1 Introduction

There is a widespread view that the Scientific Realism Debate (SRD) is circular in the sense that there is no science-like progress. Rather SRD is characterized by the eternal recurrence of the same seemingly unsolvable problems, which is typical of philosophy and is echoed by the enduring lack of consensus among philosophers. This view is expressed once more in a major book of the current SRD-literature, when Peter Vickers (2023: ix) states that

the methods of ‘pure’ philosophy can be frustrating: they never seem to establish anything definitively. Debates seem destined to go around in circles, or else evolve somehow, without ever reaching a firm conclusion that might be held up to outsiders as a noteworthy achievement.

Due to such concerns, several scholars have advocated the dissolution of SRD, as is traditionally conducted, in favor of reformed, (allegedly) more promising projects, which are broadly eliminativist towards the distinctively philosophical aims and methods. Vickers’ own proposal, which (lacking a better term) we dub sociological externalism, is that the identification of scientific claims worthy of realist commitment should be done by investigating the social features of the scientific communities making those claims, rather than by looking for types of truth-conducive evidence, epistemic virtues and the usual stuff chronically haunting SRD. Another (not so) recent and ever-growing dissolution-trend is summarized in the motto ‘SRD should go local’. Roughly, according to localism, SRD-scholars should abandon abstract philosophical reflection and be content with particular case-studies. A goal of this paper is to point out the limitations of the above approaches. We suggest that there are no tricks to break the circle. If SRD is circular, we should learn to live with it. Put in context, our endeavor can be seen as a resistance to a form of scientism that tends to prevail in contemporary (meta)philosophy of science.

Our plan is as follows. In §2, we outline the project that SRD-participants have been involved in for some time now, which largely consists in the articulation of epistemic principles. We call it the Received View, and we distinguish two versions of it, diverging on their conception of principles: an absolutist and a pro tanto one. Both are underpinned by a two-level meta-epistemological framework, according to which the epistemology of science is a second-order enterprise aimed at stating principles that indispensably refer to aspects internal to scientific methodology.

In §3, we discuss and criticize sociological externalism and localism as attempts to abolish the said two-level framework. We argue that the projects they lead to are incomplete—at best complementary to SRD as we know it—and that, even when judged in themselves, they cannot flourish in the absence of the ‘traditional’ philosophical reflection they wish to get rid of. However, although not substitutes for the Received View, sociological externalism and localism have some insightful features. These are assigned their proper place in the dialectics of SRD, which, in §2, is shown to be an instance of the method of reflective equilibrium (MRE).

A worry about MRE is that it is epistemically circular. In §4, we provide a response to this. Following Michael DePaul (1993, 1998), we claim that MRE is the only rational method of philosophical inquiry. Hence, since all alternatives are irrational, the remaining one—sceptic epoché excluded—is imposed, even if circular. Obviously, this is ipso facto an argument for why SRD cannot but proceed in accordance with MRE and why SRD is epistemically circular. But we also take it as an argument for why this should not bother us! This can be explained by saying that, overall, our goal is therapeutic in the following stoic sense: to mitigate the anxiety caused by simultaneously (i) believing that SRD is circular and (ii) suspecting that there is a good way out of the circle that we just haven’t found yet. For, having dismissed this suspicion—i.e. having renounced (ii)—one may learn to live with (i) and be more calm in carrying on with what one has.

Finally, in §5, we draw some overall conclusions.

2 The Received View of the Scientific Realism Debate

In order to explain the relationship between science and epistemology, it is customary to use a two-level image, where the first level is occupied by science and the second by epistemology, while a typical way to conceptualize this layering is by recourse to the following questions:

1st Order Question: Which particular beliefs are justified?

2nd Order Question: What are the criteria (principles/rules) for deciding which beliefs are justified?

These questions suggest an obvious distinction between science and epistemology. For, clearly, scientists seek an answer to the first-order question regarding particular theories falling into their research fields. The epistemological inquiry, on the other hand, is conceived as a “second-order criteriology” (Losee 1972/2001: 2), aiming to respond to the second-order question: what criteria do, or should, scientists apply in the epistemic appraisal of particular theories? Of course, one might claim—and we’ll do—that this way of delineating the territories of science and philosophy is to a large extent arbitrary, since, to say the least, scientists cannot but be puzzled about the rules they (are supposed to) employ in doing their job. Although the distinction between the above two questions is clear-cut, the prospects of being able to handle one of them without dealing with the other are dubious.Footnote 1 Be that as it may, let’s stay within the boundaries of this two-level meta-epistemological image and focus on the role of epistemologists.

Depending on how the interplay between the two questions is viewed, Roderick Chisholm (1982: 66) has distinguished two opposing meta-epistemological approaches: particularism and methodism. The former suggests that epistemological inquiry should begin with an answer to the first-order question and then proceed to the second-order. That is, particularists privilege our pre-philosophical intuitions about the epistemic status of (some) particular beliefs and state that criteria should be evaluated according to how well they accommodate those intuitions. On the contrary, methodists give methodological and normative priority to criteria. It is claimed that the epistemic appraisal of particular beliefs can be made only after having a response to the second-order question at hand, while the latter is usually supposed to be answered via a priori reflection.

