise for the left-to-right part of the Evidence view, i.e. the claim that if X is a reason for S to F, then X is evidence that S ought to F. The first line of objection to this part of the view appeals to the existence of undercutting defeaters. Some considerations might function as a defeater for a piece of evidence that one has for the proposition that one ought to F, without defeating one’s reason to F. Consider the following case that may be taken to illustrate this sort of possibility. You see a toddler drowning in front of you in a shallow pond as you go through a park. That the toddler is drowning is clearly a reason for you to jump into the pond. It is also at this point evidence that you ought to jump in. But now, here is a twist. As you prepare to jump in, an employee of the park approaches you and says: ‘Oh, please, don’t! It’s not a child. It’s just a hologram of a child. We are conducting a social experiment here, measuring the percentage of people ready to jump in when it appears to them that a child is drowning. Thanks for your participation and sorry for any distress caused’. Arguably, this new information functions as an undercutting defeater for you. That a child is drowning in front of you is no longer evidence for you that you ought to jump into the pond. Now, the twist is that the employee didn’t tell the truth (we can fill in the details in more or less evil ways: the employee is a sadistic serial killer, or alternatively they are indeed conducting a social experiment, but as it happens by an incredible and unlucky turn of events there is a real child in the place of the hologram in the pond today). In such a dramatic scenario, that the child in front of you is drowning is still a reason for you to jump in, but arguably it is not a piece of evidence for you that you ought to jump in. Even more can be said: the fact that you see a child drowning (even if you don’t know that a child is drowning) is a reason for you to jump in, but not a piece of evidence for you that you ought to jump in.

Now, Kearns and Star might reply that what happens in the case of undercutting defeaters is that while the new information defeats one’s evidence – that is, in our case, the testimony from the employee effectively makes it the case that you no longer have evidence that a child is drowning in front of you – there is still a sense in which that the child is drowning in front of you is evidence that you ought to jump into the pond. And it might even be maintained that there is a sense in which that you see a drowning child is evidence that you ought to jump in. It is evidence, according to this line of thought, in a more objective, reliable indicator sense of evidence. Also, they might insist that, given some further specification about the relevant background information, that a child is drowning in front of you does increase the probability that you ought to jump in, even if you don’t know that the child is drowning.

The problem with this move, however, is that it appears arbitrary, or even incompatible, given Kearns and Star’s other theoretical commitments. A major selling point of the Evidence view was that it was said to accommodate perfectly the guidance function of reasons in reliable/good reasoning (see in particular their argument from the role of reasons in deliberation; Kearns and Star 2009: 224–226). In other terms, reasons have a function of guiding one to justified actions and attitudes through reasoning. Now, if the normative reason for you to jump into the pond in our case corresponds to a piece of objective evidence (e.g. reliable indicator), then we may ask how the guidance function of reasons that Kearns and Star have advertised can be really satisfied. It is well known now that the concept of evidence, as we typically understand it both in common usage and in theoretical practice, is associated with various roles or functions. Timothy Williamson identified three crucial functions of our ordinary concept of evidence as (i) the function of ruling out hypotheses that are incompatible with it; (ii) playing a role in inferences to best explanation; and (iii) playing a role in probabilistic reasoning (see Williamson 2000: 194–207). Thomas Kelly (2016) goes a step further and suggests that the four main functions that we associate with evidence stand in tension. According to Kelly, it is doubtful that any single sort of thing could play all of the roles that we standardly associate with evidence. In particular, and of most interest for our present discussion, Kelly argues that one and the same sort of thing cannot always play the role of justifying evidence and at the same time the role of reliable objective indicator evidence. The relevant point for us is that Kearns and Star do insist on the justificatory function of evidence in the guidance of reasoning; this is what allows them to claim that their view accounts for the guidance role in reasoning of reasons. But if Kelly’s suggestions are on the right track, this would rule out, on the pain of accepting an incoherent conception of evidence and reasons, that reasons on their account can also correspond to the more objective, reliable indicator sense of evidence. In short, insisting that in the pond case the relevant unknown fact that there is a child drowning in the pond is still evidence that you ought to jump in stands in tension with the professed claim of Kearns and Star that their account of reasons fully respects the guidance in good reasoning, that is, the justifying function of reasons.⁸

