Abstract
This paper defends the view that in a certain sense evidence is normative. Neither a bit of evidence nor the fact that it is evidence for a certain proposition is a normative fact, but it is still the case that evidence provides normative reason for belief. An argument for the main thesis will be presented. It will rely on evidentialist norms of belief and a Broomean conception of normative reasons. Two important objections will be discussed, one from A. Steglich-Petersen on whether having evidence is sufficient for having a normative reason for belief and another one from S. Street.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
This paper is about the normativity of evidence. In order to state the main thesis and argument, however, a lot of unpacking is needed. We need to clarify what ‘normative facts’ are and what it means to ‘provide normative reasons’ for belief. This is the main purpose of this first section. To jump ahead, and very crudely, the main thesis can be stated as follows: neither a bit of evidence nor the fact that it is evidence for a certain proposition is a normative fact, but it is still the case that evidence provides normative reason for belief. In this latter sense, then, evidence is normative. Once the unpacking is done, an argument for the main thesis will be presented (section 2). It will rely on a pair of evidentialist norms of belief (a retrospective and a prospective one) and a Broomean conception of normative reasons. The third and last section will take a look at two important objections, one extracted from the writings of A. Steglich-Petersen and another one from Sharon Street’s work. The objections will be shown to fail, and in the course of the discussion both the proposed view and its advantages hopefully will become even clearer.
First, we need to do the unpacking. When asking the question whether evidence is normative the first fundamental point to note is that the claim that evidence is normative is ambiguous. It might mean that facts that are evidence or facts of the form that so and so is evidence for p (for subject S), where so and so is a fact, are normative facts. Call facts of this form ‘evidential facts’. And following common terminology, call the embedded fact, that so and so is the case, ‘evidence’ or a ‘bit of evidence’. So the first understanding of the claim is that bits of evidence or evidential facts are normative facts. Alternatively, the claim that evidence is normative might mean that bits of evidence and/or evidential facts are or provide normative reasons (for S to believe that p).Footnote 1 These two kinds of claim are very different. One could hold that evidence is normative in the sense of providing normative reasons for belief without holding that bits of evidence or evidential facts are normative facts. Indeed, this is what I want to propose and argue in this paper. We can get what we want and need by having bits of evidence and evidential facts provide normative reasons without having to think of them as normative facts themselves.Footnote 2
What is it for a fact to be normative? – We can settle this question roughly in the way Kurt Sylvan has proposed.Footnote 3 Normative facts are paradigmatically normative facts or facts that are partly or completely grounded in paradigmatically normative facts, or facts that are relevantly similar to the paradigmatic normative facts (even though not paradigmatic) or partly or completely grounded in facts of the latter kind. Paradigmatically normative facts are deontic facts like ought facts, permission facts, normative reasons facts,Footnote 4 etc., or normative evaluative facts like goodness simpliciter facts, or hypological facts like accountability facts, blameworthiness facts, etc. Here, ‘normative reasons facts’ are (complex) facts of the following form: there is some fact – usually called ‘the normative reason’ – and this fact is a normative reason (for subject S) for behaving in a certain way. In the following, we can focus on the deontic facts, since the evaluative and hypological normative facts will not be relevant. So the question is whether bits of evidence and evidential facts are grounded in normative facts or are relevantly similar to paradigmatic normative facts or grounded in these facts. I will argue for a negative answer to this question: the facts that are evidence and the evidential facts are not normative facts, neither paradigmatic nor non-paradigmatic.Footnote 5
It is interesting to note that similar questions arise for other states or standings, such as knowledge or warrant, for example. Is knowledge normative? Some epistemologist think it is, others deny that knowledge is normative. The issue crucially hinges on whether knowledge is constituted by (grounded in) some normative item, and presumably justification and warrant are the most relevant candidates here.Footnote 6 The question whether evidence is normative joins the other ones about the normative status of knowledge and warrant. So a further goal of the investigation is to study and clarify the general difference between the normative and the non-normative. In addition, it is highly relevant to any investigation into epistemic-moral/practical analogies or disparities (which I will not go into here).Footnote 7
Which facts are the non-normative bits of evidence and evidential facts? – I think it is quite natural to take bits of evidence and evidential facts to consist of ordinary descriptive facts plus causal-informational facts, sensitivity or safety facts (modal facts),Footnote 8 or probabilistic facts, or combinations thereof.Footnote 9 In a way, this is what we start with. We call spots on the skin evidence for the fact that the person has measles; we take traces on the ground to be evidence for the recent presence of some kind of animal; and we take the sounding bell as evidence for someone’s standing at the door. Whatever these facts exactly are, they belong to the non-normative causal, informational, modal, and probabilistic facts. Bits of evidence and evidential facts are made of these non-normative facts. They are not grounded in paradigmatic normative facts, and neither are they grounded in facts that are relevantly similar to paradigmatic normative facts. John Skorupski calls these facts ‘indicative evidence’, and he sets them apart from epistemic reasons facts, which are normative facts. “Indicative evidence that p consists of facts that indicate that p”, he writes.Footnote 10 And for a fact to indicate something requires no more than “a productive connection, deterministic, or propensity-based”.Footnote 11 Leaving aside possible quibbles about what exactly the relation of indication is, it is clearly a non-normative one. Thus we arrive at a notion of evidence that does not make the bits of evidence and the evidential facts into normative facts. And so we can raise the question of what turns the water of evidence/evidential facts into the wine of normative reasons for belief.
Evidence can indeed give rise to something normative. Even though bits of evidence and evidential facts are not normative facts they still provide normative reasons facts. What does ‘providing …’ mean here? – This question for clarification leads us to the second fundamental point of this paper. The appropriate discussion of normative issues will be lead in terms of grounding. So ‘providing normative reasons’ will be taken as grounding normative reasons facts. The idea that some facts are more fundamental than others and dependent on the more fundamental ones is an old one. Euthyphro questions already invoke the idea of grounding. By now, grounding has become one of the major tools for metaphysical analysis, and the present paper happily subscribes to the grounding approach.Footnote 12 It will pay, as will become clear in due time, I believe.Footnote 13
An interesting debate then arises as to what kind of fact bits of evidence are. Are they (always) psychological facts or (mostly) non-psychological facts? This is what the debate of psychologism versus anti-psychologism about evidence is concerned with. I will not take a stand on this debate here. But of course, some paradigmatic evidential facts – like the ones already mentioned: that traces on the ground indicate some animal, that spots on the skin indicate measles, and that the bell’s sounding indicates that someone is at the door – are commonly thought of as non-psychological facts. In order to not complicate the discussion too much (by moving from these facts to the corresponding psychological facts that friends of psychologism will take recourse to) we will occasionally use some paradigmatic non-psychological examples of evidential facts. This should be taken as non-committal since a suitable reformulation is always intended to be available.Footnote 14
Let us now state explicitly our first fundamental point mentioned above in terms of this regimented terminology. We need to distinguish between the following two claims:
-
(1)
Bits of evidence and evidential facts provide normative reasons facts concerning belief.