2.1 The Method of Reflective Equilibrium

This meta-epistemological division is often supplemented with a third, ‘mixed’ approach known as the method of reflective equilibrium (MRE). Instead of granting priority either to rules or to judgements about particular cases, MRE promotes a bidirectional justification process. As Nelson Goodman put it in his seminal paper (1955: 67):

[Particular] inferences are justified by their conformity to valid general rules, and… general rules are justified by their conformity to valid inferences. … A rule is amended if it yields an inference we are unwilling to accept; an inference is rejected if it violates a rule we are unwilling to amend. The process of justification is the delicate one of making mutual adjustments between rules and accepted inferences...

Let’s take a closer look at how MRE proceeds. The inquirer begins by collecting a variety of particular epistemic judgements. A filtering process follows: the judgements are classified according to a scale of certainty. Those that inquirers are very confident about, based on widely shared pre-philosophical intuitions regarding the soundness of the relevant inferences, comprise what MRE-theorists call considered judgements. (Call the set of these judgements J.) These act as “provisional fixed [starting] points” (Daniels 1979: 267) for the epistemological theory construction, to the effect that the wanted epistemic principles (which make up the epistemological theory) must largely be articulated to account for them. That is, the epistemologist tentatively proposes a set of principles P by answering the following question about each member x of J: if judgement x is justified in virtue of conforming to an epistemic principle, then what is this principle?

Suppose that, after following this procedure for a proper subset of J, one reached a certain set of principles. It may, however, happen that one will then realize that a certain principle implies a judgement in conflict with a particular judgement z within J. In this case, a choice must be made between adjusting the principle or revising z. Presumably, if z occupies the highest levels of our scale of certainty, or, in Jared Bates’ apt words, z is a strongly nonnegotiable judgement (2004: 48), then in all likelihood the principle will have to be dropped or modified.

But, this need not be so. If the principle fits many of the most important members of J in a satisfactory way, or z is a marginal judgement, then perhaps it is z that will have to be revised. Ultimately, there is no considered judgement that is immune from revision. Considered judgements, however, cannot be discarded unrestrictedly because they are our initial “best guesses about the topic”, hence the account we arrive at “should be recognizable as an improvement upon them” (Elgin 2017: 66). And, although we cannot impose a principled limit to the number of revisions, we can require that extensive revisions of considered judgements are permitted only when accompanied by compelling reasons for doing so and by an explanation of “why [the revised judgements] seemed as reasonable as they did when they did” (ibid. 67).

This moving back and forth between principles and particular judgements is a key part of MRE, so that a main aim is to arrive at the best fit between sets P and J. But, according to MRE-theorists, this fitting-state—defined as narrow equilibrium point—doesn’t suffice. Instead, we should seek after a wide equilibrium state (Daniels 1979; DePaul 1993), taking into account all of our relevant background beliefs. These beliefs—call their set B—provide the basis for the advancement of arguments aimed at revealing the strengths and weaknesses of the proposed principles. If the latter were not tested this way, the reason for accepting them would largely be that they explicate the particular judgements; and this would be a form of particularism. Wide equilibrium requires extending the range of considerations beyond the initial judgements used to articulate our principles. (In fact, it is the use of these independent, wider reflections that makes substantive revision of particular judgements possible.) There is no point in trying to fix what these background beliefs or arguments are. With SRD in mind, they can be thought of as all the usual suspects: historical reflections, the no-miracle intuition, psychological and sociological considerations, metaphysical presuppositions, and so on. Needless to say that, once again, no belief is granted a privileged epistemic status. When an argument brings about a conflict between background beliefs and principles (or particular judgements), it’s an open question which will be revised.

The ultimate goal of MRE is to obtain an ordered triple <J, P, B> constituting the optimal wide equilibrium point. What makes a point optimal? It’s clear that coherence within <J, P, B> is basic. That is, <J, P, B> must be consistent, and the inquiry must establish a dense network of relations of mutual support between its members, i.e., a network of (omnidirectional) logical, probabilistic or explanatory relations. There are also desiderata imposed by general requirements regarding theory building (conceptual clarity, simplicity etc.). But, it’s crucial to emphasize that these desiderata are adjusted during equilibrating; their configuration “is a product of, not a precondition for” the process of equilibrating (Baumberger & Brun 2021). This reaffirms the holistic, omnidirectional character of MRE, or as we say, the circle we are trapped in.

2.2 Pro Tanto Principles

Having set this meta-epistemological landscape, let’s turn to SRD. For a long time, SRD has been focused on three tasks that together make up what might be called the Received View. These three tasks involve the articulation, evaluation, and refinement of epistemic principles (or criteria for realist commitment). In the words of Kyle Stanford (2021), SRD.

has been most fundamentally concerned with whether there is any categorical or general variety of empirical success or evidential support that serves as a reliable indicator that a theory … will be retained and ratified throughout the course of further inquiry [227]. … For… realists and instrumentalists,… more important than any remaining point of… disagreement… is the shared epistemic project in which they are jointly engaged: … [they are]… seeking to identify, evaluate, and refine candidate indicators of epistemic security for our scientific beliefs, and both see the historical record of scientific inquiry itself as the most important source of evidence… for pursuing that joint project together [236].

