Intuitions are Used as Evidence in Philosophy

Abstract

In recent years a growing number of philosophers writing about the methodology of philosophy have defended the surprising claim that philosophers do not use intuitions as evidence. In this paper I defend the contrary view that philosophers do use intuitions as evidence. I argue that this thesis is the best explanation of several salient facts about philosophical practice. First, philosophers tend to believe propositions which they find intuitive. Second, philosophers offer error theories for intuitions that conflict with their theories. Finally, philosophers are more confident in rejecting theories to the extent that they have several (intuitive) counter examples involving diverse cases. I argue that these facts are better explained by philosophers' using intuitions as evidence than by any plausible contrary explanations. I further argue that aspects of philosophical practice that my thesis may initially seem ill-suited to explain are in fact unsurprising whether or not my thesis is true.

Do intuitions function as evidence in philosophy? This question has generated a flurry of recent metaphilosophical work. In this paper I will argue that on one significant reading of the question, the answer to it is yes. To set the stage for that answer, though, we must first clarify what we mean by ‘intuitions’ and ‘evidence’, and whether the question concerns how philosophy is actually done or how it should be done.

1. Clarifying the question

I take intuitions to be mental states that we find ourselves in when considering particular propositions. I take it that when one has an intuition that P:

• it seems to one that P;

• this seeming¹ is not the conscious result of an inference;

• this seeming is not the conscious result of an apparent memory that P, a sensorial experience as of P, or someone else’s testimony that P.

When I say in (ii) and (iii) that one’s seeming is ‘not the conscious result of’ X, I mean, roughly, that were one to reflect on whether one’s seeming is the result of X, it would not seem to one that it is. So, if one were asked why it seems to one that P, one would not cite an inference, memory, sensory experience, or others’ testimony (unless one of those things had independently caused it to seem to one that P). (ii) is compatible with one’s intuition being the result of some kind of tacit or subconscious inference. It is simply meant to rule out its seeming to one that P because one has drawn a conscious inference that P from other premises. Similarly, I mean (iii) to be understood in such a way that one of the sources mentioned there could have been the cause of your intuition, provided that this fact is not accessible to you. Finally, as I use the term, ‘intuition’ is non-factive: one can have an intuition that P even if P is false.

I take it that inferential seemings, memorial seemings, sensorial seemings, and testimonial seemings have a distinct phenomenology. (ii) and (iii) are meant to distinguish intuitive seemings from these, because in asking whether intuitions function as evidence in philosophy we are not interested in whether people’s apparent memories, sensory experiences, etc. function as evidence. However, the above leaves open the positive nature of intuitions. For example, some philosophers hold that one's having an intuition that P involves its seeming to one that P is necessary (Bealer 1998: pp. 207-8), and some hold that it involves its seeming to one that P being caused (or seeming to be caused) solely by one’s understanding of P or competence with the concepts involved in P (Sosa 2014: p. 46). I take no stand here on whether these conditions are necessary for a mental state to be an intuition.

I also do not commit myself to any positive view on the nature of seemings. Williamson (2007), who expresses scepticism towards the idea that there is a unique mental state called ‘intuition’, says that when considering a putatively ‘intuitive’ proposition, ‘I am aware of no intellectual seeming beyond my conscious inclination to believe [that proposition]’ (p. 217). Cappelen (2014b: p. 595) says the same. For all I have said here, a seeming that P may just be a conscious inclination to believe that P (cf. Sosa 2007: pp. 47-50 and 2014: p. 40, Earlenbaugh and Molyneux 2009).² A conscious inclination to believe is still a phenomenological state, and so I take it that even Williamson and Cappelen are acquainted with something to which they can take me to be referring when I talk about intuitions in philosophy (cf. Bengson 2014: p. 569).³

The above gloss is thus non-committal enough that almost all philosophers should recognize some phenomenal state as meeting (i)-(iii). At the same time, it is specific enough to distinguish this phenomenal state from other states, such as apparent memories and sensory experiences, with which we are not here concerned.

So much for ‘intuition’. As I use the term ‘evidence’ in this essay, E is evidence for T relative to background knowledge K iff P(T|E&K) > P(T|K) – that is, E raises the probability of T relative to K.⁴ A person takes E to be evidence for T or uses it as evidence relative to K iff his conditional credence in T given E&K is greater than his conditional credence in T given K.⁵ So, like ‘intuitions’, my use of ‘evidence’ is non-factive in the sense that E can be evidence for T even if E is false. (If you dislike this consequence of my account, you can substitute ‘E is evidence for T’ with ‘E would be evidence for T if E were true’, and take the former sentence to be true only when E is true. As far as I can tell, nothing crucial hangs on this semantic issue.)

Two consequences of this analysis are immediately worth flagging. The first is that evidence is relative to a context. E can raise the probability of T relative to K but not relative to K*. For example, drawing a red marble out of a two-marble bag is evidence that the other marble is red relative to the background knowledge that the bag contains either two red marbles or two blue marbles, but not relative to the background knowledge that it contains one red marble and one blue marble. Similarly, we will see that intuitions may be (treated as) evidence for philosophical claims relative to some contexts but not others.

A second consequence of this analysis is that all evidence is propositional, since only propositions stand in epistemic probability relations (Williamson 2000: pp. 194-200). Because of this, strictly speaking intuitions cannot be evidence because intuitions are mental states, not propositions. However, I take it that when we ask whether intuitions are evidence in philosophy, we are asking whether the fact that some philosopher has an intuition is evidence in philosophy. This phrase is cumbersome, however, and so for ease of notation I will continue to speak of intuitions as themselves being evidence. This should be taken as elliptical for the fact that some philosopher has the intuition being evidence. When, for instance, I speak below of philosophers having a higher conditional credence in a theory T given their intuitions and background knowledge than given their background knowledge alone, what I mean is that their credence in T given their background knowledge and the fact that they have those intuitions is higher than their credence in T given their background knowledge alone.

With that clarification in hand, we can return to our central question: do intuitions function as evidence in philosophy? As the above analysis of evidence makes clear, we can understand this question in two ways. The first is: do philosophers use intuitions as evidence, or take intuitions to be evidence? The second is: are philosophers’ intuitions actually evidence for the kinds of claims they make? The answer to the first question is yes just in case philosophers have a higher conditional credence in their theories given their intuitions and background knowledge than they do in their theories given their background knowledge alone. The answer to the second question is yes just in case the conditional probability of philosophers’ theories is higher given their intuitions and background knowledge than it is given that background knowledge alone. Presumably, if the answer to the second question is yes – if intuitions are evidence in this sense – then philosophers ought to use them as evidence in the sense spelled out above. Conversely, if the answer to the second question is no – if intuitions are not evidence – then philosophers ought not to use them as evidence.

2. The centrality of intuitions

In this paper I will argue that the answer to the first question above is yes – philosophers do use intuitions as evidence. Although one would expect this to be the orthodox view,⁶ a surprising number of philosophers have denied it in recent years, including Williamson (2007), Earlenbaugh and Molyneux (2009), Deutsch (2009: p. 459, 2010: pp. 448-50), Cappelen (2012), Ichikawa (2014a, 2014b), Chudnoff (2014: pp. 13-4), and van Roojen (2014).⁷ In several of these cases, this denial appears to be partly motivated by a conviction (or worry) that the answer to my second question is no – that is, that intuitions are not really evidence (or at least not good evidence) for philosophical theories – a conviction that is sometimes based on general philosophical considerations (how could something psychological provide evidence for something philosophical?) and sometimes based on experimental philosophers’ empirical critiques of intuitions. (For the former, see, e.g., Williamson 2007: p. 211, Ichikawa and Jarvis 2013: p. 280, Chudnoff 2014: pp. 13-4, and van Roojen 2014: pp. 159-60; for the latter, see, e.g., Williamson 2007: p. 190, Deutsch 2009: pp. 450-3 and 2010, Ichikawa 2014a and 2014b, and Cappelen 2012: ch. 11 and 2014c.)⁸

How have so many philosophers been misled about philosophical methodology, if these sceptics are right? Ichikawa (2014b) notes that the word ‘intuition’ is ambiguous: it can refer either to a psychological state or to the content of that state. (I have been using it to refer to the state, not the content.) It is uncontroversial, he thinks, that philosophers often use as evidence intuitive propositions; but he denies that they use as evidence intuition states (p. 5).⁹Williamson (2007: ch. 7), similarly, thinks that propositions that are sometimes described as ‘intuitive’ are propositions that philosophers know; and are thus sometimes evidence in philosophy. (Williamson’s primary claim seems to be about what is actually evidence, but I take it that he would endorse the claim that intuitive propositions are used as evidence as well.) On both authors’ views, to take intuition states as evidence is to ‘psychologize’ the evidence in an objectionable way (cf. also Deutsch 2010, Ichikawa and Jarvis 2013: pp. 279-81, Chudnoff 2014: pp. 13-4, van Roojen 2014: pp. 159-60, and Cappelen 2014b: pp. 578-80, 593-4).