-
(2)
Bits of evidence or evidential facts are normative facts.
The goal will be to argue that (1) is true and (2) is false.
Alternative views endorse (2). Typically, they will take the bits of evidence to be non-normative facts and take the evidential facts to be normative facts. These are ‘normativist views about evidence’, as we can call them. Some Bayesians will endorse (2).Footnote 15 And one might think of an epistemic reasons fundamentalism that takes evidential facts to be irreducible normative facts. (Some versions of Bayesianism may be instances of such a reasons fundamentalism.) Of course, if (2) is endorsed claim (1) (or something even stronger) will follow more or less immediately, since evidential facts will be normative reasons facts.Footnote 16
I will present and argue for a view that rejects (2) and endorses (1). Neither bits of evidence nor evidential facts are normative facts, but they still provide normative reasons facts. So, in a sense I will make it harder for myself to provide a normative standing for evidence in epistemology. She who straightforwardly endorses the normativity of evidential facts has an easier way. She can simply identify evidential facts and normative reasons facts (concerning belief). In short, she can say that the evidence-for relation is the normative-reason-for relation (in the domain of belief). That provides evidence with a normative standing, of course, and the strongest it could have, I take it. My view too will provide a normative role or standing for evidence, but a weaker one. Bits of evidence and evidential facts are not normative facts; they ground normative reasons facts about belief. Bits of evidence are normative reasons for belief (and not just normatively relevant in some way or other).
In any case, it is important to become clear about the flow of grounding in epistemology. (And elsewhere too.) Types of ground can be used to determine what ‘source’ some normative facts have.Footnote 17 Normative facts in epistemology might be grounded in normative facts of some other normative domain, such as the domain of moral facts. This would constitute an ‘inter-domain grounding’, and a view that takes the epistemic value facts (facts of true belief, e.g.) to be grounded in some moral facts would be an example of such an inter-domain reduction.Footnote 18 Epistemic normative facts might also be grounded in non-normative facts, i.e., descriptive or ‘positive’ facts. This would constitute an ‘external grounding’, as I would like to call it, since it would ground the normative in something external to the normative. External grounding amounts to naturalization. Of course, one could combine an inter-domain reduction with an external reduction. What is most important is to be clear about which way the grounding flows. The flow of grounding is what structures and shapes the discussion of normativity in epistemology (and elsewhere). This is the upshot of the second fundamental point mentioned above.
To complete the picture, we need to acknowledge a third kind of grounding, the ‘intra-domain grounding’. Some normative facts within a given domain, say epistemology, might be grounded in other normative facts within the same domain. Correctness facts (that a certain belief is correct) might be grounded in normative reasons facts, for example. It is plausible that we can identify a certain grounding pattern within a normative domain like epistemology. It will display the normative facts that are not (fully) grounded in other normative facts of that domain, i.e., the relatively fundamental normative facts of the domain. It is important, however, to not mistake the relative fundamentality for an absolute fundamantality, for relatively fundamental normative facts in a domain might very well be externally grounded (directly or indirectly, i.e., via some inter-domain grounding) or grounded in some other normative domain. The view that will be presented here will be a norm-first view in epistemology in the sense of taking two general normative norm facts – a pair of evidentialist norms of belief – as the only relatively fundamental normative facts in the epistemic domain.Footnote 19 Is this general norm fact absolutely fundamental? – I favor the view that it can be naturalized, and so is not absolutely fundamental but externally grounded, namely, in teleofunctionalist facts. Naturalization would remove doubts about the existence of normativity engendered by queerness considerations pushing for error theory. Clearly, this is a further topic and it shall not to be argued here.Footnote 20 What will be argued here is that a norm-first view, with a pair of evidentialist norms of belief, can vindicate the idea that evidence and evidential facts provide normative reasons facts without being themselves normative facts.
2 The Main Argument for (1)
In this section, an argument for claim (1), that evidence and evidential facts provide normative reasons for belief, will be given. It comes in two parts. The first part is argument A1, the ‘argument from the Broomean conception of normative reasons’. This part of the argument exploits the idea that facts that ground deontic facts (like correctness facts or ought facts) are normative reasons. The second part will be argument A2, the ‘argument from the evidentialist norm of belief’. It hinges on an evidentialist conception of what makes a belief correct: being held on the basis of sufficient evidence. Put together, the first pair of arguments tells us that pieces of (sufficient) evidence are normative reasons because they (partly) ground the correctness of a belief which consists in its being held on the basis of sufficient evidence.
This first pair of arguments is concerned with the deontic status of correctness and the case in which the subject already holds a belief. (This is the ‘retrospective case’.) In fact, as we will see soon, there is a second pair of arguments, A1’ and A2’, with a fully parallel structure, concerning the deontic status of ought and cases in which the subject does not have the relevant belief but ought to form it. (This is the ‘prospective case’.) Let us begin with the former one, concerning the correctness of beliefs already held. An intensive discussion of this argument will then pave the way for a much quicker discussion of the second pair of arguments concerning ought and the prospective case.Footnote 21
-
(A1):
-
(i)
The Broomean conception of normative reasons: a normative reason is what (partly or completely) grounds a deontic status of correctness.Footnote 22
-
(ii)
Bits of evidence (partly) ground the correctness of beliefs.
------.
-
(iii)
Bits of evidence are normative reasons.Footnote 23
Let us clarify the premises of the argument.