When it comes to the first task, here is the state of play. SRD-participants specify types of theoretical success, which consist in some privileged ways in which theoretical hypotheses might relate to observations (e.g., being essentially deployed in the derivation of novel predictions) or in the instantiation of some theoretical virtues (e.g., simplicity, unificatory power and the like). Call Criterion CR(T) an arguably complex and disjunctive open formula of the following form: ‘theory T enjoys such-and-such type of success or such-and-such type of success…’. These types of success are then assumed to be truth-conducive, i.e., the satisfaction of CR(T) by a theory is (very probably) an indicator of its (approximate) truth. On the basis of this “high-level” realist hypothesis (Henderson 2018: 152), we can articulate epistemic principles of this form:

For every epistemic agent A and every theory T, if CR(T) holds and CR(T)’s holding is within A’s epistemic ken (at time t), then A is requiredFootnote 2 (at t) to believe that T is approximately true.

In its more ambitious version, which might be called the Absolutist, or Monist Received View, the latter’s ultimate goal is the articulation of such a principle. Still, more modest versions are possible. These are based on a different conception of principles. Let us explain.

First, note that the aforementioned principle identifies a set of non-normative properties (types of success), whose (certifiable) instantiation is said to be a sufficient condition for a belief to have a determinate epistemic status, i.e., obligatory, permissible, or impermissible. Principles of this sort are called absolute principles (Dancy 2017). Unlike them, the so-called pro tanto, or contributory principles (Dancy 2017; Ridge & McKeever 2020), do not attempt to directly determine the epistemic status of beliefs. Rather, they specify some non-normative properties, whose instantiation is claimed to offer reasons in favor of, or against, doxastic attitudes. Pro tanto principles are like this:

If a doxastic attitude D—which is held by, or ‘is available’ to, an epistemic agent A at time t—instantiates property F (e.g., being a belief whose propositional content entails novel predictions) and F’s instantiation is within A’s epistemic ken at t, then this gives rise to a reason with this polarity (i.e. a positive or negative reason, as the case may be, for A to adopt D).

The overall epistemic status of D is then said to be determined by ‘weighing’ all the reasons present in a given situation to see where the balance lies (see Lord 2018).

A key tenet of the pro tanto approach, distinguishing it from the absolutist one, is that the weighing process cannot be algorithmic. An indispensable role for context-dependent judgements is acknowledged; a kind of Duhemian bon sens is inevitably employed in assigning weights to reasons (cf. Berker 2007: 110). So, based on this approach of principles, we can admit a more ‘relaxed’ version of the Received View that might be called Pro Tanto, or Pluralist Received View. This is probably what Stanford’s quote corresponds to. Given that, the articulation of principles can be thought of as including inter alia (i) the identification of features that provide reasons with certain context-independent polarities, (ii) the codification of the (supposedly systematic) behavior of some contextual factors capable of enabling/disabling or intensifying/attenuating reasons (see Bader 2016), (iii) the discovery of some possible ‘ranking principles’ dictating invariable orderings of some virtues (e.g., that logical consistency always outweighs simplicity; cf. Douglas 2013), and (iv) the determination of the “combinatorial function” (Berker 2007: 120) leading from contributory reasons to all-things-considered epistemic properties.

Now, the second task within the Received View is the evaluation of principles, and the history of science is a major battleground. As is known, anti-realists point to historical counterexamples to proposed criteria, i.e., cases where theories that satisfied these criteria were eventually rejected. These act as defeaters for the corresponding principles. Ultimately, anti-realists seek to motivate their epistemic parsimony by showing that no epistemic principle dictating inferences to unobservables is counterexamples-free.

In the face of counterexamples, realists proceed to the third task, viz., the refinement of their criteria by imposing further constraints that must be met for a theory to be regarded as successful or for a posit to be regarded as essential for the generation of a success (Psillos 1999: ch.5). But, apart from the exclusion of counterexamples, there are further desiderata for the refinement of principles. Τhe adjustments must be non-ad hoc, independently justified in part. Moreover, they should not be overshooting, i.e., the resulting criteria shouldn’t be so strict as to render untrustworthy (lots of) widely accepted scientific assertions (cf. Stanford’s “trust argument” (2003: 556)). It is almost unanimously admitted that the refined principles must still be ‘realist enough’. Note that this burden lies on the shoulders of anti-realists as well. That is, if their own strict epistemic principles render unjustified belief in scientific claims so deeply entrenched as deserving to be dubbed ‘common knowledge’, so much the worse for them. Even anti-realism must be realist enough!

In fact, this is the key idea behind a standard realist dialectical move in SRD. Realists argue that the inferences precluded by the strict anti-realist principles (i.e., inferences about unobservables) are relevantly similar to some inferences that all sides are willing to permit (e.g., some inferences about unobserved observables). It is thus claimed that either anti-realists end up with an incoherent epistemic stance—viz. refusing the former kind of inference, whereas arbitrarily allowing the latter—or their view is reduced to a sort of far-fetched scepticism, which they explicitly want to avoid (Stanford 2006: sec. 1.2). And the reason why (even) anti-realists want to avoid radical scepticism is because it amounts to the extensive rejection of strongly nonnegotiable judgements. Therefore, the only viable option for anti-realists tangled up in this situation is to liberalize their epistemic principles (e.g., Stanford’s inclusion of “projective evidence” (2011)). So, a chief aim of advocates of epistemic principles is to maintain a good track-record—i.e., to avoid the celebrated Pessimistic Meta-Induction (PMI)—while at the same time accommodating the (majority of) nonnegotiable instances of actual scientific inference.