Cappelen (2012) takes a more radical route. He denies that there are any interesting mental states of the kind I have endeavoured to describe, and thinks that most uses of the word ‘intuition’ and cognate terms in philosophical theorizing can be replaced with other words without loss of meaning. In some cases, he says, terms like ‘intuitive’ can simply be removed; in other cases, ‘intuitive’ can be replaced by ‘pre-theoretic’. In the remaining cases, Cappelen thinks, ‘intuitively, P’ simply means that the judgment that P has been (or can be) ‘reached with relatively little effort and reasoning’ or is easy to understand (p. 65).¹⁰ In many of these cases ‘intuitively’ is also used as a hedge – one does not want to flat-out assert that P and so instead one simply asserts that P is intuitive. Cappelen thinks that none of these interpretations support the claim that intuitions are used as evidence in philosophy.

I will argue that, contrary to these philosophers’ remonstrations, intuitions are used as evidence in ordinary philosophical practice. Following Cappelen (2012: p. 3), I will call this thesis Centrality.¹¹ In defence of Centrality, I will examine some familiar facts about how philosophy is commonly done, and argue that these facts are only explicable if we accept that philosophers use intuitions as evidence.

There is some unavoidable vagueness in the dialectic here. Centrality should not be read as saying that every time a philosopher has an intuition, she uses that intuition as evidence for some theory – indeed, my argument will partly turn on the fact that in some cases philosophers do not do this. But neither should all sceptical views be read as saying that no philosopher has ever used an intuition state as evidence. Rather, the disagreement is over fuzzier in-between claims, like the claim that philosophers frequently use intuitions as evidence, or that this is standard or normal philosophical practice. It is these more modest claims that I will be defending.¹²

3. Kripke-exegesis¹³

Many of the recent arguments for and against Centrality have focused on the correct interpretation of specific ‘intuition’ language in key philosophical texts, especially Kripke’s Naming and Necessity (1981). By contrast, my argument is based primarily on more general facts about how philosophy is done. Nevertheless, before coming to those facts, it will be instructive to see what Centrality-sceptics say about the use of ‘intuition’ language by philosophers like Kripke.

The most widely discussed passage from Naming and Necessity is the one in which Kripke asks us to imagine that Gödel did not discover the incompleteness of arithmetic. Rather, the incompleteness of arithmetic was discovered by a man named Schmidt, whose proof Gödel stole and published as his own. In such a case, a crude descriptivist theory of reference would have it that those of us who associate the name ‘Gödel’ only with the description ‘the man who discovered the incompleteness of arithmetic’ are referring to Schmidt when we use the name ‘Gödel’. ‘But’, says Kripke, ‘it seems to me that we are not. We simply are not’ (p. 84). So this descriptivist theory is mistaken.

This thought experiment looks like a paradigmatic example of intuitions being used as evidence in philosophy. It is taken to be so, for example, by experimental philosophers who purport to show that Kripke’s intuition that ‘Gödel’ does not refer to Schmidt is not universally shared (Machery et al. 2004; Mallon 2009). But sceptics of Centrality insist that here and elsewhere in Naming and Necessity, Kripke is not using intuitions as evidence. Here I will discuss the views of three Centrality-sceptics on Kripke: Deutsch (2009, 2010, 2015), Ichikawa (2014b), and Cappelen (2012, 2014b).

In the case at hand, Deutsch contends that Kripke’s evidence is not that it is intuitive that ‘Gödel’ does not refer to Schmidt, but, more simply, that ‘Gödel’ does not refer to Schmidt:

[Kripke] does not say that it is intuitive that we are not talking about Schmidt; he says straight out, and emphatically, that we are not talking about Schmidt. Of course, the Gödel case is an intuitive counterexample to descriptivism for many readers of Naming and Necessity, but this is a logically inessential feature of the case. Kripke’s argument against descriptivism succeeds if the Gödel case is a genuine counterexample. Whether it is an intuitive counterexample is not clearly relevant, and there is nothing in Kripke’s presentation of the case that would lead one to believe that Kripke thinks it is relevant. (Deutsch 2009: p. 451, emphases in original; cf. 2010: pp. 450-1)

After saying much the same thing about the above case, Ichikawa argues that this point holds more generally:

It’s undeniable that philosophers rely on many propositions that are intuitive; but it does not follow that philosophers rely on psychological states, intuitions, and allow them to play evidential roles. Neither does it follow that intuitive propositions are either available or relied upon because they are intuitive. (Ichikawa 2014b: p. 236, emphasis in the original)

Deutsch and Ichikawa admit that it is intuitive that ‘Gödel’ does not refer to Schmidt, but deny that Kripke endorses this claim because it is intuitive. At first glance, this looks implausible. Right before Kripke says ‘straight out’ that ‘We simply are not [referring to Schmidt]’, he says ‘it seems to me that we are not’. This sentence appears to report an intuition which Kripke is using to support the claim that we really are not referring to Schmidt.

However, Deutsch and Ichikawa may receive help here from Cappelen’s ‘paraphrasing-away’ strategies. Cappelen (2012: pp. 71-5) does not discuss this passage in his commentary on Kripke in his book (2012: pp. 72-5), but in a later article (2014b: p. 593, emphases in original), he agrees with Deutsch and Ichikawa that ‘In Kripke’s Gödel-Schmidt case, the evidence is that the subject refers to Gödel and not to Schmidt (not that we think it)’. He then suggests that we are misled about this and similar cases because ‘[w]e philosophers often qualify our claims with phrases like, I think that C (or it seems that C or Intuitively C) but, as I argue in Part I of PWI, those are hedges, not descriptions of evidential sources’.

In a more recent paper, Deutsch (2015: p. 10) endorses the claim that Kripke uses ‘seems’ here as a hedge. He notes that in an earlier paper, ‘Identity and Necessity’, Kripke (2011: p. 6) writes ‘I, like other philosophers, have a habit of understatement in which “it seems plainly false” means “it is plainly false”’. This aside supports the suggestion that in the passage quoted above Kripke is not reporting an intuition at all, but simply presenting his judgment in his understating way and then strengthening this a sentence later.¹⁴

Other passages from Naming and Necessity are harder for intuition sceptics to explain. On page 48, Kripke says that the thesis that names are rigid designators ‘satisf[ies] the intuitive test mentioned above’, where ‘satisfying this test’ means agreeing with our intuitive judgments about a case. On page 60, Kripke claims that the Greek letter ‘π’ is a name for a real number (as opposed to being shorthand for the description ‘the ratio of the circumference of a circle to its diameter’). He says that this is a claim that he has ‘nothing but a vague intuitive feeling to argue for’, implying that this intuition and nothing else (that Kripke has available to him) is evidence for this claim.

Contra Deutsch and Ichikawa, in these passages Kripke does appear to be taking intuitiveness as evidence. Cappelen’s paraphrasing strategies do not seem to apply to these passages. Perhaps ‘intuitive’ means pre-theoretic in the rigid designator case, but if this is so it is a counterexample to Cappelen’s claim that ‘there is no evidence that Kripke at any point treats “being endorsed prior to a theory” as carrying evidential weight’ (2012: p. 73 n15). And the π case is about as explicit an example as one could ask for of a philosopher asserting an intuitive proposition P without any evidential support apart from its intuitiveness, which Cappelen claims to be unable to find in philosophical texts (2012: pp. 121-2).

It seems to me that, in light of these last two examples, Kripke remains a problematic case both for moderate intuition-sceptics like Ichikawa and Deutsch and more radical sceptics like Cappelen. But the dialectic about the Gödel/Schmidt case makes clear that intuition sceptics do have resources available to explain many passages that initially look problematic for them. Cappelen is certainly right that words like ‘seems’ and ‘intuitively’ can be used in multiple senses, and many (although, I think, not all) passages can thus be plausibly interpreted in both Centrality-friendly and Centrality-hostile ways.¹⁵ These textual disputes are unlikely to be settled anytime soon. To move the debate about Centrality forward it will be helpful to take a step back from textual exegesis and look at some more general facts about philosophical practice.

4. Fact 1: philosophers tend to believe intuitive claims

We saw that Deutsch and Ichikawa both grant that in Kripke’s Gödel/Schmidt case, the proposition that ‘Gödel’ refers to Gödel is intuitive for many people. Cappelen would not grant this because he denies that intuitions exist. However, as I argued in §1, since Cappelen does admit the existence of conscious inclinations to believe, he is committed to the existence of some phenomenal state satisfying conditions (i)-(iii), which I have been calling an intuition. And Cappelen presumably would grant that many philosophers have an inclination to believe that ‘Gödel’ refers to Gödel.¹⁶

More generally, all parties to the debate should agree that philosophers tend to believe intuitive propositions: most philosophers in considering many cases (especially thought experiments) tend to believe that P when they have the intuition that P.¹⁷ It is true that, as Ichikawa says, it does not logically follow from this fact that philosophers believe intuitive propositions because they are intuitive. However, that philosophers believe intuitive propositions because they are intuitive, and that they use their intuition-states as evidence for those propositions, provide a very plausible explanation for the fact that philosophers tend to believe intuitive propositions. For this fact notes a striking correlation between philosophers’ beliefs¹⁸ and philosophers’ intuitions. In the absence of any causal connection between the two, this correlation is quite surprising.

