Defending a Realist Stance

ABSTRACT

Should the scientific realist admit that their realism involves what Chakravartty has called an epistemic stance? I argue that the realist should accept the need for a realist stance that licenses the use of inference to the best explanation. However, unlike Chakravartty, I maintain that the realist should insist that their realist stance is rationally obligatory. This requires an anti-voluntarism about stances that involves theoretical reasons for adopting one stance rather than another. I present one account of what these reasons might be that is tied to principles that identify some strong connections between evidence, knowledge and genuine explanation.

1. Introduction

Let us suppose that a scientific realist is a person who claims to know of the existence and character of some unobservable entities. To defend their scientific realism, the realist must offer a non-trivial and non-circular argument supporting their position. A defense counts as non-trivial when it involves substantial claims that clarify the basis for the knowledge that is claimed. And a defense is non-circular when it does not presuppose either the knowledge or the existence of the unobservables in question. In this essay I consider how epistemic stances should figure into a defense of scientific realism. A rough characterisation of an epistemic stance is that it is a collection of policies that settle how evidence supports new knowledge claims. In section 2 I argue that a scientific realist must adopt a kind of epistemic stance that I call a realist stance. This conclusion agrees with other discussions of epistemic stances, such as those found in van Fraassen and Chakravartty (Chakravartty 2017b; Van Fraassen 2002).

In sections 3 and 4 I argue for a kind of anti-voluntarism about realist stances: the scientific realist should insist that their realist stance is rationally obligatory. The appropriateness of their stance is not tied to their choices, values or attitudes, but is instead argued to be mandatory. There are two ways to reach this conclusion. The first argument is that a scientific realist who claims that their realist stance is merely rationally permissible lands in pragmatic incoherence (section 3). This makes their scientific realism indefensible. The second argument involves a diagnosis of the source of this incoherence in the relations that obtain between evidence, knowledge and explanation (section 4). There are principles connecting evidence, knowledge and explanation that clarify the mandatory basis of a realist stance. Articulating these principles thus turns out be to a crucial part of a defense of scientific realism. This paper concludes in section 5 by considering some objections and replies to this approach to scientific realism.

2. Epistemic Stances

According to Chakravartty, an epistemic stance ‘incorporates an epistemic policy or collection of policies, reflective of and including various attitudes and commitments regarding how one learns things about the world’ (Chakravartty 2017a, 227).¹ The policies involve ‘which principles and methodologies are appropriate or inappropriate to producing knowledge’ (Chakravartty 2017a, 227). There are thus two questions we can ask about a given epistemic stance. First, what epistemic policies are involved, and, second, how do those policies reflect attitudes, commitments or values? For the first question the key issue is how an epistemic policy settles the relationship between evidence and what one may claim to know. In the philosophy of science, both the realist and the antirealist endorse ampliative inferences where the conclusion of the inference says more than the premises. However, only the realist adopts inference to the best explanation (IBE), at least for claims to know about unobservables: when one’s evidence includes that there are phenomena P₁, P₂, … , P_n whose only available potential explanations are E₁, … , E_r, and of these E₁ is the best, one may claim to know E₁. These available potential explanations are explanations that have been proposed, and that may or may not be genuine. At most one of these proposed explanations can be true. The basic idea behind IBE is that someone can come to know the truth by considering these potential explanations (Lipton 2004). Some anti-realists might also endorse the use of IBE, but this use is typically severely limited. As a result, when a realist and an anti-realist start reasoning from some common basis of evidence, the realist will often claim to know more about the world than the anti-realist.

The main consideration in favour of epistemic stances is that there is a gap between one’s evidence and what that evidence is evidence for. Here I will suppose that one’s evidence is some collection of propositions. Different accounts of evidence pick out these propositions in different ways. As the scientific realist claims to know about unobservables, the appropriate notion of evidence to work with maintains that if E is evidence for P, then believing E provides a sufficient basis to claim to know P. Different accounts of evidence flesh out this conditional in different ways, but I will not take a stand on this contentious question.²

When a scientist puts forward a claim to know, we expect them to base this claim on their evidence. If this evidence is marshalled in a selective or biased way, then the scientist has done something wrong. But if, after reflecting on all their available evidence, the scientist reaches an all-things-considered judgment that their evidence is evidence for P, then the scientist has acquired a reason to claim to know P. This does not mean that every element of their evidence favours P or indicates that P is true. What is required is that the totality of their evidence provide this sort of all-things-considered basis to conclude that P is true. Only then is the scientist acting appropriately when they go on to claim to know P.³