The above characteristic snapshot of the dialectics of SRD demonstrates that the Received View is conducted in accordance with MRE. It reveals the justificatory omnidirectionality of the process of evaluation and refinement of epistemic principles. On the one hand, SRD-participants recognise that second-order arguments against the reliability of scientific methods (such as PMI) have the capacity to defeat scientific claims. Schematically put, acknowledging the role of second-order considerations in the epistemic appraisal of scientific theories reflects the methodological and normative arrow heading from background beliefs (B) and principles (P) towards particular judgements (J). On the other hand, the opposite direction—from J to P and B—is echoed in the recognition that principles are (at least partly) accepted in virtue of being able to accommodate our (pre-philosophically) more trusted particular scientific assertions. That is to say, extensive revision of considered judgements, albeit not forbidden, can hardly be accepted. Everyone must be realist enough!

Taking stock, the Received View exemplifies the MRE-scheme and a fortiori is committed to the two-level meta-epistemological image.

3 Dissolving the Received View

The Received View has been accused of reaching a stalemate, so there are growing voices for its dissolution. This section examines two kindred dissolution-proposals—sociological externalism and localism—whose revisionist visions are based on their dissent from the two-level meta-epistemological image. Sociological externalism redefines second-order discourse by proposing that it shouldn’t be concerned with ‘things’ internal to scientific practice. Philosophers wishing to identify trustworthy scientific assertions should focus on external aspects, and specifically, on the social features of scientific communities, which act as indirect means for the epistemic assessment of first-order evidence. Talk about types of theoretical success is replaced by talk about types of communities. Localism, on the other hand, fosters the elimination of second-order reflection. Rather than seeking criteria, philosophers are advised to be content with case-studies taking into account nothing more than the first-order evidence cited by scientists. In a (literal) sense, localists claim to be simply “doing science” (Magnus 2013: 50). For them, talk about types of theoretical success is off the mark; what matters in epistemic appraisal are particular instances of success.

In what follows we provide some reasons for holding that the dissolution of the Received View in favor of an externalist or localist project would impoverish SRD or, all the more, that these projects are untenable without the ‘traditional’ epistemological reflection they wish to abolish.

3.1 Against Sociological Externalism

Vickers’ recent book expresses the worry that SRD has become a battlefield of endless controversies, and suggests “another promising methodology”, which employs “the methods of sociology” and “move[s] us… further away from ‘pure’ philosophy” (2023: ix). In this way, Vickers claims “one might still dream of saying something definitive about science, something that could draw a consensus of opinion in a way that is vanishingly rare in philosophy” (ibid.).

His proposal is roughly as follows. Answering the second-order question doesn’t involve a criterion specifying types of (supposedly) truth-conducive success. Instead, philosophers looking for “future-proof” scientific assertions should focus on the features of the communities making these assertions. In this spirit, Vickers puts forward the following externalist criterion (ibid. 217): A scientific claim is very likely true if.

(1) 95 per cent of the relevant scientific community [is] willing to describe the claim in question as an ‘established scientific fact’ [and]

(2) The relevant scientific community [is] large and incorporate[s] a substantial diversity of perspectives.

Unfolding this criterion reveals several points that merit more careful examination than possible here. Having said that, we’ll make two critical remarks on the viability of the proposed research program. First, the sociological enterprise envisioned by Vickers cannot be substituted for the Received Project, but is merely supplementary. Vickers is, of course, correct in saying that, when laymen are to fix their belief regarding scientific matters, all they can do is “canvass the opinions of a large number of scientists… and see what kind of consensus… [they] can identify” (ibid. 44). However, this admission does not deprive epistemology of the responsibility or ambition to answer its central question, which is not “under what circumstances can I trust scientists’ testimony while being ignorant about their subject?”, but rather “what are the characteristics of truth-conducive (first-order) evidence?” In other words, a response to the second-order question that is relevant to SRD must be one that could guide scientists themselves, or, less ambitiously, it must amount to the explication of the most reliable scientific methods.

This can also be expressed as follows. Even if Vickers’ project succeeded in identifying trust-worthy science, it would leave us without the kind of understanding that is the appropriate end for a properly epistemological inquiry. Epistemology’s answer to ‘why should we believe p?’ cannot just be ‘because p is asserted by experts’. Instead, as DePaul (1993: 62–3) would have it, “we want to know in something like the way the expert knows. We do not merely want to be able to sort true from false… propositions; we want to know what distinguishes them [epistemically]… [O]ur goal is fundamental, primary… knowledge”, not testimonial. Vickers’ project forsakes the most interesting questions by converting SRD to a branch of epistemology of testimony—which is a kind of secondary source of knowledge, regarding the transmission of knowledge—when SRD should be part of a fundamental epistemology about basic sources of knowledge. Sociological externalism, even if profitable, is incomplete and unsatisfying, unless one succeeds in delimiting one’s curiosity.

Vickers probably wouldn’t refuse that. Given his success in identifying future-proof science, he might concede that remaining jobs do exist for philosophers, but they are not among his aims. Nevertheless, our second remark is that, even when judged on its own terms, as an enterprise about the transmission of knowledge, sociological externalism cannot accomplish its goal without resources from internalist epistemology. Putting aside several subtle issues (see Vickers 2023: ch.9), our main objection is unsurprisingly this: testimony alone, i.e., testimony accompanied by total obliviousness to considerations internal to scientific practice, fails to secure epistemic guarantee for the testimony-receiver.Footnote 3 In her effort to identify future-proof science, Vickers’ philosopher-demographer cannot avoid facing issues of internalist epistemology.