For example, there are infinitely many views one could have about what ‘Gödel’ refers to in the case Kripke describes – e.g., Gödel, Schmidt, nothing, both Gödel and Schmidt, the fusion of Gödel and Schmidt, the moon, the year 2137, etc. So, there are presumably infinitely many intuitions that one could have, in line with the above. However, Kripke’s belief about the case was that ‘Gödel’ refers to Gödel, and his intuition was also that ‘Gödel’ refers to Gödel. If Kripke’s belief and intuition were simply randomly selected from the above classes, or caused by independently predictable but unrelated processes, this would be quite a coincidence.

Similar remarks go for other philosophers’ beliefs in intuitive propositions. Even if we thought that in Kripke’s case, it was just a coincidence, and that there was some explanation other than his intuition for why Kripke, it is implausible in the extreme that there would be no general explanation of the correlation between philosophers’ beliefs and their intuitions – that this is just a massive coincidence.

So there is almost certainly some explanatory relation between philosophical beliefs and intuitions. This does not necessarily mean that the latter cause the former. It could be that the former cause the latter or that they have a common cause.¹⁹ The first option is not very plausible as a general explanation of the correlation of philosophical beliefs and intuitions (cf. Bealer 1998: pp. 210-1). Before considering Kripke’s thought experiments, most philosophers were descriptivists. If their philosophical beliefs were causing their intuitions, rather than the other way around, we would have expected them to have the intuition that ‘Gödel’ refers to Schmidt.²⁰ And yet, even if the experimental philosophers are right that the intuition that ‘Gödel’ refers to Gödel is not as widespread as Kripke might have thought, many philosophers did share this intuition and formed the belief that ‘Gödel’ refers to Schmidt – a belief that they did not have beforehand. Similar remarks go for Gettier’s thought experiments and other famous philosophical counterexamples.

The second option – that philosophical beliefs and intuitions have a common cause – is more plausible. For example, there may be some intuition-generating faculty – perhaps one somehow connected to linguistic and/or conceptual competence – that caused both Kripke’s intuition that ‘Gödel’ refers to Gödel and his belief that ‘Gödel’ refers to Gödel. Let us then suppose that either philosophical intuitions cause philosophical beliefs, or the two have a common cause. I will now argue that Centrality is true in either case.

Recall my definition of using E as evidence above:

A person uses E as evidence for T relative to K iff his conditional credence in T given E&K is greater than his conditional credence in T given K.

At first glance, it seems that philosophers need not use intuitions as evidence even if their beliefs are caused by their intuitions. After all, that philosophers believe propositions they find intuitive because they find them intuitive does not imply that they believe those propositions because they believe that they find them intuitive. For most philosophers in most cases of intuition-based beliefs, their belief that P could be caused by an intuition that P directly, and not by an awareness or belief that they have an intuition that P. So one might argue that while intuition-states are sometimes a precondition for philosophical beliefs, they are not treated as evidence for those beliefs (cf. Chudnoff 2014: pp. 12-4, Pust 2014: §2.4).²¹

I am open to the possibility that frequently, intuitions cause beliefs or beliefs and intuitions are both caused by some third factor (say, a cognitive faculty), without one consciously basing one’s belief that P on a belief that one has an intuition that P. However, I maintain that once one has become aware that one holds an intuition and that this is causally connected to one’s belief in one of the above ways, if one retains the same level of confidence that P, one is now basing it evidentially on the proposition that one holds that intuition.

Consider first an intuition’s causing a belief directly. Suppose that an intuition causes me to raise my credence in P. If I come to realize this and maintain that credence in P, I tacitly take the fact that I have an intuition that P to have increased the probability that P.

To see this, compare, say, a brain lesion leading me to believe that P. Were I to come to think that the only reason I believe that P is that I had a brain lesion unconnected to the truth of P, I would be inclined to give up my belief that P. However, if I were to reflect and realize that the only reason I believe that P is that I find it intuitive, this would not normally incline me to give up my belief that P. This suggests that I take P to be more probable on my background knowledge K plus my intuition that P than on K alone, whereas I do not take P to be more probable on K plus my having a brain lesion than on K alone. In the latter case, I realize that I have irrationally adopted a high confidence in P, and so I revert to a credence matching what I take to be the probability of P on K alone; since in the former case I do not do this, I must take the probability of P to have actually been increased by my intuition.

Consider now an intuition and belief being caused by a common source. I take it that whatever the common source is (e.g., some cognitive faculty), it is less transparent than one’s intuitions or other mental states, and so more like the brain lesion in that respect. Now compare learning that my belief that P was solely caused by whatever also caused my intuition that P with learning that my belief that P was solely caused by whatever also caused my brain lesion. In the latter case I would be inclined to give up my belief that P (assuming that P has nothing to do with brain lesions), because I would not take P to be more probable on my background knowledge K plus the existence of whatever caused my lesion than on K alone. (This is because whatever caused my lesion is likely to be irrelevant to P.) If in the former case I would not be inclined to give up my belief that P, this suggests that I take whatever caused my intuition to be truth-conducive, so that I take it that my intuition that P raises the probability of P for me by implying that this truth-conducive source for my belief exists.

I think, then, that whatever the general causal relation between intuitions and beliefs, philosophers tacitly use intuitions as evidence when they form beliefs on their basis or on the basis of a common source, at least when they are aware of this connection. Since this happens quite often, Centrality is true. So the fact that philosophers believe intuitive claims is best explained by an account that implies Centrality. Examination of more complicated cases of the interplay of intuition and philosophical belief will provide further support for Centrality.

5. Fact 2: philosophers offer error theories

Although philosophers tend to believe propositions that they find intuitive, this natural disposition can be defeated. Deutsch (2009) offers an example of a theory of reference, Millianism, whose adherents admit that it has an unintuitive consequence, and takes this to show that not all philosophers of language rely on the ‘method of cases’, as experimental philosophers claim. One might go further and claim that such cases show that philosophers do not in general rely on intuitions at all. In fact, however, when we scrutinize these cases more closely, we find that they in general support my thesis, rather than disconfirm it. Deutsch closes his discussion by noting that Millians will seek ‘to “explain away” the contrary intuitions in a way that leaves their semantic view unscathed’ (p. 449). In short, this supports the claim that philosophers rely on intuitions as evidence because they would seek to explain intuitions away only if they thought non-veridical intuitions needed explaining. Let’s look at a few examples so that we can lay out this argument more explicitly.

Debates about knowledge provide some of the most salient examples of intuition-based philosophy. As Weatherson (2003: p. 1) notes, there are ‘literally dozens of cases where an interesting account of knowledge was jettisoned because it clashed with intuitions about a particular case’. With so many intuitions being bandied about in support of so many different theories, opponents of a theory often feel obliged to offer an explanation of the intuitions adherents of that theory invoke in support of it, one that does not presuppose the truth of the theory. Consider, for example, the following pair of cases (Turri 2014: p. 169; modified from Cohen 1999), used to argue that whether S knows that P depends on the practical stakes of S’s context:

Low Flight: Stewart is in the Atlanta airport, waiting to board his flight. A fellow traveller seated nearby looks up from his laptop, stretches, turns to Stewart and says, ‘I’ve been traveling all day and it’ll be a relief to get home to Detroit. A layover would be annoying. Say, do you happen to know whether this is a direct flight to Detroit?’ With his itinerary in hand, Stewart answers, ‘Yes, I do — it’s direct to Detroit’.

High Flight: Stewart is in the Atlanta airport, waiting to board his flight. Suddenly a man dressed in a uniform and carrying a small hardshell cooler rushes down the concourse, stops in front of Stewart’s gate and breathlessly says to Stewart, ‘I’m an organ courier transporting a kidney to a patient in Detroit. I need a direct flight to Detroit, or the kidney will spoil. Do you know whether this is a direct flight to Detroit?’ With his itinerary in hand, Stewart answers, ‘Sorry, I don’t know [whether it is]. You should check with an airline official’.

The intuition that the philosopher who puts forward these cases expects you to have is that in both cases Stewart speaks truly, so that in Low Flight, Stewart knows, but in High Flight, he doesn’t. One popular explanation of this is that whether S knows that P depends on whether S is in a position to act on the basis of P, which depends on pragmatic features of S’s situation.²²

Turri (2014) canvasses several explanations that epistemologists sceptical of the claim that Stewart’s knowledge differs in the two cases have put forward for our (alleged) intuition that it differs in these two cases – explanations that do not involve the truth of this claim. These explanations claim that in having this intuition we are confusing truth with linguistic propriety – it would be appropriate for Stewart to say that he knows or doesn’t know, but one of these claims is false. Different explanations differ on why truth and appropriateness come apart. On one view (see, e.g., Unger 1975: ch. 2),²³ Stewart knows in neither case, but it is appropriate for him to say that he knows in the first case because this is ‘close enough to the truth for practical purposes’ (Turri 2014: p. 172) – it correctly pragmatically implies, for example, that Stewart has strong evidence that the flight is direct to Detroit. On another view (see, e.g., Rysiew 2001 and 2007), Stewart knows in both cases, but it is appropriate for him to say that he doesn’t in the second case because this allows him to avoid pragmatically implying certain falsehoods – e.g., that he can rule out all alternatives to the flight’s being direct.