To consider a concrete case, recall Franklin’s famous kite experiment. As Franklin described the experiment in his 1752 letter to Collinson, Franklin flew a kite during a thunder storm. The kite was tied to twine, and a key was attached to the twine by a silk string. As the storm developed, the key exhibited all the signs of being electrified, e.g. ‘you will find it [the electric fire] stream out plentifully from the key on the approach of your knuckle’ (Franklin 1941, 266). After this experiment was conducted, Franklin had a wide range of propositions that any proposal should count as his evidence. Using this evidence, Franklin concluded that (L) lightning is an electrical discharge. The realist aspires to treat this case as one where Franklin’s evidence provided him with a reason to claim to know L. For this to occur, it is not enough for Franklin to simply have this evidence, e.g. believe or even know all these propositions. For Franklin might believe or know these propositions and still lack a reason to claim to know that lightning is an electrical discharge. In order to acquire a reason to claim to know L on the basis of his evidence, Franklin must grasp or recognise that his evidence is evidence for L. I claim that a realist epistemic stance that includes IBE is essential to this grasp or recognition. The policy to employ IBE is part of that stance and provides Franklin with his reason to claim to know L.

Let us suppose that Franklin has adopted a realist stance that includes IBE.⁴ This makes it possible for Franklin to be aware of a principle of inference that links his evidence to L, and this is what is involved in Franklin acquiring a reason to claim to know L on the basis of his evidence. Here I am supposing a somewhat demanding notion of what it means to say that E is evidence for P: E is evidence for P only when E provides a reason to claim to know P. On this reconstruction, the reason that the evidence provides does not arise exclusively from the content of the evidence. Another sort of contribution is required from the principles of inference that the agent has adopted. Whenever such a principle is part of an agent’s stance it is possible, at least in principle, for them to recognise that their inferential principles license a given claim to know. On the broadly internalist approach that I am arguing for here, this is the only way for the agent to acquire a reason to claim to know something whose content goes beyond their initial evidence.

One might think that something else besides an epistemic stance could lead a scientist like Franklin to claim to know L when their evidence includes the propositions that Franklin had in his evidence. One proposal is that Franklin merely needs to be disposed to claim to know L when he has these propositions in his evidence. This disposition might involve much less than a standing policy to employ IBE in this and similar circumstances. For example, it could be an unthinking reaction that is similar to a reflexive pain utterance. I maintain that such a disposition is unable to give Franklin a reason to claim to know. This becomes clear when we imagine how Franklin could respond to a challenge to his claim to know L. If this claim issued merely from a reflexive disposition, then he would be unable to back up his claim to know L, and his colleagues would conclude that he did not know L. By contrast, if Franklin’s claim to know resulted from the application of his policy to employ IBE, then he should be able to articulate these considerations and present them when criticised. The same point applies to any proposal that maintains that the mere reliability of this inference is sufficient for Franklin to acquire a reason to claim to know L. If the inference in question is in fact reliable, then Franklin’s conclusion is very likely to be true. But Franklin will be unable to address doubts about the inference, and so his colleagues will conclude that he does not know L. The policies making up an epistemic stance are essential for someone to have a reason to claim to know because it is their potential awareness of these policies that allows them to defend their claims to know when challenged.⁵

As we saw at the beginning of this section, for Chakravartty the epistemic policies that make up an epistemic stance are ‘reflective of and including various attitudes and commitments regarding how one learns things about the world’. These attitudes, commitments and values typically motivate an agent to adopt a determinate epistemic stance. For example, it might be that someone places a special value on precise experimental results, while someone else values comprehensive theoretical explanation. These differences in value are sometimes reflected in the personalities of actual scientists from the history of science. Chang relates how Berthelot characterised Regnault: ‘It seemed that the very spirit of precision had been incarnated in his person’ (Chang 2004, 75; Langevin 1911, 44). By contrast, Einstein claimed that a scientist like Planck

tries to make for himself in the fashion that suits him best a simplified and intelligible picture of the world; he then tries to some extent to substitute this cosmos of his for the world of experience, and thus to overcome it. (Einstein 1982, 225)

In some cases, at least, the values of a scientist help to explain their epistemic stance: Regnault’s preoccupation with precision goes along with his positivism or anti-realism, while Planck’s value for explanation makes sense of his scientific realism. Debates between anti-realists and realists can thus be partly explained by connecting these debates back to a difference in epistemic stances, which in turn is often based on a difference in values.

3. The Pragmatic Incoherence of Stance Voluntarism

Now that the character of epistemic stances is clear, we can consider how best to think about the basis for an epistemic stance. Chakravartty is a voluntarist about epistemic stances. I take a voluntarist about stances to claim that there is a collection of viable stances, and one may be perfectly rational and adopt any of the stances in this collection. So voluntarism about stances requires one to admit that there is no reason in favour of one’s own stance. By ‘no reason’ here and in what follows, I mean that there is no rational obligation to adopt that stance, and so nothing irrational about adopting some other stance. There is a more expansive notion of reason that includes practical reasons. One has a practical reason to perform an action when it promises to satisfy one’s desires or to maximise expected utility. I take it that Chakravartty’s voluntarism about stances is consistent with there being practical reasons to adopt a given stance, namely that it helps to satisfy one’s desires. My focus is on theoretical reasons to adopt a stance like the realist stance.