Moore-paradoxical considerations and self-undermining beliefs. Another set of worries that the Evidence view shares with the Reasoning view of reasons comes from Moore-paradoxical considerations and self-undermining beliefs. We have seen above (cf. Chapter 2) that these cases appear problematic for the views that define reasons as (fitting) premises in good/fitting patterns of reasoning. Here is how these cases also cause trouble for the Evidence view. Consider first the following Moore-paradox-style affirmation: (m) ‘the building is on fire but John doesn’t believe that the building is on fire’. We argued above (cf. Chapter 2) that considerations of the form (m) can be reasonably taken to be a reason for one to F in a sense. In this case, (m) can be reasonably taken to be a reason for John to investigate/check/reconsider etc. the state of the building (more precisely, consider the hypothesis (h) ‘the building is on fire’). It does appear to speak in favour of John’s checking the state of the building. Here, I would like to suggest that (m) is not evidence for John that he ought to check the state of the building (e.g. reconsider the hypothesis h). More specifically, (m) is not evidence that John ought to check the building, if we endorse Kearns and Star’s proposal that a central function of evidence is to guide one’s reasoning towards justified F-ing. The following quotation sums up that apparent commitment:

Premise (4) [of the Deliberation Argument] says that evidence that an agent ought to φ can help this agent conclude that she ought to φ. The plausibility of this idea stems from the very notion of what it is for a fact to be evidence for something. We use evidence precisely to work out which propositions are true. If a fact is evidence that one ought to φ, then such a fact is able to help an agent conclude that she ought to φ. […] If the agent is reasoning well, she can use this fact to conclude that she ought to φ on those occasions she ought to φ.

(Kearns and Star 2009: 225)

They also write:

We conclude, then, that reasons to φ are evidence that one ought to φ and that evidence that one ought to φ is a reason to φ. This is equivalent to RA. Because a fact’s being able to play a certain important role in practical reasoning is both necessary and sufficient for this fact to be a reason and for it to be evidence, we conclude that reasons are evidence.

(Kearns and Star 2009: 226)

Elsewhere they also commit themselves to the specific characterisation of evidence as facts that can be knowable:

More plausibly, a fact can be evidence only if it is knowable. There is evidence that P only if someone is able to have this evidence.

(Kearns and Star 2008: 52)

I take these passages to constitute sufficient evidence for the claim that Kearns and Star are committed to the understanding of evidence according to which a consideration that is a piece of evidence for a subject has to be able to play a role in one’s practical reasoning, and, in particular, it has to be able to guide one’s practical reasoning towards a justified F-ing (e.g. one has to be able to F on the basis of evidence and be able to do it justifiably).

And this sort of thing, evidence as knowable and being able to guide us in the specific way, is precisely what we don’t have in Moore-paradoxical cases. The Moore-paradoxical considerations are not able to guide one’s reasoning towards justified F-ing. One cannot F on the basis of the Moore-paradoxical considerations. At the heart of the paradox lies the very fact that the subject of Moore-paradoxical beliefs cannot know the Moore-paradoxical consideration. John cannot know [that the building is on fire but he doesn’t believe that the building is on fire]. Knowing one of the conjuncts of ‘the building is on fire but John doesn’t believe that the building is on fire’ would undermine the truth of the other. Thus, given Kearns and Star’s understanding of evidence, Moore-paradoxical considerations cannot be evidence for one that one ought to F, and yet, plausibly enough, they are still reasons in a sense for one to F.

Similar considerations apply for self-undermining beliefs. Consider: (d) ‘I just took a drug that erased all my memories of the last 5 minutes’. It seems that (d) can well be a reason for me to suspend judgment about what I did during the last 5 minutes. But, again, I cannot use (d) in my reasoning towards suspending judgment about what I did during the last 5 minutes. It seems that (d) cannot play a role in reasoning that Kearns and Star expect evidence to be able to play. Hence, (d) seems to be a reason for me to suspend judgment without being evidence for me that I ought to suspend judgment about what I did during the last 5 minutes. Thus we have identified another sort of case where pace the Evidence view, a consideration can be a reason to F without being evidence that one ought to F.