These explanations are error theories for the intuition that Stewart knows in the first but not the second case. Such error theories are common in philosophy.²⁴ Here are two more examples. First, Grice (1989: ch. 1) offers an error theory similar to those discussed above for Ryle’s (1950) intuition that morally neutral actions cannot be voluntary or involuntary. Grice suggests that this intuition was based on confusing the propriety of asserting P with the truth of P. It would be conversationally inappropriate to describe scratching one’s head as voluntary or involuntary, because it would be inappropriate to describe an action as voluntary or involuntary unless there was reason to do so (e.g., one was offering an excuse for one’s action). But this does not imply that scratching one’s head cannot be voluntary or involuntary. Second, after arguing that desires do not provide us with reasons to act, Parfit (2011) offers ten(!) explanations of why so many philosophers have thought that they do (pp. 65-70). For example, he notes that we often desire ends we have independent reasons to bring about, and that we have reasons to do what we enjoy, and often desire to do what we enjoy.

I think the most natural way to describe what is going on in these cases is as follows. Suppose we have two philosophers, Deon and Connie. Deon takes his intuition-state E to support theory T. Connie objects by saying that there’s a better explanation of E than T – namely, error theory T*. T* implies that E was to be expected whether or not T, so that E does not substantially confirm T. For example, consider the footbridge variant of the trolley thought experiments, where a trolley is hurtling down a track and about to run over and kill five people (Thomson 1985). You are standing on a footbridge overlooking the track with a very heavy man in front of you; if you push him off, the trolley will be stopped by his body, killing him but saving the five others. Here we can let E be the intuition that it is wrong to push the man. We could imagine T to be some elaborate deontological theory, but for simplicity’s sake let’s consider the limiting case in which T is simply the proposition that it is wrong to push the man. Deon proposes that E supports T because T is the simplest explanation of E: Deon has the intuition that it is wrong to push the man because it is wrong to push the man. Connie, a consequentialist, proposes T*: in most situations like the above in salient respects, pushing the man would have overall worse consequences than not pushing him.²⁵ She then claims that, although T would lead us to expect E, so would T*, because our intuitions about these kinds of cases respond mainly to coarse-grained features of the cases (e.g., that the case involves pushing a man to his death). Thus, E is not as good evidence for T as Deon thought, because it is equally well explained by T*.²⁶

This way of describing these cases presupposes that philosophers rely on intuitions as evidence.²⁷ If they do not, then it is not clear what relevance error theories of the above sort would have. If Deon does not take E to be evidence for T, why should he care when Connie observes that there’s another explanation of E that does not presuppose T?

Notice, moreover, that the error theorist in this kind of example is not offering an explanation of the content of an intuition. Connie does not grant the content of Deon’s intuition (that it is wrong to push the fat man); she thinks it is false.²⁸ So the exchange between Deon and Connie cannot be explained simply by the fact that Deon uses the content of his intuition as evidence. Rather, Connie’s error theory threatens Deon because he is using his intuition state as evidence.

The most natural explanation of why philosophers propose error theories, then, is that philosophers use intuitions as evidence. In the next section I will consider three alternative explanations of error theorizing that do not presuppose Centrality, and argue that they all fail.

6. Explaining away Fact 2: error theories about error theories

The first alternative explanation is that error theories threaten philosophical beliefs directly. While considering experimentalist critiques of intuition-based philosophy, Ichikawa (2014b) argues that although experimentalists are wrong to think that philosophers use intuitions as evidence, there is an experimentalist argument that does not rely on this assumption. According to this argument, experimental philosophical data call into question what our evidence is:

Empirical data about human reliability can provide undercutting defeaters for judgments of any kind. Call this the defeater critique. The defeater critique sees traditional armchair methods as insufficiently self-reflective; armchair philosophers proceed without enough regard for their abilities to discern relevant facts. (Ichikawa 2014b: p. 240, emphasis in original)

Along these lines, one might claim that an error theory about philosophers’ judgments that P provides a general reason to doubt those judgments, one which does not presume that those philosophers relied on intuitions in forming those judgments. For example, Parfit does not explicitly say anything about subjectivist intuitions; rather, what he says he is explaining is why subjectivists have wrongly judged desires to be reason-giving, not why they have had the intuition that desires are reason-giving. On this reading of Parfit, all he is doing in giving his error theory is saying ‘here are some epistemically questionable considerations that could lead you to form your judgment about this matter, so you’d better be less confident in that judgment’.

Although this is a plausible description of what Parfit is doing in this particular passage, it is not promising as a general description of the dialectical force of either the experimentalist critique or error theories. Consider the experimentalist critique first. Empirical data showing that people are frequently unreliable about a subject matter are only relevant to me if I am forming my beliefs in the same way or on the same basis as those who were studied. Psychologists have found that people’s judgements about probability are frequently irrational (Kahneman, Slovic, and Tversky 1982). But this should do nothing to undermine my confidence that A is no less probable than A&B, if I have deduced that proposition from the axioms of probability theory. The above research shows that people’s intuitive probability judgments are frequently wrong; it says nothing about inferred ones. Similarly, experimental data showing that people’s intuitive beliefs are generally unreliable constitute a threat to my beliefs only if my beliefs are formed on similar bases.

Return now to error theories. Suppose that I believe that pushing the fat man in front of the trolley is wrong not because I find that proposition intuitive but because it follows from a more general theory of mine – say, the Doctrine of Double Effect. In this case, while Connie’s objections may give Deon some reason to become less confident in T, they give me no such reason. So Connie’s error theory does not cast doubt on belief that T in general – it only does so if that belief is based in a certain way. Thus, if I raise the fat man case as a counterexample to consequentialism, and Connie responds with her error theory, she is presupposing that I am basing my belief on an intuition that it is wrong to push the fat man. Likewise, if I become less confident in the correctness of my counterexample in light of Connie’s error theory, I am presupposing that I am basing my belief that the counterexample works on an intuition that it works.

So taking error theories to be general reasons to doubt philosophical beliefs won’t work. Instead we must say that in many cases (including all the examples above, besides Parfit) they cast doubt on philosophical beliefs by casting doubt on the trustworthiness of intuitions, as in my above reconstruction of what goes on when philosophers offer and react to error theories. And if a philosopher does not take his intuitions to be evidence for his beliefs, it is unclear what relevance error theories of the above sort would have. Again, suppose Deon does not take E to be evidence for T. Then it seems that he should not care when Connie observes that there’s another explanation of E that does not presuppose T, or that P(E|T&K)/P(E|∼T&K) is not very top-heavy. After all, Deon’s belief that T was not based on E in the first place.

Second, consider the anti-Centrality explanations of the belief-intuition correspondence discussed in the previous section: that intuitions and beliefs have a common cause, or that intuitions cause beliefs, but that in neither case do philosophers use those intuitions as evidence. These will not enable us to explain error theories either.

If we explain the correlation between beliefs and intuitions by giving some common-cause explanation, and hold that philosophers only use intuition contents (T),²⁹ and not intuition states (E), as evidence, even when they are consciously aware that they are in those states, then we are unable to explain why philosophers are moved by error theories. If they do not use E as evidence for T, then it shouldn’t matter to them that P(E|T&K)/P(E|∼T&K) is not very top-heavy, or that T* provides as good an explanation for E as T.

Similarly, error theories would have no bite if intuition-states were simply a precondition for or cause of philosophical beliefs without being evidence for them. Suppose that I see a sparrow and believe that a sparrow is in front of me. Suppose that as a matter of human brain chemistry I can only believe that a sparrow is in front of me if I am in some brain state S, described in purely physical terms. However, I do not know this. You then challenge my belief that a sparrow is in front of me by showing me that the prior conditions of my brain and certain neurological laws make it highly likely that I am in S now whether or not there is a sparrow in front of me. Since I do not even know that I am in S, nor that S is a precondition for my belief, your remarks about S would leave me completely unmoved.

In this example, S was a precondition for belief, but we can also make S a cause. We could suppose that S caused me to form my belief on the basis of sense experience in the usual way. As far as I can see, my reaction to your neurological evidence would be exactly the same.

So, if intuitions were merely preconditions for or causes of philosophical belief that philosophers did not take to be evidence, it remains unclear why they would offer error theories and respond to them in the way they do. If I believe that I see a sparrow because I am in brain state S, but not because I believe I am in S, accepting the above ‘error theory’ would not move me to give up my belief. But if I believe that pushing the man off the footbridge is wrong because I know that I have an intuition that this is so, then I will be moved when a consequentialist observes that I am likely to have this intuition even given the assumption that pushing the man is the right thing to do.