Given the connections between an epistemic stance and a person’s attitudes or values, it might seem obvious that one should be a voluntarist about epistemic stances. To see why, let us make the simplifying assumption that one’s values uniquely determine one’s epistemic stance. So, to continue the example from the end of section 2, once we fix Regnault’s values, there is only one epistemic stance that he will adopt, and once we fix Planck’s values, there is only one epistemic stance that he will adopt. In addition, many suppose that are no theoretical reasons for Regnault to have the values that he has, and no theoretical reasons for Planck to have the values that he has. So, assuming that values determine one’s stance, and that there are no theoretical reasons for one’s values being what they are, it follows that there are no theoretical reasons to adopt a given epistemic stance. This is voluntarism about epistemic stances.⁶

I reject voluntarism about epistemic stances, and so I must reject at least one step of this argument. The key problem is that the argument assumes that there no theoretical reasons for one’s values. This is a contentious and problematic assumption. Most people assume that at least some values can be supported by theoretical reasoning. For example, moral values are often taken to be properly based on theoretical reasoning. This is consistent with many people adopting moral values without any theoretical basis. I would maintain that the same point applies to epistemic values, such as the value that one attaches to having a scientific explanation. That is, just as people should adopt some moral values (and not others), people should also adopt some epistemic values (and not others). Perhaps Planck adopted his epistemic values without any theoretical reasoning. But these values could still be supported by theoretical reasoning. If this sort of support is available, then there will be theoretical reasons in favour of a realist stance. I develop this sort of argument in section 4.

Some problems created by voluntarism are especially severe for the realist. For if the realist admits that their realist stance is voluntarily adopted, then they cannot claim to know of the existence of unobservables. That is, there is an instability for a realist who both admits the need for a realist stance and also concedes that their stance is voluntarily adopted in the sense that it has no theoretical reasons supporting it. As Psillos has put the point: ‘if adopting a stance is governed by voluntarism, there cannot be any knowledge. ‘I do voluntarily believe that p; hence I know that p’ is a fallacy’ (Psillos 2021, 27).⁷ Psillos strengthens this position further by paraphrasing Bernard Williams from ‘Deciding to Believe’: ‘it is (pragmatically) incoherent to say that I believe at will’ (Psillos 2021, 27, fn. 34).⁸ Here is an argument from Williams for this conclusion: to believe something is to believe it is true. So any mental state that is formed without regard to truth, but instead to satisfy some desires unrelated to a desire for truth, is not something that I can regard as a belief, at least if I am aware that this is how the belief was formed. A key premise here is that I would know if a mental state was adopted through an act of will that is not directed at getting at the truth. Given this knowledge, I cannot sincerely maintain that the mental state is a belief, i.e. as Williams puts it, ‘something purporting to represent reality’ (Williams 1973, 148).

There is a connection between Williams on deciding to believe and Chakravartty’s voluntarism about stances. A realist who is also a voluntarist about stances will admit to their own realist stance, but also allow that there is no reason that obliges them to adopt their realist stance. Consider a claim to knowledge that the realist advances on the basis of their evidence and their realist stance, such as Franklin’s L. A realist who is also a voluntarist about stances and who reflects on this situation is immediately landed in a pragmatically incoherent situation that parallels what Williams describes. First, they put forward the claim L as something they believe to be true. Second, they are aware that this belief is due in part to their realist stance. Third, as a voluntarist about stances, they admit that they have not adopted their realist stance on the basis of any reasons that reflect the truth.

As we have seen, the voluntarist explains the adoption a stance by appeal to the person’s desires or values. If these desires or values have no connection to the truth, then the realist must admit that their resulting beliefs are not appropriately connected to the truth, and so not known. What if their desires or values are connected to the truth? For example, we attributed to Planck a desire to explain the hidden structure of the universe. If Planck appealed to this desire and its satisfaction when defending some belief that is tied to his stance, he would be forced to admit that satisfying this desire does nothing to connect this belief to reality. That is, unless there are theoretical reasons for a realist stance (or the underlying values that determine that stance), there is no way for the realist to employ the stance’s policies to claim to know. So, just as with Williams’ case of acquiring a belief at will, the realist must admit to themselves that they know that they could acquire this belief whether it was true or not. This is the pragmatic incoherence: by the realist’s own lights, one of their beliefs, which they take to be true, is also something that they admit to themselves that they would have whether it was true or not. The only way to restore coherence and to maintain one’s voluntarism about one’s stance is to withdraw any claim to know that is based on one’s stance. For the realist, this means abandoning their realism.⁹