Seen from an externalist point of view, it would make it easy to produce counterexamples to Vickers’ criterion. Isn’t geocentrism, after all, a theory that the 95% of the fifteenth century academics were willing to accept as true? Vickers anticipates this and responds that he admits science only in “the modern sense of that term” (ibid. 224). However, Vickers’ response presupposes a criterion distinguishing modern from pre-modern science. And this cannot be purely sociological—say, that in modern times the community is larger and more diverse—for it would then be subject to the objection that an even larger and more diverse future community will probably see our current community as just as liable to error as we see pre-modern communities (cf. Wray 2013). The criterion must refer to a qualitative jump in features internal to scientific practice correlated with the advancement of reliability. The internalist criterion need not be an absolute one. But, non-absolute internalist criteria are still internalist criteria! In short, Vickers’ project cannot but shift talk to features internal to science. Sociology won’t work.

This point has also been made by Philip Kitcher (2019) while criticizing Naomi Oreskes’ account (2019), on which Vickers heavily relies. Kitcher argues that consensus within a diverse community is not enough to establish trust. We also “need to know why to believe the scientific consensus”, for “[b]ringing together a diverse group of people is not likely to achieve very much… [unless] collective investigation uses reliable methods” (2019). Besides, as Kitcher suggests, mere invocation of the consensus of a diverse scientific community is not.

likely to provide an adequate answer to the skeptic who takes … appeal to diversity at face value and wonders why an even more diverse community—say, one containing [skeptics]—wouldn’t be still more objective, and why it shouldn’t operate according to the rules of intellectual exchange that those who challenge [science] prefer. [ibid.]

Bottom line: The social features of the community do matter in trusting (or not) scientific consensus, but that’s only part of the story. Sociological considerations must be supplemented with reasons for believing that the community’s methods are reliable. The reason laymen think science is trust-worthy is “because it works” (ibid.), and this is a simplified version of the (hopefully more sophisticated) answer to the (quasi-)demarcation-problem that the Received View seeks; an answer that inevitably involves reflection upon “the distinctive ways that scientists engage with the world, the rules of evidence they deploy… and the ways that evidential standards… are grounded in a history of successful practice” (ibid.).

3.2 Against Localism

Let’s now turn to localism. Admittedly, it would take much more than is possible here to deal with it adequately, given the multitude of localist proposalsFootnote 4 and the fact that discussion fetches up on intricate issues about the nature of normativity. Our main goal is to neutralise some considerations that might be seen as bolstering localism, and to undermine it by demonstrating how heavy the burden is for localists to debunk the generalist meta-epistemological framework underpinning the Received View, which, crudely put, consists in the thesis ‘no rules, no normativity.’

Recall that we’re concerned with localism as a meta-philosophical view that abolishes the two-level image by rejecting the quest for principles as superfluous. That’s why we’ll focus on Penelope Maddy’s influential approach (2001, 2007, 2022)—the so-called “Second Philosophy”—since it is explicitly described as a “reaction to… two-level positions” (2007: 5), being eliminativist towards second-order reflection and refusing to depart from the “ground-level” (2022: 88) of science itself.

Maddy’s concerns against two-level positions boil down to an argument to the conclusion that epistemic rules lack normative force (2007: 402; 2022: 86). And if this is the case, then the quest for rules is superfluous, and consequently the Received View should be dissolved.Footnote 5

To see Maddy’s argument, consider a criterion referring to a type of success; e.g., ‘we must be committed to the existence of a particle if and only if we’re able to detect it’. Maddy casts doubts on this by saying that the definition of ‘detection’ will inevitably be “vague and/or easily disposable”, i.e., “given the poor prospects for a strict definition of ‘detection’, the proposed criterion must either decline to specify precisely what counts as detected or leave open the possibility of shifting its current specification in light of scientific progress” (2007: 402). But, she says, this means that the criterion lacks normative force. Why? Because it cannot provide epistemic guidance. If scientists claim to have detected a so-and-so by some new method, the criterion won’t have “any independent judgment to pass on this matter” (ibid.). For, according to Maddy (ibid. emphasis added).

at issue [will be] whether the new method falls under the old, vague notion of ‘detection’… or whether the old notion… should be replaced with one that admits the new method…, and the only grounds available on which to base either decision are the ordinary scientific grounds we gave for so-and-sos… In other words, the ordinary science, not the criterion, is what’s doing the work… This isn’t to say that the scientist’s judgment can’t be questioned… but the criterion alone isn’t providing any leverage.

In short, Maddy’s argument seems to be this: If a rule has normative force, then the non-normative categories it mentions (e.g., ‘detection’) must be non-vague and non-easily disposable. But, the non-normative categories mentioned in any rule cannot but be vague and/or easily disposable. Therefore, rules cannot have normative force.