Lastly, consider Cappelen-style scepticism about the existence of intuitions. The discussion above should lead us to reject such scepticism. If there are no such mental states as intuitions on which philosophers rely in constructing theories and which are challenged by error theories, then it again becomes unclear why philosophers propose error theories or feel threatened by them. Suppose, for example, that in describing the intuitive judgment about the footbridge case, Low Flight and High Flight, and whether scratching one’s head is voluntary, as ‘intuitive’, philosophers simply mean that this is their pre-theoretic judgment or one arrived at without much reasoning. Then it seems that Cappelen would have to say that error theories about these judgments are simply error theories about the judgments. But, again, my pre-theoretic or snap judgment is only threatened by an error theory if it is formed on the basis of a fact that that error theory seeks to independently account for (e.g., its seeming to me that that judgment is correct).

How might Cappelen respond? He acknowledges that ‘[m]any philosophers engage in an activity they describe as “explaining away intuitions”’ (2012: p. 89).³⁰ But, oddly, he doesn’t see this as a problem for his rejection of Centrality. Instead of explaining how this practice is compatible with his thesis, he simply argues that philosophers have no obligation to engage in it. But whether or not this is right, it is a normative thesis. And Cappelen’s stated thesis in his book is a descriptive one: philosophers do not use intuitions as evidence. Cappelen needs an error theory of his own, to explain why philosophers engage in explaining away intuitions when there are no intuitions to explain away.

Cappelen does suggest, in passing, that philosophers put forward error theories because they mistakenly endorse Centrality (pp. 89-90). Cappelen is officially agnostic on why Centrality has become so widely endorsed, but suggests that it is due to the ‘verbal tick’ of saying ‘Intuitively, BLAH’ a lot – a tick perhaps due to the influence of Kripke (p. 71) – and perhaps also Chomsky, Rawls, Moore, and ordinary language philosophers (pp. 22-3). So let us take Cappelen’s error theory of error theories to be the pernicious spread of the metaphilosophical view that philosophers use intuitions as evidence.

This is not a promising explanation of error theories. Cappelen offers no reasons to think that philosophers offer theories because they have come to accept Centrality, and it is implausible when one considers the full range of cases discussed above. For example, error theories put forward in debates about contextualist and pragmaticist views of knowledge (or ‘knowledge’) owe their most obvious methodological debt not to any of the philosophers mentioned above, but to Grice (who was largely reacting against ordinary language philosophers).

Inasmuch as Cappelen takes the endorsement of Centrality to be a recent development (the ‘verbal tick’ apparently started spreading ‘thirty to forty years ago’ (p. 50)), older examples of error theories are also harder for Cappelen to account for. For example, Unger expressed a version of his error theory about fallibilist intuitions in his ‘A Defence of Skepticism’, published in 1971. Along with the aforementioned Grice, Searle argued in his 1966 that we need to distinguish, for example, the conditions under which an act is voluntary and the conditions under which it is linguistically appropriate to call an act voluntary, and that failure to distinguish these explains ordinary language philosophers’ mistaken intuitions.

More generally, the above error theory about error theories faces a dilemma. Either philosophers’ metaphilosophical beliefs significantly influence their practice or they do not. If they do not, then Cappelen’s explanation is implausible. Many error theories take place in the context of ordinary philosophical discussion, for example about our concept of knowledge. Metaphilosophy is (often) not salient.

So if metaphilosophical views impact philosophy in this case, they must do so pretty generally. But in that case, philosophers’ endorsement of the claim that philosophers use intuitions as evidence will plausibly lead them to use intuitions as evidence. For instance, they will become more sceptical of their own views when their interlocutors offer error theories about their intuitions. And as soon as they do this, they are using intuitions as evidence, for reasons already rehearsed. And in that case Cappelen is no longer right that, as a matter of description, philosophers do not use intuitions as evidence.

7. Fact 3: diverse intuitions provide greater confidence

Error theories present a more complicated case of intuitional support than an intuition that P simply supporting a belief that P in all cases. I argued that the complexity they introduce – that philosophers are moved by considerations that suggest that the truth of their theory is not as good an explanation of the intuition as they thought – supports Centrality. As we will see in this section, other more complicated cases likewise support Centrality.

In arguing against consequentialism, Ross (1930: ch. II) presents three counterexamples to the claim that it is always wrong for someone ‘to do an act which would produce consequences less good than those which would be produced by some other act in his power’. The first counterexample involves choosing between fulfilling a promise to A and bringing about slightly more good to B, to whom one has made no promise. In the second case, one is choosing between benefiting A by fulfilling one’s promise to him or by doing some other act that would benefit A slightly more. In the third case, one is again choosing between helping A or helping B slightly more, but now ‘A is a very good and B a very bad man’. In each case Ross thinks it intuitively clear that we ought to choose the first option, contra consequentialism.³¹

Two features of this case should be familiar from philosophical argumentation. First, Ross offers several counterexamples, not just one. Second, the details of his counterexamples are carefully constructed so as to vary in significant ways. For example, not all three involve choosing between actions affecting different agents, and not all three involve breaking promises. Ross, like many other philosophers, appears to be more confident in rejecting a theory to the extent that he has several counterexamples of diverse kinds. I maintain that this pattern is explicable on Centrality, but not otherwise. To illustrate this, I will set up an artificial but representative case of this phenomenon, and then formally analyse the evidential relations at play.

Consider the following three propositions:

1. It is wrong to push a fat man off a footbridge with your hands in front of a speeding trolley to save five others.

2. It is wrong to open a trapdoor that causes this man to fall in front of the trolley.

3. If an anthropologist promises an indigenous assistant of his from the tribe he is studying that he will not photograph him, he ought to refrain from doing so even if the assistant is asleep and would not know he had been photographed (Foot 2001: p. 47).

Now imagine two moral philosophers, Sam and Vera. Sam has intuitions that (1) and (2) are true, and Vera has intuitions that (1) and (3) are true. (Sam, we can suppose, has not considered (3), and Vera has not considered (2).) We would expect Vera (like Ross) to be more confident in concluding that consequentialism is false on the basis of her intuitions than Sam. This is because Sam’s intuitions are about cases that are much the same whereas Vera’s are about cases that vary with respect to morally significant details.³²

But why does this kind of difference make Vera more confident in rejecting consequentialism than Sam? I maintain that Centrality can explain this and that it is not clear how it can be explained otherwise. In sum, given Centrality, we can offer the following reconstruction of Sam and Vera’s logic. Since Sam’s intuitions are much the same, his having the second one is all but ensured by his having the first one, even if consequentialism is true. So, given that he has the first one, his having the second one is not (much) more predicted by non-consequentialism than consequentialism, and hence is not (much) evidence against consequentialism. By contrast, Vera’s intuitions are varied enough that it is surprising that she has her second intuition, given that she has her first, if consequentialism is true. However, it is less surprising if consequentialism is false, because then her second intuition could be explained by her apprehending the moral facts. Hence, Vera’s second intuition is better predicted by non-consequentialism than consequentialism, and is evidence against consequentialism.

To be clear: I am reconstructing the tacit logic behind Sam and Vera’s epistemic practices, not endorsing that logic. If moral intuitions are not sensitive to the moral facts (and Sam and Vera know this), then Sam and Vera both err in lowering their credence in consequentialism at all after forming intuitions that (1)-(3) are true. Whether or not moral intuitions are actually evidentially relevant, the important point here is that the above explanation implies that Sam and Vera tacitly take them to be so. Inasmuch as it is not clear how to explain Sam's and Vera’s differing confidences otherwise, their differing confidences thus give us reason to accept Centrality.

To make clear why we can only explain Sam's and Vera's differing confidences by appeal to Centrality, I will now spell out the evidential relations Sam and Vera tacitly take to be at play more explicitly:³³

P(intuit(2) | ∼(2) & intuit(1)) ≈ P(intuit(2) | (2) & intuit(1)) ≈ 1, so that

P(intuit(2) | consequentialism & intuit(1)) ≈ P(intuit(2) | non-consequentialism & intuit(1)).

Conversely,

P(intuit(3) | (3) & intuit(1)) ≈ P(intuit(3) | (3) & ∼intuit(1)), and

P(intuit(3) | ∼(3) & intuit(1)) < P(intuit(3) | (3) & intuit(1)), so that

P(intuit(3) | consequentialism & intuit(1)) < P(intuit(3) | non-consequentialism & intuit(1)).

In words: given that one has an intuition that it’s wrong to push the man, it’s pretty likely that one will have an intuition that it’s wrong to open the trapdoor. The cases involve very similar kinds of factors, and so whatever led one to intuit that it’s wrong to push the fat man will likely also lead one to intuit that it’s wrong to open the trapdoor, even if it is not wrong (as consequentialism implies). So, that one has an intuition that (1) (approximately) screens off the truth of consequentialism³⁴ from the proposition that one has an intuition that (2); so, consequentialism does not significantly raise or lower the probability that one has an intuition that (2), relative to the fact that one has an intuition that (1).