A crucial ingredient is the Broomean conception of normative reasons, as spelled out in premise (i). Why rely on it? There are two main alternative views about what normative reasons are.Footnote 24 First, one could hold that normative reasons facts are primitive and cannot be explained, are not grounded in anything else. They favor responses, but this is not to be understood as an informative analysis or account. This view is the reasons fundamentalism made popular by Scanlon and Parfit, for example. It has the obvious cost of being explanatorily inferior to any account that can explain normative reasons facts (ceteris paribus, at least). Second, one could hold that normative reasons are the starting points of good or correct (theoretical) reasoning and so normative reasons facts are grounded in facts of good or correct reasoning.Footnote 25 Some version or other of this view is held by Kieran Setiya, Daniel Whiting, Jonathan Way, and Stephen Kearns and Daniel Star, but it also comes with significant costs: the goodness or correctness of reasoning and attitudes is assumed and there is a problem with undercutting defeaters (even though perhaps not with rebutting defeaters).Footnote 26 If the goodness or correctness of theoretical reasoning can be explained solely in terms of the correctness of beliefs, as Way envisages, the costs go down. (Facts of correctness – fittingness – of beliefs are then the relatively fundamental normative facts in epistemology.)
Here I cannot go into a detailed investigation of the merits and shortcomings of these two alternative views. I have indicated some costs, but more can and should be said. The focus of this paper is to put forward a positive alternative view and to show its attractions. If the view proves to be attractive, it is thereby justified to some extent. So let us move on to the second premise.Footnote 27
The second premise raises the obvious question of how evidence grounds correctness. Here the second part of the argument comes in, the ‘argument from the evidentialist norm of belief’ which tells us how bits of evidence plus a general norm fact ground correctness of belief. (We will later see that the norm which is called ‘evidentialist norm of belief’ here is just one of the two evidentialist norms of belief, the ‘retrospective’ evidentialist norm of belief.)Footnote 28
-
(A2):
-
(iv)
Evidence and evidential fact: That X is a fact and the fact that X is sufficient evidence for p (for subject A).
-
(v)
Belief based on evidence: A holds Bp on the basis of the fact that X (S believes p for the reason that X).
-
(vi)
Evidentialist norm of belief (retrospective):
For any S and p, S’s Bp is correct iff S holds Bp on the basis of sufficientFootnote 29 evidence for p.
------ (ground).
-
(vii)
A’s Bp is correct.
(A2) tells us how pieces of evidence and evidential facts ground correctness facts. They do so by fulfilling the conditions that the evidentialist norm of belief (vi) relates to correctness. Thus, the flow of grounding is clarified. We have evidence that the subject possesses and responds to and the general normative principle, the norm of belief, and together they ground correctness facts. No non-epistemic normative facts are involved (i.e., no inter-domain grounding takes place). Epistemology is portrayed as normatively autonomous and epistemic normativity is partly grounded in the descriptive – a quite transparent and perspicuous view we end up with, I believe.
Premise (vi) of (A2) expresses a commitment to evidentialism, of course. Evidentialism should not be understood as a claim about what epistemic reasons there are, but as a normative first-order claim, as (vi) has it.Footnote 30 To defend evidentialism would go beyond the scope of this paper. Suffice it to mention that Selim Berker has recently provided an interesting critique of certain kinds of reliabilism that point to the lesson that merely promoting true belief cannot be what grounds epistemic normativity.Footnote 31 Incidentally, Berker sees evidentialism as a promising alternative. In my view, evidentialism can be understood as a special form of reliabilism, one that prescribes a special way of arriving at true beliefs, namely, the evidential way (following one’s evidence).Footnote 32 For that you need the assumption that ‘sufficient evidence’ is truth-guaranteeing (in close worlds).Footnote 33 So epistemic normativity is about both, forming true beliefs (and avoiding false ones) and following one’s evidence. This yields a kind of perspectivism about epistemic normativity since not only truth matters, but also one’s perspective (in the form of mental states that constitute possession of evidence). It is a form of ‘objective evidentialism’.
If in (A2) the evidentialist norm of belief is replaced by the truth-norm, evidence vanishes from the picture, and thus bits of evidence would no longer count as normative reasons.Footnote 34 This comes as no surprise. It is quite clear that only on an evidentialist view evidence will be normative. If evidence played no role for justification and correctness, one could not see how it should be normative. The defense of the normativity of evidence (in the sense of claim (1)) rests on adopting some form of evidentialism.
Premise (iv) of (A2) mentions evidence and evidential facts. Each of them is a partial ground of correctness, the one being more encompassing than the other. This raises the question which of them should be taken to be a normative reason for belief, the bits of evidence or the evidential facts or both? – In principle, both could be normative reasons. It depends on what it is for which the subject believes that p. In many ordinary cases, we believe things simply on the basis of evidence, and not on the basis of evidential facts. This is then how the argument proceeds. Premise (v) characterizes this kind of case. And in this case the bits of evidence are normative reasons – the normative reasons for which the subject holds the belief that p. But it is possible to believe for an evidential fact as well. This is a more sophisticated case since the subject must now possess the evidential fact, and the evidential fact comprises the relation of being evidence for (which is not required if the subject believes simply for the facts that are evidence). Then the evidential facts are the normative reasons – the normative reasons for which the subject holds the belief. Either way is fine for the present purposes. There is no real conflict or competition, just two different (but still saliently similar) possibilities. Because the first case is simpler, I will choose it for the discussion in the following. As long as at least one of the two possibilities is accepted, the (corresponding) argument goes through. (Of course, (A1) could be modified correspondingly, letting evidential facts take the place of bits of evidence.)
(A2) may give rise to the following worry. Does the fact that the subject has to possess the evidence, and has to follow it (i.e., believe that p on the basis of the evidence), pose a problem? Doesn’t this smuggle in a further normative fact? – This worry brings up an interesting issue. We can acknowledge that, indeed, if following one’s evidence is (partly) grounded in a normative fact, a further normative ingredient is relied on. The question is whether following one’s evidence is in fact a normative matter. One reason for thinking so might come from virtue epistemology, and it has recently been articulated by Ernie Sosa and Kurt Sylvan.Footnote 35 It is simply the idea that following one’s evidence requires the proper exercise of an epistemic virtue (competence) and that epistemic virtue facts are normative facts. After all, it is clear that one might hold the view that epistemic virtue facts are among the relatively fundamental epistemic normative facts. Indeed, this would be what makes the view a true virtue view. And so there is a challenge for the proposed view, namely, to argue that no normative epistemic virtue facts are needed.