As I see it, there are two ways to resist this conclusion. One could object to Williams’ original point about deciding to believe. There is of course an ongoing debate about the possibility of deciding to believe. It is not feasible to enter into this debate here, except to say that a scientific realist should adopt the ‘evidentialist’ position on questions like Pascal’s Wager. Otherwise, they will be vulnerable to antirealist concerns that their claims to know are not supported by adequate evidence.¹⁰ Suppose, for example, that a realist claimed to know that the universe is eternal on the basis of the fact that humans will be much happier if they believed that the universe is eternal. This argument parallel’s a version of Pascal’s Wager that maintains that we know God exists because humans will be much happier if they believed that God exists. The realist claim to know that the universe is eternal is clearly vulnerable to the worry that it is not sufficiently supported by scientific evidence. There is no connection between the realist’s evidence and what they are claiming to know. More generally, a realist needs to insist that their scientific claims are adequately supported by their evidence.

Another objection would grant Williams’ point about deciding to believe and yet maintain that voluntarism about stances is not analogous to deciding to believe. It might seem that the whole discussion of this section trades on a confusion between voluntarism about belief and voluntarism about stances.¹¹ As Chakravartty is at pains to emphasise, ‘The idea of voluntary commitment in the context of scientific ontology applies to stances in the first instance, and only thereby, secondarily or indirectly, to belief’ (Chakravartty 2017b, 219, emphasis original). This suggests that even if voluntarism about belief is problematic, there is no reason to think that voluntarism about stances is problematic. My response to this objection is to repeat that voluntarism about stances is the claim that there is no theoretical reason to adopt one stance rather than another. Anyone who admits this is thus forced to admit that the input that their stance provides to their claims to know is unrelated to the truth. So in this respect the role of their stance is analogous to the role of a desire that is used to justify a claim to know. In both cases, no claim to know is defensible.

Here it is important to distinguish two scenarios. There is one scenario where the realist relies on a realist stance, but is not aware of this or forgets the significance of their stance when they reflect on their beliefs, such as the claim to know L. In this scenario, let us suppose that voluntarism about stances is correct, and so the realist has no theoretical reason for adopting their realist stance. But given that they are not aware of the relevant details of the situation, they avoid pragmatic incoherence. That is, it seems to the realist that everything they believe is just based on evidence and theoretical reasons. I would concede that in this scenario the realist avoids pragmatic incoherence. However, it should also be plausible that in this scenario any defense of realism that this realist offers is fundamentally flawed. They claim that their realism is based on evidence and theoretical reasons, while it is in fact based in part on a realist stance that they have adopted without regard for theoretical reasons. They are confused as the only reasons they have to adopt their stance (if voluntarism is correct) are practical reasons. So even though this realist can maintain their realism with an easy conscience, a confusion or self-deception remains that compromises their epistemic situation.

In the second scenario, I imagine a realist who is aware of their realist stance, its essential role in their realist claims to know, and who also maintains voluntarism about stances. That is, they admit the crucial role of their realist stance in supporting their claims to know, and also maintain that there is no reason for one to adopt such a stance. How could this realist sincerely maintain that L is known to be true? They admit that the antirealist who starts from the same evidence as they do is fully entitled to complain that the evidence does not support this claim to know. And they also suppose that the only difference between themselves and the antirealist is that they have adopted an optional, additional stance tied to their practical reasons. This is very similar to the pragmatic incoherence that Williams identified: if one admits that one’s evidence does not adequately indicate the truth of their claim to know, and that their belief is tied to their desires, then they cannot sincerely believe that this claim is true.

4. A Realist Stance is Rationally Obligatory

The challenge for the realist should now be clear. Based on the argument of section 2, the realist should admit that their claims to know depend on their realist stance. Based on the argument of section 3, the realist should reject voluntarism about stances, and so insist that their stance is tied to theoretical reasons. But realists have not yet consisted how best to articulate these theoretical reasons. It is clear that most realists would try to avoid pragmatic incoherence by rejecting voluntarism about stances, but it not clear how to do this. My suggestion is that the realist should maintain that their realist stance is rationally obligatory. It follows that a defense of realism involves a defense of the claim that one’s realist stance is rationally obligatory. How might this defense go? As we have seen, the distinctive element of the realist stance is endorsing a policy to use IBE. So the realist must say something to connect their initial stock of evidence, the purported explanation that is deemed the best and the appropriateness of claiming to know that this explanation is genuine, i.e. true.