Conversely, by Sam and Vera’s lights, (3) screens off the proposition that one has (or doesn’t have) an intuition that (1) from the proposition that one has an intuition that (3). That is, that one has an intuition that (1) only makes it likely that one will have an intuition that (3) by making it likely that (3) – that is, by being evidence for its content. (I am not, after all, inclined to judge that I ought to do just anything that doesn’t bring about the best consequences. So if (3) is not true, there’s no special reason to expect that I would have the intuition that it is, given just that I have the intuition that (1) is true.) So, one is less likely to have an intuition that (3) if ∼(3) than if (3), even given that one has an intuition that (1). So the truth of consequentialism – which implies ∼(3) – would make it significantly less likely that one would have this intuition (given, as seems plausible, that the truth of consequentialism doesn’t raise the probability of one’s having this intuition by some other means). So, the fact that one does have this intuition is significant evidence against consequentialism.

The difference between Sam and Vera, then, is that Sam tacitly takes his first intuition to (approximately) screen off the moral facts from his having the second intuition. By contrast, Vera tacitly takes the moral facts to (approximately) screen off her first intuition from her second intuition. So Sam does not take his second intuition to provide any greater disconfirmation of consequentialism, because that intuition was just as much to be expected given consequentialism as given non-consequentialism. But Vera does take her second intuition to provide more disconfirmation, because that intuition seems more surprising if consequentialism is true than if it is false, even given her first intuition.

This difference is related to the fact that when different intuitions supporting the same theory are about various cases, it appears harder to offer an error theory for all of them. If your error theory for the intuition that it’s wrong to push the fat man is simply that it involves killing a person, then you can explain the intuition that it is wrong to open the trapdoor, but not the intuition that the anthropologist ought not to photograph his native assistant.³⁵ You could give a separate error theory for that, but if your explanations for the intuitional data are all ad hoc and unrelated to each other, then the more plausible explanation is that the intuitions are correct.

The above account of why Sam and Vera would take themselves to have differing reasons to reject consequentialism presupposes that they are using their intuitions as evidence. Suppose that I do not take as evidence against consequentialism my intuitions that it’s wrong to push the man off the footbridge, that it’s wrong to open the trapdoor under the man, and that the anthropologist ought not to photograph his assistant. Instead, I only take as evidence against it the facts that it’s wrong to push the man, that it’s wrong to open the trapdoor, and that the anthropologist ought not to photograph his assistant. Given that these facts all entail the falsity of consequentialism, it’s unclear why on coming to believe the third I should lower my credence in consequentialism more than when I came to believe the second. The fact that given my first intuition-state, my second intuition-state is equally probable on consequentialism and non-consequentialism, but my third intuition-state is less probable on consequentialism than non-consequentialism should make no difference to my epistemic practices in the two cases, for I do not take my intuition states to bear in any way on the truth of consequentialism.

My discussion so far has focused on intuiting counterexamples to a theory. Similar remarks go for the converse phenomenon: philosophers are more confident in a positive theory when they have multiple independent arguments for it. If the first-order premises of those arguments are the only evidence they take themselves to have, this is a puzzling phenomenon. For, as above, each argument individually (let’s suppose) entails the positive theory; so why be more confident given more (and more diverse) arguments? But we can reconstruct what’s going on in these philosophers’ heads by appeal to logic similar to that glossed above. Informally, the relevant principle was expressed by Darwin (1876/2009: p. 278): a theory is more greatly confirmed to the degree that it ‘explains various large and independent classes of facts’. In the case of philosophy, the relevant facts are intuitions: the more of them there are and the more varied they are, the more support they provide for a theory.³⁶

In these last four sections I have canvassed several familiar facts about philosophical practice:

Fact 1: Philosophers frequently believe propositions they find intuitive.

Fact 2: Philosophers offer error theories for intuitions that they reject.

Fact 3: Philosophers are more confident in rejecting (accepting) theories when they have intuitions about diverse cases opposing (supporting) them.

The best explanation of these facts is that philosophers treat intuitions as evidence. When they become aware that they have an intuition, they become more confident in the content of that intuition and thus more confident in the theory that content supports. If they did not do this, the above facts would not be obviously explicable. These facts are, moreover, like (1) and (3) above, fairly different in kind apart from Centrality, so that a sceptical explanation of one of them is unlikely to explain the others. But that philosophers use intuitions as evidence neatly and simply explains these otherwise disparate facts in a natural and non-ad hoc way. This gives us strong reason to accept that philosophers indeed do use intuitions as evidence.

8. Objections

My argument is, broadly speaking, an inference to the best explanation: Centrality best explains salient features of philosophical practice, and so is very probably true. Critics, however, might maintain that there are other facts about philosophical practice that tell against Centrality. In this section I will consider two such alleged facts, and argue that they do not constitute strong evidence against Centrality. Doing so will further clarify what Centrality does and does not say and reinforce some earlier points.

The first alleged fact is that philosophers do not tend to make ‘I have an intuition that P’ an explicit premise in the arguments they put forward for their theories – instead, they just use the premise ‘P’. Moreover, they will not even use P as a premise of their argument if they think their interlocutors are unlikely to have the intuition that P is true (or accept P for other reasons). All this is surprising, the objection goes, if Centrality is correct.

In reply to this objection, it is important to distinguish Centrality (as I am understanding it) from closely related claims. My thesis is that in most cases philosophers are more confident in P when they take themselves to have an intuition that P than they are when they do not take themselves to have such an intuition (holding constant other background knowledge). My thesis does not imply that philosophers do not also use intuitive contents as evidence. I can use E as evidence for P relative to K and use P as evidence for T relative to K.³⁷ So my thesis does not predict that philosophers will not use intuitive contents as premises in their arguments.

My thesis also does not imply that for philosophers to use intuitions as evidence they must consciously infer that P from the fact that they are in an intuition-state with P as its content. So many philosophers may use intuitions as evidence without being aware that they are doing so.

Even if philosophers are aware that they use intuitions as evidence, this still does not imply that they will cite their intuitions as evidence in their published work. (Knowingly) using something as evidence oneself is distinct from offering something as evidence in a public discussion. It is true that the former will often make the latter more likely. But in some cases features of the dialectical context will make one unlikely to offer one’s private evidence as public evidence. In the case at hand, there are good explanations for why philosophers do not tend to cite their intuitions as evidence in their published work.³⁸

First, it is dialectically unhelpful to cite as evidence propositions that you know your interlocutors will not accept. So I may not cite P as evidence if I know my interlocutors do not find it intuitive, not because I do not take P to be evidence for my theory, but because it will be dialectically ineffective. Similarly, I may use my intuition that P as evidence for P even while knowing that my interlocutors will not be moved by the fact that I have that intuition.

Second, it would be cumbersome always to make explicit the fact that my reason for thinking that P is that I have an intuition that P. Frequently this will be clear from context (e.g., if P is a verdict about a thought experiment which I have offered no independent arguments for), and so I would not be conveying relevant information (and wasting valuable space) by making it explicit (cf. Grice’s (1989: p. 26) maxim of quantity).

The second alleged fact that is supposed to tell against Centrality is that philosophers do not become more confident in philosophical claims when learning that others have intuitions that they are true (Earlenbaugh and Molyneux 2009: pp. 98-9). But, so the objection goes, if philosophers treated their own intuitions that P as evidence for P, they would also treat others’ intuitions that P as evidence for P. Since they do not do the latter, they must not do the former either.

It is empirically doubtful that philosophers never give others’ intuitions any evidential weight. We frequently ask our colleagues for their intuitive reactions to cases. However, I agree with the objection that philosophers tend to be less moved by other philosophers’ intuitions. I maintain, though, that there are Centrality-friendly explanations for this.

First, that I have an intuition that P is verifiable by me in a more direct way than that you have an intuition that P. It is possible, from my perspective, that what appears to be an intuition-report from you is in fact something else: perhaps you are reporting a theory-based conviction in a misleading way (either intentionally or through carelessness). Even if you do have the intuition, perhaps you are exaggerating its strength for dialectical effect. In such cases it is natural for me not to accord your apparent intuition-report much evidential weight.

Second, as I have stressed throughout, the evidential strength we attach to intuitions is context-relative.³⁹ I may take my intuition that P as evidence for P relative to my initial background knowledge but not relative to my new background knowledge which includes the error theory you have proposed for P. And typically the context of learning that you have an intuition that P will be different from the context of learning that I have an intuition that P. If I learn that you have an intuition that P, I will usually have already considered P, and either intuited it or not.⁴⁰ Relative to the background against which I have already intuited it, your additional intuition may raise my credence slightly, but it will usually not raise it as much as my intuition did, especially if my credence relative just to my intuition is already near 1. And relative to the background against which I have not intuited it, I have evidence against P which will make me suspicious of the reliability of your intuition. If I can come up with an error theory for your intuition, this will make me even less inclined to treat it as evidence for P. Indeed, to reiterate one of the main points from §§5-6, to the extent that I feel compelled to construct such an error theory, I am treating your intuition as evidence (relative to my initial background knowledge) – for I am conceding that your intuition would make P more probable if it cannot be explained except by P’s truth.