To address and properly rebut this worry would probably require a rather lengthy investigation. But for the present purposes, the following reply seems to suffice. We can acknowledge that if the worry is correct, a further normative ingredient is needed. Two things, however, can be added. First, it might be the case that this further normative ingredient is itself grounded in non-normative facts (abilities, understood as non-normative dispositional facts) plus the norm of belief. For it is a plausible idea to think that what makes certain dispositions – and many virtue epistemologists, like Ernie Sosa for example, think of epistemic virtues as dispositions – virtues is the fact that they make us form beliefs in accordance with the norm of belief. And so epistemic virtue facts could be grounded in dispositional facts plus the norm of belief. Then the further normative ingredient is not ungrounded in the epistemic normative domain, not relatively fundamental. Therefore, the view would remain a norm-first view, and there is no objection to (A2) left over. Second, even if this reduction did not work, evidential facts would still be providing normative reasons since they would still be partial grounds of normative reasons for belief. They are not full grounds. But that has already been assumed, due to the appeal to the norm of belief. The view would thus be a norm-plus-virtue-first view. Evidential facts would still play the normative role of providing normative reasons (with the help of the norm and some virtue facts). So even in case virtue facts cannot be grounded suitably in non-normative facts and the norm of belief we can still maintain claim (1) and reject claim (2). (A2) would not be disproved but would mean something different: non-normative evidential facts plus relatively fundamental normative virtue facts plus the norm of belief ground correctness of beliefs. Because the core of the view presented in this paper would not be touched, I will from now on ignore the virtue facts.
We can conclude by summarizing the first pair of arguments of this section, comprising A1 and A2, as follows. The argument concerns the case of beliefs that are already held and their correctness. Since pieces of evidence are (partial) grounds of correctness, they are normative reasons and they provide normative reasons facts. At the same time, bits of evidence are not normative facts. The norm of belief is also a further (partial) ground of correctness. The evidential fact (comprising the evidence) plus the fact of the subject’s following it plus the norm of belief are the full ground of correctness. No normative facts belonging to another normative domain (such as the moral facts) enters into the determination of epistemic correctness. Epistemic normativity thus is autonomous (even though grounded in descriptive facts, at least partly).
Now, let us quickly look at the second pair of arguments which is parallel to the first one. It concerns the case of subjects who have sufficient evidence for p but do not hold the belief that p (at the same time). Intuitively, they ought to believe it by following their evidence (supposing that the question whether p is properly activated for them and they have all the abilities needed for following the evidence).Footnote 36 We need to introduce a second norm since the correctness norm does not cover the cases where the belief has not (yet) been formed.Footnote 37 The productive and the retrospective norm should of course dovetail, but their content guarantees that no conflict arises.
The parallel argument turns around the ought for belief – and thus is closer to Broome’s view. We can quickly write it down by replacing the relevant parts. It has two parts, again.
-
(A1’):
-
(a)
The Broomean conception of normative reasons: a normative reason is what (partly or completely) grounds the fact that someone ought to believe that p.
-
(b)
Evidential facts (partly) ground the fact that someone ought to believe that p.
------.
-
(c)
(c) Evidential facts are normative reasons.
The second part brings in the second norm of belief, a ‘productive’ or ‘prospective’ norm. We can call the first norm – norm (vi), the evidentialist norm of belief – a ‘retospective’ norm, correspondingly.Footnote 38 The second part now reads like this:
-
(A2’):
-
(d)
Evidence and evidential fact plus possession (plus activation and abilities):
F is a fact that is sufficient evidence for p (for A).
-
(e)
Possession of evidence (plus activation and abilities) and lack of belief:
A possesses F (and the question whether p is activated for S and S has all the abilities needed for following F) and A does not believe that p.
-
(f)
Productive evidentialist norm of belief:
For any S, if there is a fact, X, which is sufficient evidence for p (for S) and S possesses X (and the question whether p is activated for S and S has all the abilities needed for following X), then S ought to form the belief that p on the basis of her sufficient evidence for p (i.e., on the basis of the fact X).Footnote 39
----- (ground).
-
(g)
A ought to believe p on the basis of F.
In other words, having sufficient evidence for p brings the subject under an ‘obligation’ in the sense of an ought fact; she ought to believe that p on the basis of her sufficient evidence. The ought is of course a normative ought. Again, a norm-first view arises. And we can again ask whether following one’s evidence smuggles in some relatively fundamental virtue facts. The reply will be the same, mutatis mutandis.Footnote 40
3 Two Objections
3.1 An Objection from the Corner of Teleology (Steglich-Petersen): The Insufficiency Argument
In this subsection, I will discuss a possible argument against the proposed view. It is taken from materials to be found in the writings of Steglich-Petersen.Footnote 41 I hesitate to attribute the argument to him, however. The idea seems to be somewhat present, but it is not clearly or explicitly expressed in the text. But we can leave aside the exegetical issue and look at the systematic possibility of providing an argument while being inspired by what Steglich-Petersen has written.
The argument turns on the idea that evidential facts alone are not sufficient for providing normative reasons facts. We can state it as follows:
The insufficiency argument:
Having (sufficient) evidence for p alone is not sufficient for having a normative reason to believe p. For, one has a normative reason to believe p only if in addition one has a normative reason to form a true belief about p at all.
So evidential facts alone do not provide normative reasons for beliefs;
some other normative reason (not constituted just by evidence or evidential facts) must come into play and provide the normative reason for believing p.
The picture then is this. The right principle about normative reasons has to appeal to reasons to form true beliefs. It reads something like this:
(BA) Iff, and because, S has a normative reason to form a belief about whether p is true,
it is the case that if S has sufficient evidence for p S has a normative reason to believe that p.