Here is one proposal for how to do this: suppose that if E is evidence for P, then there is an explanatory connection between E and P. The explanatory connection can come in three forms. First, E may be part of an explanation for P. Second, P may be part of an explanation for E. Third, there may be a third element P’ that is part of an explanation of E, part of an explanation of P and also part of an explanation of why E is evidence for P. Here are some illustrations of these three forms of explanatory connection. First, an earthquake (E) may be evidence that a tidal wave will occur (P). In addition, the earthquake (E) is part of an explanation for why the tidal wave occurs (P). Second, the symptoms of some disease (E) are evidence for the presence of the disease (P). But of course the presence of the disease (P) is part of an explanation for the symptoms (E). Third, a falling barometer (E) is evidence that a rain storm will occur (P). Here we find that the change in pressure (P’) is part of an explanation for the falling barometer (E) and part of an explanation for the rain storm (P). In addition, the change in pressure (P’) is part of an explanation for why the falling barometer (E) is evidence for the rain storm (P). Through this explanation we come to appreciate why having E in one’s evidence gives one a reason to claim to know P.

It should thus be plausible that in many cases where E is evidence for P there is an explanatory connection between E and P. My suggestion is to strengthen this link in the following way: if one says that E is evidence for P, then one is rationally obligated to suppose that there is an explanatory connection between E and P. Here is an argument for this conclusion that clarifies the demanding notion of ‘evidence for’ that is involved:

1. An agent S who claims that E is evidence for P should believe that the truth of E is a sufficient basis to claim to know P.¹²

2. For the truth of E to be a sufficient basis to claim to know P, there must be an explanatory connection between E and P.

3. Therefore, an agent S who claims that E is evidence for P should believe that there is an explanatory connection between E and P.

There are thus rational obligations to believe in explanatory connections that arise from the nature of evidence.

If we accept principle (3) about explanatory connections, then we can clarify the basis for the realist’s use of IBE to make claims to know about unobservable entities. There are cases where E is evidence for P, and yet there is no plausible explanatory connection that links E to P that avoids unobservable entities. Consider, again, Franklin’s kite experiment. Based on significant experience with artificially created electrical discharges, Franklin designed his kite experiment in the hopes that it would reveal a number of salient similarities between naturally occurring lightning and these artificially created electrical discharges. During the experiment, conducted during a thunderstorm, Franklin obtained evidence E that (P) his kite string had become electrically charged. For example, he noticed that (E) the key attached to the string emitted electrical sparks and could be used to charge a capacitor (Leyden jar).¹³ In this case, the only plausible explanatory connection between E and P involved (P’) the clouds being electrically charged. This in turn supported the conclusion that (L) lightning is an electrical discharge.¹⁴ So this instance of IBE involves uncovering an explanatory connection between evidence and what this is evidence for. The realist can thus defend their use of IBE as an appropriate means to uncover the explanatory connections that their evidential claims rationally oblige them to suppose.

A voluntarist about stances could question this argument for the realist stance by questioning the significance of premise (1). Suppose that S has adopted a realist stance. Then premise (1) applies to S, and it follows that (3) applies to S as well. But consider another agent S’ who has not adopted a realist stance. The voluntarist will maintain that premise (1) does not apply to S’ and so (3) does not apply to S’ either. This will suggest to the voluntarist that even if S acquires a kind of theoretical reason for their realist stance through this argument, this theoretical reason is not generally available. In particular, S’ has no theoretical reason to adopt the realist stance. So there is no way for the realist to provide any argument that their realist stance is obligatory.

My position is that the theoretical reason developed in this argument is generally available and does provide each of us with a reason to conclude that the realist stance is obligatory. Here I suppose that everyone claims to know something, and that sometimes these claims involve explanatory connections that are not observable. Everyone that is placed in this position is thereby committed to the existence of an unknown explanatory connection between their evidence and what they claim to know. This provides each of us with a theoretical reason to employ the only available form of inference that can reveal the character of this otherwise unknown explanatory connection, i.e. IBE. Of course, many anti-realists will say that they do not wish to know about these explanatory connections. This may be an accurate report on their values, but their values do not determine their theoretical reasons. My position is that the nature of evidence and knowledge provide the theoretical reasons for the realist stance. This is why adopting the realist stance is obligatory.

A voluntarist about stances says that a realist stance is rationally permissible. That is, there is no theoretical reason against adopting a realist stance. But their voluntarism means they do not think there is any theoretical reason for adopting a realist stance. The last four paragraphs have summarised a theoretical reason for adopting a realist stance: the realist stance, with its policy of endorsing IBE for unobservables, includes an appropriate means to identify the explanatory connections that are required for their evidential claims to be legitimate. For the realist stance to be rationally obligatory, it needs to offer the only means of identifying these explanatory connections. If IBE is the only way to discharge this rational obligation, and an agent is rationally obligated to employ the only means available to meet such obligations, then the realist stance will be mandatory.