Finally, even if philosophers ought to be equally moved by others’ intuitions, philosophers are (alas) not completely rational. The objection at hand might be right that if intuitions are evidence, yours are as good as mine. But it does not follow that if I use my intuitions as evidence, I similarly use yours: psychological biases might lead me to treat mine differently from yours.

9. Conclusion

My argument in this paper has been that philosophers use intuitions as evidence. I have not made any claims about whether intuitions are in fact evidence. However, my thesis does have normative implications. A higher conditional credence in a theory T on E&K than on K alone is rational only if E raises the probability of T relative to K. Hence, using intuitions as evidence for T is rational only if those intuitions raise the probability of T. Thus, philosophers who reason in this way are committed to the view that intuitions like this make their theories more probable – and are thus committed to viewing intuitions as evidence. So philosophers who reason in the manner described in this paper ought to either stop doing so or accept that intuitions in fact are evidence in philosophy. If they deny that intuitions are evidence but continue treating them as such, they are being inconsistent.

This implies that philosophers who use intuitions cannot shrug off experimental philosophy’s critique of intuitions in the way many are wont to do. Experimental philosophers who purport to show that intuitions are unreliable are challenging something central to philosophical methodology.

More generally, those of us who rely on intuitions in our theorizing can’t beg off sceptical worries about the connection between psychological states and facts about the world. We don’t necessarily need a full account of how the former are connected to the latter to do philosophy, but we should be sensitive to the demand for one, and careful about relying on intuitions when it seems unlikely that there is any such story (e.g., intuitions about which possible worlds are ‘nearer’ to us than others). If it is unlikely that there is any connection between my intuitions about X and the facts about X, then my intuitions about X can’t be explained by the facts in such a way as to make them probable for me.

In this paper, I have argued that most philosophers use intuitions as evidence for their philosophical theories: they are more confident in their theories when they take themselves to have intuitions that propositions supporting those theories are true. In support of this thesis I canvassed three facts: philosophers believe intuitive claims, philosophers offer error theories, and philosophers are more confident in theories supported by diverse intuitions. Various sceptical worries about whether these practices are rational and whether intuitions are really evidence can be raised. I have not addressed those worries here. But my arguments imply that philosophers cannot ignore them, for the use of intuitions as evidence is central to contemporary philosophical practice.

Footnotes

1

I say this seeming because it might seem to one that P both intuitively and as the result of an independent inference. (For example, one might find a mathematical proposition intuitive and then prove it; it would then seem true in both these ways.)

2

More cautiously: it may be that the only seemings that satisfy (ii) and (iii) are conscious inclinations to believe. It is compatible with this that there are other seemings that are not conscious inclinations to believe, such as the perceptual seeming that a stick in water is bent. My thanks to an anonymous reviewer for pushing me to clarify this.

3

If a seeming is a conscious inclination to believe, then we could rephrase (ii) and (iii) to say that if one were to reflect on whether one’s conscious inclination to believe that P is a result of inference/memory/sense experience/testimony, one would not be conscious of an inclination to believe that it is. Perhaps Williamson and Cappelen would say that this is still too open-ended, and that intuitions thus defined are not unique to philosophy, but ubiquitous in ordinary life. But nothing in my argument presupposes that intuitions are unique to philosophy, and so my argument is compatible with this.

4

The probability here is epistemic probability – roughly, the epistemic probability of P given K is the degree to which an agent with knowledge K ought to believe P. On my understanding of epistemic probability here, a priori truths and falsehoods (including mathematical and conceptual claims) can have epistemic probabilities between 0 and 1. For instance, my epistemic probability that the one-trillionth digit of pi is 2 is .1. For more on the nature of epistemic probability, see Keynes 1921, Swinburne 2001: ch. 3.

5

If K is what one currently knows (or believes to degree 1), then one’s conditional credence in T given K is simply one’s current credence in T. One’s conditional credence in T given E&K is the credence one would have in T were one to also believe E to degree 1, i.e., conditionalize on E. Conversely, if one already believes both E and K to degree 1, so that one’s conditional credence in T given E&K is one’s current credence in T, we can understand one’s conditional credence in T given K to be the credence one would have in T if one were to cease believing E to degree 1, and believe it instead to whatever degree one believed it before. There are sticky issues here, related to the old-evidence problem (Glymour 1980: pp. 85-93) of how to spell out the latter case in particular; I set those aside here.

6

Cappelen says that it has been assumed by ‘practically all participants in [metaphilosophical] debates’, but complains that ‘no one has ever presented a detailed case for’ it (2014a: p. 513, emphases in the original). Whether or not he is right about the former, this paper will hopefully falsify the latter claim.

7

Some of these philosophers, Chudnoff for example, defend views on which intuitions play an important justificatory role in philosophy. They all maintain that intuitions are not used as evidence, however. See also the distinction between intuition states and intuitive contents in the next paragraph and footnote 8.

8

In suggesting that these philosophers are motivated to deny that intuitions are used as evidence by a conviction that intuitions aren’t really evidence I am not saying that in most cases this is their stated motivation. However, the fact that so many philosophers make both of these claims, and the complete absence (as far as I know) of philosophers who think that intuitions are evidence but are not taken to be such, suggests that many philosophers’ beliefs that intuitions are not evidence are leading them to deny that they are used as such. This is similar to the way in which philosophers are motivated to deny classical foundationalism to avoid radical scepticism, and to endorse compatibilism to avoid the conclusion that we have no free will. In neither case are these usually philosophers’ stated motivations (though occasionally they are), but it is quite plausible that they largely explain the popularity of the above views. (For the first point, see Fumerton and Hasan 2010: §3; for the second, see Nichols 2007.)

9

Other philosophers make similar points. Even prominent defenders of the use of intuition in philosophy like Pust and Bealer are reticent to say that we use intuition states as evidence. Bealer (1998) says that ‘When I say that intuitions are used as evidence in philosophy, I of course mean that the contents of the intuitions count as evidence’ (p. 205, emphasis in original). Pust (2014: §2.4) similarly says that ‘most of our putative evidence in philosophy consists of non-psychological propositions’, but suggests (contrary to Ichikawa) that ‘what qualifies such propositions as our evidence’ – what justifies us in believing them – is the fact that we intuit them.

10

These, Cappelen thinks, are the ordinary English meanings of ‘intuitively’.

11

Cappelen (2012: p. 13) understands Centrality more weakly than I do. He understands it as claiming that philosophers use intuitions as evidence or as sources of evidence; on the latter view it is intuitive contents that are evidence. I am using ‘Centrality’ to denote the stronger claim that intuitions (i.e., intuition states) are themselves used as evidence in philosophy.

12

I am open to the claim that intuitions are used differently in some subfields in the sense that philosophers are more willing to use particular kinds of intuitions as evidence than particular other kinds. It is plausible that epistemologists and metaphysicians tend to give intuitions more evidential weight than ethicists (cf. Weatherson 2003: pp. 1-2); and error theories about deontological intuitions are probably more popular than error theories about Gettier intuitions. My thesis in this paper is simply that philosophers typically do use intuitions as evidence; I think this is true of philosophers in all subfields (although I do not claim that it is true of all philosophers), and hence I consider examples from several different fields below. It is compatible with this that this practice is less typical of some subfields than others, and that philosophers vary in the evidential weight they attach to intuitions vis-à-vis other considerations (e.g., theoretical parsimony). An interesting further sociological task would be to examine what differences of this kind exist and why.

13

This section benefited greatly from comments by an anonymous reviewer.

14

I am grateful to an anonymous reviewer for pointing me to both this Deutsch paper and the above quote from ‘Identity and Necessity’. Here is fuller context for the Kripke quote:

According to this view, whenever, for example, someone makes a correct statement of identity between two names, such as, for example, that Cicero is Tully, his statement has to be necessary if it is true. But such a conclusion seems plainly false. (I, like other philosophers, have a habit of understatement in which ‘it seems plainly false’ means ‘it is plainly false’. Actually, I think the view is true, though not quite in the form defended by Mrs. Marcus.) At any rate, it seems plainly false, (emphasis in original).

As this fuller context makes clear, Kripke is here contrasting this common use of ‘seems’ with his use in this case. So in this case Kripke is not using ‘seems’ as a hedge. However, he could arguably be using it in Cappelen’s ‘pre-theoretic’ sense, to mean ‘pre-theoretically, we would judge this to be false’.

15

Another passage like this is on page 42 of Naming and Necessity, where Kripke seems to explicitly say that he takes intuitions to be evidence. Bengson (2014: pp. 215-6) cites this passage in support of Centrality. Cappelen (2014b: p. 597) and Deutsch (2015: pp. 21-4), however, raise legitimate doubts about its interpretation.