As a next step, we can envisage the grounding of reasons to form a true belief about whether p in some value of forming a true belief about whether p. This would bring in a form of teleology, and perhaps an improved form of teleology that allows us to deal with some problems of classical teleology (having to do with the well-known difficulty that not all true beliefs seem to be valuable).Footnote 42 The second step, however, is not necessary for the insufficiency argument and is no part of it, and so I will ignore it.Footnote 43
The insufficiency argument faces a decisive problem. Let us put difficulties of understanding what a ‘reason to form a true belief about whether p or not’ really is aside. We might have some good grasp on this notion, let us suppose. Then the first problem is that the distinction between modality and grounding is not taken into account. And once we do take it into account, the argument breaks down. Somewhat ironically, it is in fact the case that evidential facts are modally sufficient conditions for having a normative reason for belief. They are not full grounds of having normative reasons, but this does not prevent them from being partial grounds, of course. That evidential facts are modally sufficient comes from the fact that arguably, the evidentialist norm of belief is metaphysically necessary, and so holds in all metaphysically possible worlds. The only contingent element then is the evidential fact. Once it is present, we have a normative reason for belief. No further contingent normative reason – such as a normative reason to form a true belief about whether p – is needed. The argument simply assumes that it is needed, but it does not present any justification for thinking so. Thus the insufficiency argument fails as an objection against the proposed view.
3.2 Sharon Street’s Objection against Epistemic Normative Realism (in Street 2009)
Sharon Street has argues against normative realism in the epistemic domain.Footnote 44 She distinguishes between two kinds of normative realism (NR): Clifforidan NR (CNR) and normative realism about likelihood (NRL). Only CNR is relevant for the present purposes, since NRL takes the evidential facts to be normative facts (This corresponds to acceptance of claim (2), above.) It is worthwhile to point out, furthermore, that Street explicitly addresses normative realism, i.e., the view that there are objectively (attitude-independently) existing normative facts. So far, objectivity has not been an issue. But of course, it is in the vicinity. If bits of evidence and evidential facts are objective and the norms of belief hold objectively, we arrive at a view that fits the definition of normative realism, Cliffordian style. So Street’s argument is relevant. And in any case, it is interesting to see what Street’s objection against CNR is – if only to see why the proposed view is not affected by it.
CNR then is the following claimFootnote 45:
(CNR) “Facts that make p more likely are normative reasons in favor of belief that p, and facts that make p more unlikely are normative reasons against belief that p, and this is so no matter what an agent’s evaluative attitudes or what follows from within the standpoint constituted by them.” (Street 2009, p. 227)
The notion of ‘making it more likely that …’ is supposed to refer to some descriptive, non-normative relation. For the present purposes it could simply be replaced by the notion of (sufficient) evidence – non-normative bits of evidence - that I have appealed to when charterizing evidential facts. (We can leave it open whether they can be usefully characterized as ‘likelihood facts’). CNR asserts that evidential facts provide normative reasons facts. And so we can say that CNR is a formulation of the idea that evidence is normative à la (1) without (2). (NRL corresponds to the alternative view that accepts (2).)Footnote 46
Now, CNR raises the following two questions, at least:
(Q1) How can CNR be justified? Which arguments can be given in favor of CNR?
(Q2) How can we explain the fact that many people believe CNR (in the sense of taking evidential considerations to be reasons for belief)?
Clearly, these are two distinct questions. How are they related? This is not so easy to say. And Street does not explicitly say how they are related, since she does not bring up explicitly question Q1, somewhat surprisingly.
Street argues as follows against CNR, focusing on question Q2. Our belief in CNR demands an explanation: How come that we have hit the truth about this fact instead of believing one of many alternative false views about reasons? (After all, we could take comic considerations or authority-related considerations or … to be normative reasons for belief.) This is a question from within our standpoint, as Street points out. And so it is legitimate to ask for an explanation. There are, then, three possibilities:
-
(i)
It’s mere chance, sheer coincidence.
-
(ii)
Tracking account: we grasp an objective truth (by means of a special tracking capacity …).
-
(iii)
Adaptive-link account: creatures that take evidential facts to be normative reasons for belief tend to form true beliefs and this is advantageous to their survival and reproduction; as a result, belief in CNR has spread.
In Street’s view, these options can be assessed as follow.
-
(i) is not good, “defies common sense” (Street 2009, p. 234).
-
- (ii) is not scientifically acceptable.
-
- (iii) is good and scientifically acceptable.
-
- (iii) is not available to CNR (Street 2009, p. 234, 236).
The main problem lies with the last claim, the unavailability of the adaptive-link account as an explanation of our belief in CNR. It is false. The adaptive-link account is in fact fully available to the normative realist, and the normative realist is thus not forced to accept the tracking account (given that pleading for mere chance is excluded, as we can grant). In fact, nothing in the adaptive-link account is incompatible with CNR. It is true that the adaptive-link account does not make use of normative facts. (But why should it?) This is no reason to think it incompatible with CNR. And no further reason is given by Street. None is in sight, I submit.
One might think that this leads to the follow-up question: How then does the normative realist justify claim (1) [that evidential facts provides normative reasons facts]? – The simple answer is, by way of an argument like the main argument given above (A1 plus A2).
This invites the next worry, however: This argument presupposes some normative facts, namely, the normative fact of the evidentialist norm of belief! This worry can be developed in two ways then. Either (Wi) this is not scientifically/naturalistically acceptable, or (Wii) this is question begging in the present context, since we are in the business of questioning the idea that evidence provides epistemic normative reasons in general, tout court.
The response to (Wi) is simply that yes, indeed, we should accept that the norm of belief has to be made naturalistically acceptable. A naturalization by means of teleofunctions is the best option here, in my view: the job of our doxastic capacities it so produce true beliefs in response to one’s sufficient evidence. The reply to (Wii) is that there is no question begging. We don’t have to be reasons-firster, we can be norm-firsters and we can explain normative reasons by evidential facts and the norm of belief, and then, in addition, we can externally ground this norm in teleofunctions.Footnote 47 So the only remaining and, indeed, significant question is about the naturalization of the norm of belief (for example, in terms of teleofunctions). But as Street does not address this point, her argument does not defeat CNR.
In sum, the adaptive-link account is available to the normative realist, too. And there are no serious follow-up problems either. A challenge remains, the challenge of providing an external grounding of the norm of belief in descriptive facts, i.e., of naturalizing this norm. Teleofunctionalism might fulfill the task.Footnote 48 Then the project would be completed. But since Street does not say anything that would undermine teleofunctionalism, her argument is not successful.
Notes
We can switch from evidence and the evidential facts that so and so is evidence for p to other, more inclusive facts in the vicinity: facts that so and so is evidence for p for S and S possesses the fact that so and so, or the like. The ambiguity remains.
For the present purposes we can leave it open whether bits of evidence and evidential facts are relative to some background (background facts, background beliefs, background knowledge, or the like).