Let us split our handling of this argument into two parts, one about the accusation of vagueness, the other about that of easy disposal. Consider vagueness first, and ask: Why can’t the categories mentioned in any rule but be vague? A response could be that this is a symptom of the inevitable vagueness of language in general, since no language is capable of achieving absolute precision. However, it is unclear why this ‘natural’ vagueness could deprive rules of normative force. One might happily admit that there are simply some borderline cases, where it is uncertain whether the inference under appraisal falls into the mentioned category, and where the epistemic status of the inference is thus underdetermined by the rule. But, this does not render the rule generally normatively idle, incapable of guiding us in most other cases. Thus, it seems wrong to say that, for a rule to have normative force, it must not be vague simpliciter, in the sense of sometimes not determining the case at hand. It would be more reasonable to merely demand that rules are not, so to speak, ‘ultra-vague’, in the sense of not completely determining many cases, even cases where they are supposed to do so. But, then, one needs to explain why rules are inevitably ultra-vague without recourse to a unanimously admitted feature of language, for it is not a characteristic of language in general that it is ultra-vague; witness the fact that it generally suffices for determinate communication and regulation.

On this point, of course, localists will emphasize the demanding nature of the relevant epistemic contexts. They will claim that epistemic rules are inevitably ultra-vague due to the high conceptual precision required for determinate judgements within the context of specialized scientific research; the language of epistemology is unavoidably too coarse-grained in regard to contexts where, by epistemologists’ own lights, it is expected to decide matters.

This is undoubtedly a reasonable worry. A way to address it is by challenging the assumption that a rule must ‘completely determine’ our judgement in order to have normative force in a given circumstance. Even in cases where a rule doesn’t suffice to determine our judgement (and our bon sens must be drafted in), it is unclear why there is not a sense of ‘partial guidance’. For there is an extent to which any category—albeit vague—narrows down the range of its permissible applications.

Moreover, and this is crucial, it seems that the rule’s guiding function cannot be preempted by particular intuitions. That is, it can be argued that epistemic agents wouldn’t be cognitively capable of producing output-judgements in particular cases if they were left only with bare intuitions, i.e., information not (at least partially) codifiable by any rule. Indeed, how are we to account for an agent’s ability to process and keep track of the uncodifiable, endless variability that a rules-free account imposes on the way that reasons depend on non-normative features? Likewise, without rules, the acquisition of the capacity for epistemic judgement appears to be impossible. Scientists learn how to judge, which means that what renders judgement possible must be acquired, transmitted, and able to be used in novel situations. There is a clear explanation of how this is done: scientists extract rules from particular cases, which can then communicate and apply to new, relevantly similar cases. However, this explanation is not available to someone who views rules as superficial.Footnote 6 In sum, as one of us has argued in a similar context, even though scientific practice cannot be equated to the “algorithmic” application of rules, it “cannot be reduced just to an ineffable ‘good sense’” (Psillos 2013: 112–3). Rules, although vague, are necessary.Footnote 7

Let’s now turn to the accusation that categories in rules are “easily disposable”, in the sense of being susceptible to “the possibility of shifting their current specification”. Apparently, the claim here is that easily revisable rules lack normative force, since it is the rules’ very authority that is questioned. That is, although a rule R might be sharp enough to deliver determinate judgements, believing that R is revisable in effect means accepting that there might be cases where following R is wrong. Hence, in each particular circumstance, we cannot help but wonder whether the case at hand is such (i.e., one calling for R’s revision), and our only pillar in deciding is our bon sens. Another way to state the issue is this. To be rationally forced by a rule R, one should believe that R is (probably) right, i.e., that belief-fixation in accordance with R promotes the ends of inquiry. But, the belief that R is easily revisable seems inconsistent with the belief that R is probably right.

In dealing with this issue, much depends on the adverb ‘easily’. Why is it the case that the categories in rules cannot but be easily revisable (as Maddy claims)? One can object to this claim by noting that it’s not a feature of language in general that all categories are easily revisable, nor is it necessary that categories in epistemic rules must be easily revisable in order to account for new scientific techniques. Moreover, one could claim that if categories were believed to be easily revisable (and thus useless), then we would be left with no explanatory resources to account for the acquisition of the capacity for epistemic judgement and the (at least medium-term) stability of scientific methodologies. We think that this sort of response, as demonstrated above, is reasonable.

Of course, even though rules are not easily revisable, it remains that we cannot but consider them revisable simpliciter—and this is not solely a psychological fact, but also a legitimate attitude. Thus, let us accept that all possible rules are inevitably and rightfully acknowledged as subject to potential revision. We might as well assume that in order for a rule R to have normative force for a rational agent A, the latter must believe that R is probably right. However, unlike the case of easy revisability, no problem stems from these assumptions, since A’s belief that R is probably right doesn’t entail that A completely rules out the possibility that R be revised. Acknowledging the possibility of error doesn’t annihilate normative force, since one can still believe that a rule is probably right, despite recognising—as a fallibilist—that some time one might come to believe it is wrong.

To our mind, even if an inquirer has a forceful belief that a rule is right, it would be irrational not to treat it as revisable. And we are happy to admit that the process of testing and potential revision of rules inevitably employs our judgements about particular cases. In fact, this is a key element of MRE. Thus, instead of being a blame, the indication of the possibility of rules’ revision in the face of particular intuitions appears to simply reassert our thesis that the Received View is an instance of MRE.