16

Cappelen could hold that Kripke’s belief that ‘Gödel’ refers to Gödel is based on inference (as Deutsch (2009: pp. 451-2, 2015: pp. 9-17) suggests), so that any inclination to believe this proposition would not satisfy my condition (ii), and so not be an intuition. But plausibly many other philosophers who have considered the proposition have come to believe it non-inferentially. At any rate, not all philosophical beliefs can be based on inference, on pain of regress. There must be some philosophical beliefs that are not based on further philosophical beliefs. In most of these cases philosophers, if they consider the proposition in question, both believe it and feel an inclination to believe it. If intuitions are conscious inclinations to believe, then in these cases philosophers will both believe the proposition in question and have an intuition that it is true. So Cappelen should still grant the more general phenomenon of philosophers believing intuitive propositions, even if he were to hold that Kripke’s belief that ‘Gödel’ refers to Gödel is not an instance of this phenomenon.

17

In spite of the liberality of my definition of intuitions, some philosophers may still be sceptical that there are any phenomenal states of the kind I have attempted to describe, and so deny that Fact 1 is a fact. But even these philosophers cannot reasonably deny Facts 2 and 3 which I go on to adduce – that philosophers offer error theories and that they are more confident when they have diverse cases supporting a theory than otherwise. And, as I argue, these facts can only be explained by supposing that philosophers use intuitions as evidence. Thus, Facts 2 and 3 provide non-question begging grounds supporting Centrality for even the most extreme intuition-sceptic.

18

In this essay, when I speak of philosophers’ beliefs (e.g., in intuitive propositions), I simply mean what they have a credence above .5 in. So when I say that philosophers tend to believe what’s intuitive I just mean that they tend to be more confident in intuitive propositions than in their negations.

19

One explanation of the correlation is that both beliefs and intuitions are explained by the a priori facts to which they are sensitive. However, as an anonymous reviewer pointed out to me, we might not want to think of this explanatory connection as causal, since arguably a priori facts do not stand in causal relations. In the text I retain the language of ‘common cause’ for ease of exposition, but I wish to allow for the possibility of non-causal common explanations of this kind as well.

20

Perhaps philosophers initially had implicit beliefs about reference (inconsistent with their explicit descriptivist beliefs at that time) which caused their intuitions about Kripke’s case. (My thanks to an anonymous reviewer for suggesting this possibility.) However, this does not yet explain the correlation of philosophers’ intuitions with their explicit beliefs, which are my focus here. This correlation would need to be explained either by these intuitions then causing new explicit beliefs, or by the implicit beliefs causing the explicit beliefs. So either intuitions will cause explicit beliefs, or the two will have a common cause. I consider these two cases below.

21

This may seem especially plausible if one thinks that intuition-states are conscious inclinations to believe (cf. Earlenbaugh and Molyneux 2009).

22

Contextualists about knowledge-attributions (like Cohen 1999, who formulated the flight case) also use these cases to argue for their view. Following Turri, I focus on the pragmaticist explanation here.

23

Unger is perhaps even more explicit than Kripke that he is using intuitions as evidence. In the chapter before this, Unger is discussing why it seems that we ought not to be certain that there is no evil scientist deceiving us into falsely believing such propositions as that there are rocks. He says:

These are, I suggest, our intuitions in these matters. All of them cry out for explanation, and any sceptic should be interested in explaining them. By the same token, so should any other philosopher. The difference between the two here is only this: these intuitions favour a sceptic's case. … On the other hand, the philosopher of ‘common sense’, who tries to reverse our classical argument, now has some very damaging evidence stacked against his attempt. Quite plainly, there is no intuition here to favour his case, to hold out the opportunity for him to give a favourable explanation. Rather, there are here only feelings which cut the other way, and he must try to dispel, and to ‘explain away’, these distinctive and intuitive feelings. (1975: p. 26, emphasis in original)

24

Error theories of the kind I am concerned with here should not be confused with the similarly-named moral error theory associated with Mackie (1977). The error theorist in Mackie’s sense denies the existence of the putative objects of a domain of discourse: e.g., in the case of ethics, objects like moral rightness or obligation. This error theorist has a theory that moral discourse (say) is an error. The error theorist in my sense has a theory about error: in particular, a theory about why a certain error is made. This error theorist seeks to offer a theory of why someone has the intuition that P even though ∼P (i.e., even though the intuition in question is an error).

25

Assume, for the sake of argument, that we’ve specified the footbridge scenario finely enough that the consequentialist agrees that pushing the man would have the best consequences, and so would be the right thing to do.

26

Formally: ⁠. This is the Odds Form of Bayes’ Theorem. We can see from this theorem that the posterior probability of T given E is a function of the prior probability of T and the relative degree to which T and ∼T make E probable (the Bayes’ Factor). If E raises the probability of T, this means that P(E|T&K) > P(E|∼T&K), so that the Bayes’ Factor is top-heavy. When Deon treats E as evidence for T, he relies on a tacit assessment of this ratio as top-heavy. Connie, in offering her error theory, is arguing that P(E|∼T&K) is fairly high. (This will follow if P(T*|∼T&K) and P(E|T*&∼T&K) are both high.) Thus, the Bayes’ Factor will not be very top-heavy, and E will not be strong evidence for T.

27

I do not mean to claim that philosophers explicitly think through all of this when offering and responding to error theories. I am rather offering a formal reconstruction of their tacit epistemic practices in such situations.

28

Philosophers sometimes do offer error theories for intuitive contents. For instance, Connie could argue that, contrary to appearances, pushing the fat man would have overall worse consequences, so that consequentialism in fact does explain why it would be wrong to push the fat man. But this is a different error theory than the one she in fact does offer. All I need to claim is that many philosophical error theories seek to explain away intuition states, and this is true of all the examples I have mentioned so far in the text.

29

As before, T could be a more general theory that is not immediately intuitive, but I am considering the limiting case where T is itself the content of an intuition.

30

Cappelen seems to think that error theorizing about intuitions that P only takes place after the error theorizer has presented a compelling argument that ∼P (p. 92). This helps make sense of his claim that error theories are not necessary to reject a view. It is indeed true that error theorists frequently offer other reasons to reject the theories they oppose, or accept their incompatible theories. But these reasons are not always conclusive, and in at least some cases philosophers treat error theories as themselves providing reason to reject the view they are arguing against. It is plausible that in giving their error theories, the error theorists about Low Flight and High Flight take themselves to be offering some reason to reject the claim that knowledge depends on practical stakes, Grice takes himself to be offering some reason to reject Ryle’s claim that morally neutral actions cannot be voluntary or involuntary, and Parfit takes himself to be offering some reason to reject subjectivism about reasons.

31

It is worth noting that Ross is explicit that he thinks intuitions are evidence in moral philosophy:

We have no more direct way of access to the facts about rightness and goodness and about what things are right or good, than by thinking about them; the moral convictions of thoughtful and well-educated people are the data of ethics just as sense-perceptions are the data of a natural science. (Ross 1930: ch. II, emphasis added).

Matters are slightly complicated, however, by his assertion shortly before this sentence that sometimes we know the contents of our ordinary convictions, and thus that those contents can be our starting points in ethics (cf. the discussion of Ross in Audi 2015: pp. 58-62).

32

If it is not obvious that Vera would be more confident in rejecting consequentialism than Sam, it may help to compare both with another moral philosopher, Won, who has only considered (and endorsed) (1): it is wrong to push the fat man. Other things being equal, we would expect another philosopher who has considered multiple counterexamples to consequentialism to be more confident in rejecting consequentialism than Won. So Vera, with her two varied counterexamples, will be more confident in rejecting consequentialism than Won. But it does not appear that Sam will be much more confident in rejecting consequentialism than Won. That is because his second counterexample is a trivial variation on Won’s. So, since Vera is more confident in rejecting consequentialism than Won, and Sam is not, Vera is more confident in rejecting consequentialism than Won.

33

I suppress the background K here for ease of notation.

34

The terminology of ‘screening off’ derives from Reichenbach (1956). Formally, B screens off A from C if P(A|B&C) = P(A|B).

35

I am not denying that more sophisticated error theories could explain both these intuitions, simply noting that it will be more difficult.

36

See McGrew (2003) for a formal proof of Darwin’s principle.

37

This does not in general imply that I use E as evidence for T. However, in ordinary philosophical cases in which I use my intuition as evidence for its content and that content as evidence for a theory, this kind of transitivity will hold, inasmuch as the content will ordinarily screen off the intuition from the more general theory the content supports. See Shogenji (2003) for the formal details on the transitivity of evidence under screening off.

38

I should also note that some philosophers (e.g., Ross, Unger, and Kripke, all mentioned previously) do explicitly cite intuitions as evidence in their published work. Also, that philosophers give error theories in their published work implies that there is a tacit understanding that intuitions are being used as evidence in the debate they are engaging in, even if those intuitions are not being explicitly cited as evidence by the participants in the debate.

39

Earlenbaugh and Molyneux do not seem to recognize this possibility: ‘either a given kind of mental state is evidence for its contents, or it is not’ (p. 99). But some instances of the mental state could be evidence in some circumstances, without all instances being evidence in all circumstances.

40

Exceptions would be cases where I am unable to understand P as well as someone else or recognize myself as less reliable about such matters. However, here we are much more likely to treat others’ intuitions as evidence. For example, were I to show my brother (a professional mathematician) some putative theorem which I have no intuition about and he were to say, ‘this seems true’, I would take that as evidence for it.