To be precise, Sylvan does not mention normative reasons in his list of deontic facts, he only mentions “permissibility, rightness, and wrongness” (Sylvan 2018, 193). I take it that normative reasons facts are paradigmatic normative facts, and that this is not controversial. I believe Sylvan will not object. Furthermore, Sylvan uses the term ‘constituted by’ instead of ‘grounded in’. I take it that for the present purposes this does not make a difference.
Evidence and evidential facts are not on the list of paradigmatically normative items. Why not? – Well, it is uncontroverisal that bits of evidence are not normative facts, but it is rather controversial whether evidential facts are normative facts, and the paradigmatic examples should be uncontroversial. I believe that the normativist about evidence should not deny this. The normativist should instead hold that evidential facts are relevantly similar to the paradigmatic normative facts and should provide arguments for this claim. Needless to say that I do not see such arguments.
Among the relevant publications on the epistemic-moral/practical analogy or disparity are Heathwood (2018), Heathwood (2009), Cuneo and Kyriacou (2018), Rowland (2013), Cuneo (2007), and Olson (2018). At some points this debate is intertwined with the vexed open question argument and the analytic-synthetic question. The present paper does not concern this conundrum.
Safety facts are facts of the form ‘in all close worlds where s obtains, p is the case’ (s is a safe indicator of p); sensitivity facts are facts of
Of course, the relevant type of probability is not the epistemic probability that Bayesians speak of, but rather objective probability (in ensembles) and objective chance. Cf. Ismael (2011).
Skorupski (2010), p. 48.
Skorupski (2010), p. 47.
Just to give one example, Brian Epstein has recently analyzed issues of social metaphysics in terms of grounding. Cf. Epstein (2015). In epistemology and ethics, grounding ideas are all over the place. Let me note that of course, there are some serious remaining questions about grounding. Still, it has proved to be very useful for organizing views and arguments to rely on grounding, and this is how I will explore the terrain in this paper too.
Some epistemologists think that evidence does not, or not always, take the form of facts. Entities of other ontological categories might be, or serve as, evidence. This view will be ignored in the present paper, and with some justification, at least. For one, it might be possible to reformulate these views in terms of facts. (Particulars will be replaced by corresponding existence facts, for example.) For another, the existing literature provides good reason to think that evidence takes the form of facts or is ‘propositional’. Cf., e.g., Williamson (2000), Alvarez (2018).
Williamson’s kind of ‘objective Bayesianism’ seems to me to be such a normativist view, too. Cf. Williamson (2000), ch. 10. Normativists about evidence will at some point have to say something about what normative facts they take evidential facts to be grounded in. Alternatively, they might want to extend the list of paradigmatic normative facts by including evidential facts.
Strictly speaking, identity does exclude grounding, and so it is not exactly (1) that follows but a corresponding – stronger – identity claim: evidential facts are normative reasons facts (concerning belief, of course).
A particularly insightful analysis of the sources of practical normativity can be found in Worsnip (2018).
This view is proposed by Grimm (2009).
Cf. Hofmann and Piller (2019)
I believe that some considerations pushing into the direction of naturalizing normative reasons facts that Pekka Väyrynen has presented in Väyrynen (2011) could be taken over to argue for a need for naturalizing a norm of belief.
Here, ‘ought’ and ‘correct’ are of course to be understood in the relevant normative sense, and they are dual notions. That S ought to A means that it is correct that S As and it is incorrect that S does not A. So everything could be reformulated by using just one of these dual notions. For the sake of a more natural reading, avoiding awkward double negations, I am stating things by using both notions. ‘Ought’ also comes more naturally in prospective cases.
I take it that ‘correctness’ and ‘rightness’ mean the same. Here I depart from Broome on which denotic status it is that normative reasons ground. For Broome it is an ought (a normative, owned, unqualified ought), here it is correctness. But since ‘ought’ and ‘correct’ are dual notions, this is not a significant difference. I understand correctness as consisting in faultlessness, the absence of the subject’s committing mistakes. Other philosophers who favor a similar conception are Alvarez (2010: ch. 1), Hyman (2015: ch. 6), and Nebel (2018).
The argument contains a slight sloppiness. More precisely, things should be reformulated in terms of what makes it the case that something is a normative reasons. That x grounds correctness is what makes it the case that it is a normative reasons. This reformulation is less elegant, and so I have chosen to formulate things in the simpler way.
Here I am basically agreeing with Ralph Wedgwood. Cf. Wedgwood (2015).
The account is supposed to govern both practical and theoretical normative reasons. For the present purposes, I focus on the theoretical case.
Cf. Setiya (2007), Whiting (2018), Way (2017), Kearns and Star (2009). The overarching theme is how to come to grips with the defeasibility of reasoning. A reason which is rebutted (outweighed) is still a reason. Then there can still be a good (pattern of) reasoning leading from the rebutted reason to the conclusion. But a reason that is undercut is no longer a reason, though there might very well be a defeasible type of reasoning from it to the conclusion. It is also unclear how to deal with factual defeaters, viz., facts that turn good reasoning into bad reasoning merely thru obtaining and without being believed or detected by the subject.
Please note that all we rely on here is only that part of the Broomean conception that is concerned with what he calls a ‘perfect reason’ or ‘pro toto reason’, i.e., a reason that completely explains a deontic fact. Broome thinks there is a second type of reason that explains by being part of a weighing explanation involving more than one reason. These are his ‘pro tanto reasons’. What weighing explanations are is controversial, and it is also not clear to me. (Cf. Nebel 2018.) Fortunately, we do not need them here. All we need are the ‘pro toto reasons’. (Things may be different in the practical domain, but here I am concerend with the epistemic domain.)
Here ‘sufficient evidence’ should not be taken as evidence sufficient for correctness (which would make the norm circular and trivial) but evidence sufficient for truth. I believe that sufficient evidence should be understood in terms of safety: the fact that X is sufficient evidence for p if and only if X is a safe indicator that p. Safe indication is a modal relation, namely, guaranteeing truth in close worlds. So X is sufficient evidence for p if and only if in all close worlds in which X obtains, p is true. This would yield a kind of ‘objective evidentialism’. Alternative conceptions of sufficiency of evidence could be plugged in here, however, such as a pragmatic threshold conception.