On the first point, there is a long philosophical tradition stretching back at least to Feigl’s ‘Existential Hypotheses’ that argues that explanatory considerations are the only means of supporting claims to know about unobservables (Feigl 1950).¹⁵ One way to argue that IBE is needed to support claims to know about unobservables involves underdetermination worries that are often raised by antirealists. The antirealist grants only evidence concerning observables. If one’s evidence relates only to observables, then the evidence is consistent with adding either the claim that lightning is an electrical discharge or that lightning behaves as if it is an electrical discharge. Forms of ampliative inference that avoid explanatory considerations, like enumerative induction, do not provide adequate support for the realist claim. However, IBE is readily seen to favour the realist claim over its weaker ‘as if’ alternative. For lightning’s being an electrical discharge is part of an explanation of (part of) the evidence, while the weaker ‘as if’ alternative is not part of an explanation of that evidence. More generally, the realist needs IBE to claim to know of unobservables as this is the only way to overcome underdetermination worries. In certain special circumstances, IBE is also the only way for the realist to identify the explanatory connection between their evidence and their claims to know.

Let us suppose, then, that IBE is the only means of obtaining knowledge of the explanatory connection between the realist’s evidence and what the realist claims to know. I have already argued that when the realist puts forward a claim to know on the basis of some evidence, then they are rationally obliged to suppose that there is some explanatory connection between the evidence and what is known. It follows that the realist is rationally obliged to employ IBE. It is the only means of obtaining knowledge of something that the realist is rationally obliged to suppose exists.

5. Objections

One challenge that Chakravartty has raised when discussing his voluntarism about stances with scientific realists is for the realist to develop ‘a non-question-begging critique of a rival stance or stances held by antirealists’ (Chakravartty 2021, 58). One objection to the argument of section 4 is that it does in fact beg the question against the antirealist. We can see this because the argument advanced in section 4 will not persuade the antirealist to adopt the realist stance. My response to this sort of objection is to impose a lower standard on a defense of realism and a realist stance. The realist does not have to persuade the antirealist to become a realist in order for the defense of realism to be viable. I suppose that we are dealing with a realist who has adopted the realist stance. The main issue is what this realist can say to themselves (and other realists) when they are engaged with a debate with their antirealist interlocutor. This defense of the realist stance is thus a kind of rationalisation of one’s own choices, if one is a realist. This rationalisation can be genuine and successful even if it fails to persuade the antirealist. The antirealist is likely to object to at least one of the steps of this defense of the realist stance. Despite this impasse, the realist can provide adequate support for each step of the argument. In addition, this defense of a realist stance is both non-trivial and non-circular. It is non-trivial because it relies on substantial claims about evidence, knowledge and explanation. These claims can be assessed and evaluated in ordinary ways. In addition, this defense is non-circular: the conclusion is that one is rationally obligated to employ IBE. This conclusion does not appear among the premises, and the form of the argument is not an instance of IBE. The argument instead relies on features of rational obligation, and how those features are transferred from a rational obligation regarding some end to a rational obligation to adopt the only means for obtaining that end. For the realist, then, the antirealist’s refusal to employ IBE looks unmotivated, arbitrary or irrational. The antirealist will not agree, of course, but this by itself is not evidence that the realist should use to question their own rationalizations.

This way of characterising what an adequate defense of realism requires turns out to concede a number of substantial points that voluntarists like Chakravartty often emphasise. One consideration in favour of voluntarism about epistemic stances that I have not yet discussed is the intractability of the realism debate in the philosophy of science. It might seem that the proliferation of positions in the realism debate over the last fifty years is itself conclusive evidence for voluntarism about epistemic stances. For if there actually were theoretical reasons that identified a realist stance as mandatory, then one would hope that philosophers who worked on these questions would be in a position to clearly identify that reason. And once the reason was found, surely the realism debate would end and everyone would adopt the realist stance. So, the argument concludes, there are no theoretical reasons to adopt a realist stance. Something like voluntarism must be correct.

When formulated in this way, this argument is readily seen to be an instance of IBE: the best explanation of the ongoing realism debate is voluntarism about stances. I do not think this is the best explanation. My preferred explanation is that theoretical reasons that should govern our commitments are often very difficult to identify. These reasons can be there independently of anyone’s thoughts actually being guided in the appropriate way. This is why many philosophical debates that actually should be resolved in one way actually persist indefinitely. In this respect, the correct resolution to a philosophical debate is similar to the correct resolution of moral debate. For many years people debated moral principles such as the abolition of slavery or the equality of men and women. Eventually adequate theoretical reasons for these moral principles were formulated, and now it is easy for nearly everyone to grasp their force. Theoretical reasons in philosophy could be like this: very difficult to find, and yet ultimately persuasive when suitably developed and articulated for audiences with the right cognitive preparation.