So far, we’ve tried to resist Maddy’s localist objections against SRD. Let us now turn the tables by focusing on localism’s positive instructions. Having renounced the quest for rules in favor of particular case-studies, what’s the role of epistemology for localists? According to Maddy, it is “to flesh out [each particular] story, … to reveal and explicate the rationality of… [scientific] claims based on each of these particular varieties of evidence” (2007: 407). But, what is to rationalize a scientific claim? Doesn’t it amount to showing that the inference supporting it conforms to some “recognised general form[s] of good scientific reasoning”? (Henderson 2018: 161) What else could it be (apart from reciting first-order evidence)? Indeed, according to the dominant (and, in a way, common-sense) meta-epistemological (and meta-ethical) view known as generalism, which is presupposed by the Received View, the very possibility of normative judgements depends on the provision of rules, so that, bluntly put, an agent A is justified in believing p only if, in believing p, A conforms to a set of epistemic rules that are acknowledged as end-promoting. Hence, to rationalize a claim, even in the sense of an ex post facto rationalization (as implied in Maddy’s quote right above), requires (retrospectively) demonstrating that it is accountable by some rules.

This requirement is imposed by some irresistible intuitions. Normative properties are hard to be seen as primitive. When a particular inference x is recognised as ‘good’, we cannot rest with saying that this is a brute fact; that x is good, period. An explanation of why x is good is needed. But, isn’t explanation essentially a generalizing process, involving the subsumption of x under some ‘regularities’ (viz. rules) regarding the correlations between non-normative and (supervenient) normative properties?Footnote 8 Moreover, since justification is conceptualized as a matter of obligation, permission or prohibition, isn’t it obvious that we should treat it in terms of accordance with, or violation of, rules (Goldman 1986: 59)?

These considerations are meant to reveal the attractiveness of generalism about the nature of reasons. Obviously, they are not conclusive against a particularist conception of rationality, such as the one famously advocated by Jonathan Dancy in meta-ethics (2004) (a “piecemeal approach” according to Maddy (2007: 403)), which claims that the possibility of justification in no way depends on the provision of rules. No doubt, there do exist several particularist arguments (mostly indicating the difficulties in articulating a generalist ethics). However, there are even more overwhelming arguments for generalism (and against particularism).Footnote 9

At any rate, making a case for localism is not as smooth as several scholars think. As we see it, if there is a rightful subversion of the two-level image, it’s not one that drops the second-order level, but one that (methodologically) merges both levels by allowing ‘bidirectional exchanges’ between them. The need to conduct inquiry in accordance with this equilibrating scheme is ultimately a natural outcome of the fact that, as Stanford (2016: 93) puts it, “understanding what… scientific theories are telling us about the world and understanding how we go about entheorizing that world in the first place are not distinct challenges: both are part of the overarching and more fundamental challenge of trying to simultaneously understand both the world and our own place within it.”

3.3 No Dissolution, After All

So, the dissolution of the Received View in favor of an externalist or localist one is not promising. However, this doesn’t mean that the suggestions of sociological externalists and localists are totally flawed, but rather that the corresponding projects are incomplete. And it’s arguable that, although their suggestions cannot replace the Received View, they can add to it. In this spirit, this section acts as a partial vindication of the dissolvers by properly placing their insights into the dialectics of SRD or by showing how they are already incorporated into the Received View.

How the key insight of localism is embedded in the Received View becomes apparent via a gross meta-philosophical observation. As is known, an important aspect of the naturalistic turn in philosophy of science in the second half of the twentieth century was the recognition that epistemological theories should be assessed inter alia by “confronting” them with actual scientific practice (Schickore 2011). The Received View adopts this idea, in so far as the wanted epistemic principles are not supposed to be captured by epistemologists’ intuitions and be evaluated (solely) a priori, but should be extracted through case-studies and be (partly) evaluated according to how well they accommodate concrete instances of seemingly good scientific reasoning. We take it that, although some kinds of localism might involve more radical theses, this is the key idea of localism in general; i.e., that principles should account for particular judgements of competent epistemic agents. As long as the Received View espouses this idea, it absorbs the core of localism.

Nevertheless, according to the Received View, the goal of epistemology isn’t taken as the mere explication of intuitions about some determinate considered judgements. Instead, “no special epistemological priority is granted the considered … judgments” (Daniels 1979: 265): showing that a particular judgement conforms to principles is thought of as an indispensable part of its justification, while background considerations are deemed as capable to affect its epistemic status. Overall, localism accounts for, so to speak, the half of the Received View. Looking into case-studies is necessary, but it doesn’t exhaust the role of philosophy vis-à-vis science.

Coming to sociological externalism and, specifically, to Vickers’ criterion, its place within the MRE-scheme is rather evident. Looking for scientific assertions enjoying the confidence of the relevant community is equivalent to searching for, and filtering, the set of considered judgements meant to be the starting points for epistemological theory construction. Vickers’ criterion, to begin with, provides a quantification (95% of the community) of MRE-theorists’ demand that the considered judgements must be “widely shared” (Bates 2004: 49). Moreover, his requirement that the community must be large and diverse—as is obvious from his analysis (2023: ch.5 and ch.9)—can be thought of as ensuring MRE-theorists’ request that the collected particular judgements “have been made under conditions conducive to avoiding error” (Daniels 1979: 258); e.g., they are impartial, free of cognitive biases, sufficiently scrutinized, etc. Lastly, requiring that the community must be “relevant”, i.e. comprised of scientists with the appropriate expertise (Vickers 2023: 96–7), reflects MRE-theorists’ demand that the particular judgements taken into account should have been made by “competent judges” (Scanlon 2002: 143; DePaul 1993: 174). In sum, given the advanced nature of research at the frontiers of science, there appears to be no other option for epistemologists apart from collaboration with scientists themselves. Perhaps, this is the ultimate practical lesson to take home from aspiring dissolvers.