For a defense of this claim, and of evidentialism, see Hofmann and Piller (2019).
It is noteworthy that epistemologists which are sympathetic toward reliabilism have recently tried to incorporate some evidentialist element in their views. For example, Comesaña (2010) and Goldman (2011) have both arrived at a kind of hybrid view combining evidentialism and reliabilism (‘evidentialist reliabilism’, as Comesaña calls it).
This is in line with a teleofunctionalist understanding of the (retrospective) evidentialist norm (vi): our doxastic capacities are supposed to let us hold true beliefs on the basis of sufficient evidence.
If it is replaced by a knowledge-norm, it depends on what the connection between evidence and knowledge is. Roughly we can say that in general, if the conditions in the content of the norm of belief are not intimately related to evidence, evidence will not be normative. The general lesson is clear: whether evidence is normative depends on what the norm of belief is.
Cf. Sosa and Sylvan (2018).
I am not subscribing to ‘ought implies can’ in general. But for the present purposes it does not do any harm to make the assumption that the subject needs to be able to do what she ought to do. The relevant abilities are to be understood as dispositions, be they narrow or wide. (For a convincing general criticism of ‘ought implies can’ see P.A. Graham 2011.) The condition that the question whether p is activated for the subject allows us to avoid an implausible inflation of oughts concerning propositions that the subject is not confronted with at the relevant time and, thus, is not subject to an ought even in case she has sufficient evidence for these propositions.
Recently, this has been emphasized by Sullivan-Bissett and Noordhof (2019). They use this point to argue against an explanatory argument for the truth-norm in connection with doxastic transparency, as proposed by Shah and Velleman (2005). I do not have the space to enter the discussion of doxastic transparency here. Suffice it to say that the proposed view does not vindicate doxastic transparency as conceived by Shah and Velleman: forming the belief that p by following one’s sufficient evidence, and thus settling the question whether p, allows one to infer that one ought to believe p only if one has sufficient evidence for believing that one has sufficient evidence for p (and perhaps in addition one has (implicit) knowledge of the relevant evidentialist norm of belief). So one cannot simply move from <p > to <I ought to believe that p > without significant further material. To the extent to which we are aware of our sufficient evidence and the norm of belief, we can make this inference. But it is unclear to what extent we do so.
This terminology is inspired by Ralph Wedgwood’s discussion of ought and correctness in Wedgwood (2013).
The opposite direction of the conditional is satisfied trivially, of course. It can only be the case that you ought to (form belief that p on the basis of your sufficient evidence for p) if you have sufficient evidence (leaving aside the activation and ability conditions). In other words, we could turn it into a biconditional.
A tricky question arises as to propositions that cannot be both true and believed at the same time (like Moore-paradoxical propositions of the form <p and I do not believe that p>. Wedgwood (2013) provides a plausible proposal about how to deal with these cases.
A similar but different form of teleology is sketched by John Gibbons in Gibbons (2014). He characterizes this view as follows: “According to some optimists, if it would be good that p, then you should be more inclined to believe that p than you would be if it would be bad that p, even given the same evidence.“(ibid., p. 99) This form of teleology speaks of its being good that p; Steglich-Petersen’s form of teleology speaks of its being good that one forms a true belief.
The second step faces serious difficulties. A fully worked out analysis of these difficulties would require some longer investigation. In order to indicate where the difficulties come from suffice it to say the following. Where does the value of forming a true belief about whether p come from? What is it grounded in? If it comes from a final epistemic value, it is probably the final epistemic value of true belief or knowledge. (What else?) Then the view would indeed be a genuine epistemic teleology. But the old problem of seemingly valueless true beliefs has a come back. If, however, the value is a pragmatic value, such as the value of successful action that requires a p-dependent choice, we have left epistemic teleology behind and entered an inter-domain grounding view. In effect, we have endorsed a form of pragmatism, a form of teleological pragmatism. Perhaps, there are these pragmatic-value-based normative reasons to form true beliefs about whether p. But this would have no tendency to show that there are no reasons for belief that are grounded in evidential facts plus the general evidentialist norm fact. The proposed view does not have to deny the pragmatic-value-based normative reasons, in case there are any.
Initially, Street frames the discussion in terms of ‘epistemic reasons’. (Cf. Street 2009, p. 217.) I take it that this is preliminary. In later sections she only appeals to likelifying facts.
Actually, Street embeds her argument in a dilemma that arises from accepting or rejecting a certain account of belief, the ‘constitutive-takings account’ (cf. Street 2009, sc. 5). Accepting this account is incompatible with normative realism, as Street argues convincingly. So we can focus on the second option, rejecting the ‘constitutive-takings account’. Therefore, we do not have to go into this account any further.
I might venture an exegetical hypothesis here. Why does Street implicitly assume that the normative realist is a norm-firster? She has imported the argument from her earlier and similar argumentation against practical normative realists (cf. Street 2006). And there it is of course intended to primarily deal well with reasons-firsters like Scanlon, Parfit, and Raz. But we are not to presuppose that the flow of grounding is of the same shape in the practical and the epistemic domain. It might be that in the practical domain, reasons ground correctness and are relatively fundamental (and even non-naturalizable). In the theoretical domain, reasons are not relatively fundamental but grounded in the norm (and evidential facts).
Incidentally, if the causal-historical facts underlying teleofunctions like the one mentioned – the job or teleofunction to produce true beliefs by following one’s evidence – are the non-normative facts that ground the norm of belief, we will arrive at an objective type of normative realism. Plausibly and arguably, the relevant teleofunctions are independent of our evaluative attitudes. At one place Street, however, characterizes evaluative attitudes extremely broadly, including not only desires and normative judgements but also “unreflective evaluative tendencies such as the tendency to experience X as counting in favor of or demanding Y“(Street 2016, n. 4, p. 334). Now, if X is an evidential fact and Y is a belief (favored by the evidential fact X) we might conclude that dispositions or capacities to form beliefs by following one’s evidence counts as evaluative attitudes, according to this very broad characterization. If this is so, earlier evaluative attitudes (in our ancestors) (partly) explain/ground the fact that our doxastic capacities now have the teleofunction described by the norm of belief, according to teleofunctionalism. Would that amount to a denial of objectivity? – No, in my view. Street has stretched the concept of an evaluative attitude beyond recognition and thus made it unsuitable for the purposes of leading a significant debate between objectivists and antiobjectivists (realists and antirealists).