These critical considerations lead us to a somewhat more general observation about the Evidence view, an observation that points to a more fundamental flaw within this approach. The Evidence view faces many if not all of the problems of the Reasoning view of reasons that we have discussed above (and some additional ones as well). A natural thought at this point is that the Evidence view just is a variation of the Reasoning view. This interpretation is supported by the above observation about the central aspects of evidence according to Kearns and Star. It is implicit in much of Kearns and Star’s theorising. They clearly think that an important and good thing about their view is that it can account for the intuitive feature of normative reasons in guiding one’s deliberation/reasoning towards an overall-ought (Kearns and Star 2009: 224–226, 2013). Thus a reasonable interpretation of the Evidence view is that it is a version of the general Reasoning approach to reasons.

Daniel Star (2018) actually explores the topic of the compatibility of the Evidence view with Jonathan Way’s version of the Reasoning view. Star thinks that the Evidence view is superior to the Reasoning view, given that it appeals to a more easily graspable, more robust notion of ought, as compared to the notion of fittingness that figures centrally in Way’s version of the Reasoning view. Note, however, that the purported difference in this respect might be more artificial than Star seems to think it is. For fittingness just is, as we saw above, the property of F-ing that complies with an ought. Thus a fitting action just is an action that one ought to do. If so, the difference between the Evidence view and the fittingness version of the Reasoning view really seems to be a minimal one, if any. Now, realising that the Evidence view just is a version of the Reasoning view allows us to see a central flaw in the view. As is the case with the Reasoning approach in general, the Evidence view focuses on one of the functions of a common-sense notion of being a reason to act or have an attitude, at the expense of another crucial function of our common-sense notion of a reason to act/have an attitude. It focuses on the role of reasons in reasoning and guiding towards justified F-ing at the expense of the role of reasons in making F-ing fitting or more specifically explaining why one ought to F.

At any rate, even if my suggestion that the Evidence view reduces to a version of the Reasoning view is wrongheaded, the above objections stand. And many of the same problems that the Reasoning view faced also apply to the Evidence view – which should lead us to investigate the prospects of alternative options in our attempt to define normative reasons.

4.5 Concluding Remarks

In this chapter we have seen another major approach within the recent debates about the nature of normative reasons. We have presented Kearns and Star’s influential Evidence view of reasons, according to which a fact is a reason for one to F if and only if that fact is evidence for one that one ought to F. We rehearsed some of the main positive arguments in favour of the view. And we then spent some time exploring the pitfalls of this original proposal. At the end of the day we concluded that it is reasonable to see the Evidence view just as another version of the Reasoning view of reasons. Thus the substantial worries that we identified with the Reasoning view re-surface with respect to the Evidence view as well.

Can someone sympathetic to the Evidence view try to provide a different version of a view that maintains the insights of thinking about reasons in terms of evidence but still avoids its problems? I haven’t shown that this cannot be done. And indeed some recent proposals might be more promising in this respect than Kearns and Star’s original view (see, for instance, the view presented and defended in Whiting 2018). Unfortunately, however, a full-blown analysis of further variations of the Evidence view lies beyond the framework of the present discussion. At this point, I would like only to suggest that, given the poor track record of existing monist reductionist views, the views that appear to be close to either a version of the Reasoning view or a version of explanationist views might not be the most promising alternative to explore. I would like to suggest that we rather focus on a radically different proposal, a view that doesn’t commit itself to the idea that there is only one sort of normative reasons and that it should be understood either to be more like a [fitting] premise in good reasoning or more like an explanation of why one ought to F/why it would be good for one to F. Rather, the alternative, towards which I would like to turn now, combines the best insights from both of these general views without inheriting their respective pitfalls. Without further ado let us now turn to our positive proposal about normative reasons.