This analogy with moral debates could also clarify why, even though the arguments developed in this paper are of primary significance to realists, the arguments are philosophically significant and not just the trivial point that a realist must be a realist in order to be a realist. In moral debates, the ultimately correct arguments for a moral principle may initially function to persuade and rationalise the views of people who already accept that moral principle. This is just the first step in convincing neutral parties or those who reject that moral principle. Similarly, in the philosophy of science, the first step is for the realist to get their own house in order by clarifying the basis for their realism. I have argued that this realism is threatened with incoherence once the essential role for the realist stance is admitted. So it is highly non-trivial to see what theoretical reasons there are to adopt the realist stance. With these theoretical reasons identified, the realist will be in a better position to articulate these reasons and ultimately persuade non-realists.

Another argument that Chakravartty has offered maintains that there are no theoretical reasons to adopt an epistemic stance because of their ties to a restricted class of epistemic values. Suppose that the only theoretical reasons there could be would involve the weight that one assigns to these epistemic values, such as believing the truth or not believing what is false. However, there is no theoretical reason to set these weights at any particular value, and so no theoretical reason to adopt one stance rather than another: ‘there is no universal, stance-transcendent conception of how to chart a course between the poverty of excessive caution and the pitfalls of excessive zeal’ (Chakravartty 2017b, 234).¹⁶ Notice that this argument assumes that the only theoretical reason in favour of one’s stance would involve settling how one should weigh one’s epistemic values against one another. My position is that there are other sorts of theoretical reasons that can be offered to support the claim that one’s stance is rationally obligatory. These reasons arise from reflection on the nature of evidence, and how evidence requires the existence of an explanatory connection. This permits the realist to rationalise their adoption of a realist stance without addressing the relative weights of their epistemic values. These relative weights do not enter into their reasoning in support of the use of IBE for unobservables.

From this vantage point we can see that the assumption that the only theoretical reason to adopt a stance would turn on the relative weights of one’s epistemic values is already framing the question in voluntarist terms. This framework essentially takes for granted that we are dealing with a practical decision where means-end calculations are the only relevant considerations. The realist should not accept this framing of the question, and should insist instead that different kinds of theoretical considerations are relevant to evaluating a given stance.

In a recent article ‘Scientific Realism and Blocking Strategies’ Pils has deployed Chakravartty’s voluntarist framework to further examine the implications of epistemic stances (Pils 2023). Pils agrees with Chakravartty that an epistemic stance is adopted with reference to the truth goal that involves a ‘fundamental epistemic trade-off’ between ‘believing truths and avoiding error’ (Pils 2023, 3). This framework is a kind of ‘meta-stance’ that informs an individual’s choice of epistemic stance based on their individual weights for these two values: ‘One picks out a specific stance … if one (i) commits to a specific balancing of the truth goal, and (ii) policies of how to achieve it’ (Pils 2023, 4). For our purposes the most interesting part of Pils’ discussion concerns his criticisms of what he calls ‘steadfast’ realism. A steadfast realist adopts a realist stance that counsels the use of IBE in appropriate circumstances. In addition, the steadfast realist makes their realist stance invulnerable to overturning by the accumulation of historical or ‘diachronic’ evidence concerning how truth-conducive their modes of inference turn out to be. Pils asks us to consider the following possible future for science:

Now suppose that our best scientific theories switch ontologies 1000 times in the next 1000 years. Suppose further that in the year 3020 the data of our best scientific theories is best explained by the existence of xyz. Given that the theory is mature and predictively successful, we would have belief obligations or at least permissions in xyz. Such a version of classical realism, one that solely relies on inference to the best explanation (IBE), blocks itself from diachronic evidence as well (Pils 2023, 7–8).

Pils’ complaint is that the ‘blocking strategy’ of making one’s epistemic stance invulnerable to undermining through the historical track-record of one’s inferences renders that stance unavailable. The upshot is that the only sort of realist stance that is viable is a ‘Cautious Realism’ (Pils 2023, 9) that qualifies its inferential policies as the historical evidence dictates.¹⁷

Pils’ argument against steadfast realism is a potential challenge to my position that a realist should take their realist stance to be mandatory. For if my position is correct, then it looks like a realist has to be a steadfast realist, committed in advance to a dogmatic loyalty to their policy to employ IBE no matter what historical track-record unfolds. I agree with Pils that a realist should not endorse this sort of dogmatic loyalty to IBE, and that the only viable form of scientific realism is a cautious realism that responds appropriately to the historical record of success and failure for various modes of inference. The key point to appreciate is that even though the realist takes their realist stance, with its inclusion of IBE, as mandatory, the realist is not bound to deploy IBE independently of the historical evidence that they have at their disposal. For along with their policy to employ IBE the realist will have a number of other inferential policies that should be employed to restrict their use of IBE when the evidence warrants it. For example, every stance should include a policy to not employ a mode of inference when the available evidence indicates that that mode of inference is unreliable. As long as a realist stance includes this policy, that policy can lead the realist to suspend or qualify their use of IBE if the right historical evidence arises.