References
Alvarez, M. (2010). Kinds of reasons, OUP.
Alvarez, M. (2018). Reasons for action, acting for reasons, and rationality. Synthese, 195(8), 3293–3310.
Berker, S. (2013a). Epistemic teleology and the separateness of propositions. Philosophical Review, 122(3), 337–394.
Berker, S. (2013b). The rejection of epistemic consequentialism. Philosophical Issues, 23, 365–387.
Comesaña, J. (2010). Evidentialist reliabilism. Nous, 44(4), 571–600.
Cuneo, T. (2007). The normative web, Clarendon Press.
Cuneo, T., Kyriacou, C. (2018). “Defending the moral/epistemic parity”, In: C. McHugh, J. Way, D. Whiting (eds.), Metaepistemology, OUP, 27-45.
Epstein, B. (2015). The ant trap, OUP.
Fine, K. (2018). “Truthmaking and the is-ought gap”, Synthese, published online 11 Nov. 2018.
Gibbons, J. (2014). Knowledge versus truth. In C. Littlejohn & J. Turri (Eds.), Epistemic Norms (Vol. 2014, pp. 97–114). New York: Oxford University Press.
Goldman, A. (2011). “Toward a synthesis of reliabilism and evidentialism? Or: evidentialism’s troubles, reliabilism’s rescue package”, In: T. Dougherty (ed.), Evidentialism and Its Discontents, OUP, 2011, 254–80.
Graham, P. A. (2011). ‘Ought’ and ability. Philosophical Review, 120(3), 337–382.
Grimm, St. (2009). “Epistemic normativity”, In: A. haddock, A. Millar, D. Pritchard (eds.), Epistemic Value, OUP, 243-64.
Heathwood, C. (2009). Moral and epistemic open question arguments. Philosophical Books, 50(2), 83–98.
Heathwood, C. (2018). “Epistemic reductionism and the moral-epistemic disparity”, In: C. Kyriacou, R. McKenna (eds.), Metaepistemology. Realism And Anti-Realism, Hampshire: Palgrave Macmillan, 45–70.
Hofmann, F. (2013). Three kinds of reliabilism. Philosophical Explorations, 16(1), 59–80.
Hofmann, F., Piller, C. (2019). Norm-reasons and evidentialism. Analysis, 79(2), 202–06.
Hyman, J. (2015). Action, knowledge, and will, OUP.
Ismael, J. (2011). A modest proposal about chance. Journal of Philosophy, 108(8), 416–442.
Kearns, S., & Star, D. (2009). Reasons as evidence. Oxford Studies in Metaethics, 4, 215–242.
Nebel, J.M. (2018). Normative reasons as reasons why we ought. Mind, 128(510), 459–84.
Olson, J. (2018). Moral and epistemic error theory: The parity premise reconsidered. In C. McHugh, J. Way, & D. Whiting (Eds.), Metaepistemology (pp. 107–121). Oxford: OUP.
Rowland, R. (2013). Moral error theory and the argument from epistemic reasons. Journal of Ethics and Social Philosophy, 7(1), 1–24.
Schroeder, M. (2015). Is knowledge normative? Philosophical Issues, 25(1), 379–395.
Shah, N., Velleman, D. (2005). Doxastic deliberation. The Philosophical Review, 114(4), 497–534.
Setiya, K. (2007). Reasons without rationalism. Princeton: Princeton University Press.
Skorupski, J. (2010). The domain of reasons, OUP.
Sosa, E., Sylvan, K. (2018). “The place of reasons in epistemology”, In: D. Star (ed.), Oxford handbook of reasons and normativity, OUP, 555-74.
Steglich-Petersen, A. (2008). Does doxastic transparency support evidentialism? Dialectica, 62(4), 541–547.
Steglich-Petersen, A. (2011). “How to be a teleologist about epistemic reasons”, In: A. Steglich-Petersen, A. Reisner (eds.), Reasons For Belief, CUP, 13-33.
Street, S. (2006). A Darwinian dilemma for realist theories of value. Philosophical Studies, 127, 109–166.
Street, S. (2009). “Evolution and the normativity of epistemic reasons”, Canadian Journal of Philosophy, vol. 35 supplemental volume, ed. D. Hunter, 213-48.
Street, S. (2016). Objectivity and truth: you’d better rethink it. Oxford Studies in Metaethics, 11, 293–334.
Sullivan-Bissett, E., Noordhof, P. (2019) The transparent failure of norms to keep up standards of belief. Philosophical Studies. https://doi.org/10.1007/s11098-019-01242-y.
Sylvan, K. (2018). Knowledge as a non-normative relation. Philosophy and Phenomenological Research, 97(1), 190–222.
Väyrynen, P. (2011). In: M. Brady (ed.), New Waves in Metaethics, Palgrave-Macmillan, 185-207.
Way, J. (2017). Reasons as premises of good reasoning. Pacific Philosophical Quarterly, 98, 251–270.
Wedgwood, R. (2013). “The right thing to believe”, In: T. Chan (ed.), The Aim of Belief, Oxford University Press, 2013, 123–39.
Wedgwood, R. (2015). The pitfalls of ‘reasons’. Philosophical Issues, 25, 123–143.
Whiting, D. (2010). Should I believe the truth? Dialectica, 64(2), 213–224.
Whiting, D. (2018). Right in some respects: Reasons as evidence. Philosophical Studies, 175, 2191–2208.
Williamson, T. (2000). Knowledge and its limits, OUP.
Williamson, T. (2018). “Knowledge, action, and the factive turn”, In: V. Mitova (ed.), The Factive Turn in Epistemology, CUP, 2018, 125–41.
Worsnip, A. (2018). Eliminating prudential reasons. Oxford Studies in Ethics, 8, 236–257.
Acknowledgements
I would like to thank Hannes Fraissler, Susanne Mantel and Christian Piller for many helpful discussions and comments. A special thanks goes to Paul Noordhof who gave me the opportunity to present some of the material in his research colloquium at York University in March 2019, and also many thanks to the participants for valuable comments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Hofmann, F. Is Evidence Normative?. Philosophia 49, 667–684 (2021). https://doi.org/10.1007/s11406-020-00241-2
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11406-020-00241-2