This reaction to Pils’ criticism of steadfast realism suggests another important concession to Chakravartty’s voluntarism. One aspect of the ongoing realism debate in the philosophy of science could be that different philosophers have different takes on the evidential significance of the track-record of IBE. It may even be the case that all parties to this debate are best interpreted as adopting a realist stance that permits the use of IBE in some circumstances. If this was the case, there could still be disagreements about whether or not we have reached the point where the track-record of scientific failures undermines our ongoing use of IBE. I take this to be one motivation for Chakravartty’s voluntarism. Nothing I have said here is meant to suggest that we can now employ IBE without addressing these concerns. I take these concerns to be the primary reason to worry about realism, and until they are satisfactorily addressed, we can expect realism to remain a minority position in the philosophy of science.

6. Conclusion

An epistemic stance is something that an agent adopts, and that settles their policies for moving from their evidence to their claims to know. In this paper I have argued that the scientific realist should admit that an epistemic stance is required as part of their defense of realism. This epistemic stance includes the permissibility of the use of IBE for the purposes of claiming to know about unobservable entities. A second point I have argued for is that the realist should insist that their realist stance is mandatory. That is, the realist should be an anti-voluntarist about stances. This puts the burden on the realist to articulate some non-trivial and non-circular basis for a realist stance. I have sketched such a basis in section 4 by appeal to principles that connect evidence, knowledge and explanation.¹⁸

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 See also (Chakravartty 2017b, 46–47).

2 See, in particular, the three different accounts of evidence developed by (Bird 2022, chapter 5) (all evidence is known), (Comesaña 2020) (all evidence is a special sort of belief) and (Williamson 2015) (all evidence is what one rationally grants).

3 One complication here is that some evidence may be available to the scientist even if they are not aware of it. For example, Lackey argues for ‘normative defeaters’ that are sufficient to deprive a scientist of knowledge because the scientist should have been aware of this evidence (Lackey 2014, 292). I would argue that one’s total evidence should account for such defeaters, but I have to set aside this complication here due to reasons of space.

4 Some anti-realist stances could also try to validate this reasoning using a more restricted version of IBE or some other epistemic policy. In section 4 I argue that an unrestricted form of IBE is needed to claim to know about unobservables.

5 If externalism about justification is right, there might still be a good argument for epistemic stances. I have to reserve a consideration of these arguments for future work.

6 I am grateful to an anonymous referee for suggesting this argument, although they may not endorse it in the form presented here.

7 This claim is noted in Chakravartty’s response to Psillos (Chakravartty 2021, 53).

8 Chakravartty also cites this paper, as an instance of someone who denies that ‘human beings have control over at least some of their beliefs’ (Chakravartty 2017b, 216, fn. 4).

9 While I take Psillos’ appeal to Williams on deciding to believe to entail that a realist should insist that their stance is rationally obligatory, this does not appear to be Psillos’ own position. See esp. Psillos’ discussion of what he calls ‘the realist framework’ whose adoption is just as dependent on the will as Chakravartty’s realist stance: ‘the realist framework … is not an object of belief or doubt. It can either be shown to be dispensable or less efficient than others when it comes to achieving a certain aim’ (Psillos 2011, 312).

10 For the debate between the evidentialist and the non-evidentialist on reasons to believe see esp. (Kelly 2002; 2003; McCormick 2019; Rinard 2015; 2019a; 2019b).

11 I am grateful to an anonymous referee for raising this worry.

12 Many discussions of evidence do not require that if E is evidence for P, then E is a sufficient basis for a claim to know P. For example, some suppose only that if E is evidence for P, then E makes P more likely to be true. As I am interested in a realist who claims to know various things about unobservables, I am working with a more demanding account of when someone’s evidence E is evidence for P. A more detailed investigation of when evidence E is evidence for P is needed, but is not feasible here. See (Bird 2022, chapter 5) for an important, recent discussion of these issues.

13 See esp. (Franklin 1941, 265–266). See also (Cohen 1990).

14 As Franklin concludes his letter, ‘and thereby [is] the sameness of the electric matter with that of lightning completely demonstrated’ (Franklin 1941, 266).

15 See esp. (Lange 2023). Lange draws on Sellars to argue that ‘It is not open to an observer to do without any expressly explanatory background opinions’ that inform their use of inference to the best explanation (Lange 2023, 13).

16 See also (Chakravartty 2017b, 216–217).

17 Pils develops the same sort of criticism for anti-realists who are similarly steadfast in their anti-realism.

18 A version of this paper was presented at the 2022 Biennial Meeting of the Philosophy of Science Association. I am grateful to the audience for helpful suggestions and to Anjan Chakravartty for many insightful comments about how to improve this paper.