4 When is a Circle Rational?

So far, we have suggested—contra externalism and localism—that second-order inquiry into epistemic criteria regarding types of theoretical success is ineliminable for SRD. We also made the descriptive claim that this inquiry—the Received View—is conducted in accordance with MRE. But we haven’t argued why this must be so. In this section, we provide a brief argument for this prescriptive claim. This won’t be a ‘positive’ meta-epistemological argument. Instead, the consideration supporting MRE is simply that there cannot be another method, or, more appropriately, as DePaul has extensively argued, that “any other approach to philosophical inquiry is irrational” (1998: 300). And if there is no other rational method, the remaining one is imperative.

Note that this argument, if persuasive, is also a kind of reply to (some) criticisms of MRE, such as, interestingly for us, that MRE is epistemically circular. That is, apart from arguing that the circularity is not vicious, but “a virtuous one” (Goodman 1955: 67)—which is a prominent way to respond to the said criticism—an MRE supporter might claim that, since there is no other rational method, the only one available—sceptic epoché excluded—should be good enough, even if circular. Admittedly, this response cannot be entirely satisfying to everyone. It’s merely a response aimed at providing some therapy, in the sense of mitigating the anxiety caused by (i) the fact that we’re trapped in the circle with no visible way out and (ii) the suspicion that there must be some good way out that we just haven’t found yet. For, if one is convinced that this suspicion is wrong, one can learn to live calmly within the circle.

To see why MRE is the only rational method, let’s state very abstractly what it consists in. As DePaul puts it, MRE is a method directing the inquirer to reflect upon the logical and evidential relations between the members of J, P and B, and to resolve conflicts “by revising beliefs in a way that… seem[s] most likely to be correct… after taking into account everything she believes that might be relevant” (1998: 301). Any alternative to MRE must oppose these directives.

Putting aside non-starters (such as a “method that does not allow the results of a person’s reflection to determine what she accepts” (ibid. 302)), the only prima facie viable alternative would be a method directing the inquirer “to reflect incompletely”, i.e., “to leave certain beliefs, principles, theories, or what have you out of account” (ibid.). Such a method would allow certain beliefs to be held or discarded without reflection. Note that, to be a real alternative to MRE, this method must permit acceptance or elimination of beliefs without any ground for doing so. For, if one had any reason for doing so, then one would not simply be excluding these beliefs from account. Rather, the beliefs would have been reflected upon, and either supported by some considerations or shown to be in conflict with more strongly held beliefs.

But how could one ever be in a position to consider oneself justified in believing or disbelieving something for no reason at all? As DePaul notes, such a person must “be alienated from the part of herself responsible for the judgements being excluded and have given up on this part of herself without any reason for [trusting or] distrusting this part” (ibid. 304). Otherwise, she blindly submits to the demands of an external authority, and this for no reason; i.e., not because she believes that the authority is reliable, for this would be a reason. We take it that the said features of such an epistemic situation provide sufficient grounds for considering the corresponding method irrational. But one may be more precise. Arguably, believing things without any thought would make it highly unlikely to arrive at a coherent view about a subject. Furthermore, DePaul (ibid. 303) cautions that a method of incomplete reflection like the one just described involves a high risk of a kind of self-contradiction:

It very commonly happens that when we reflect upon something we believe, we uncover among the other things we believe, or come to believe as a result of our reflections, reasons for… rejecting our initial beliefs… Thus, I say, a person [who adopts the teachings of an authority or] who fails to reflect and goes along believing what he has always believed is liable to believe what he himself does not find most acceptable … I think it is irrational to believe what one finds unacceptable, and a method of philosophical inquiry that directs inquirers to put themselves into such a position cannot be rational.

These considerations suggest that, as Thomas Scanlon puts it (2002: 149), MRE “is the only defensible method; apparent alternatives to it are illusory.”

5 Concluding Thoughts

That progress in philosophy cannot be like that in sciences, due to the very nature of philosophical problems, is an idea that we, of course, adopt, and one contributing to the same kind of therapy as above: it relieves one from the suspicion that there must be a currently unknown way to accomplish an unattainable goal. Since this idea is obvious and has been elaborated by philosophers of, say, Bertrand Russell’s caliber (1912/1997: 154–5), we prefer not to develop it, but rather to add some thoughts that are tangential to it.

First, one could note that the reports on the lack of progress in SRD are exaggerated. Contrary to Vickers’ pessimism about the chances of a consensus, philosophers, such as Stanford (2021: 216), have noticed a substantial convergence, that is a set of commitments that are embraced by most SRD-participants. The extent of such a convergence is easy to overlook. This is partly due to our failure to distinguish between two types of questions posed within a research field. The first are what might be called the fundamental questions, i.e., the ‘big’ questions “whose resolutions are among the field’s ultimate (perhaps even constitutive) aims at any time” (Bengson et al. 2022: 150), whereas the second are the field’s ordinary questions at a specific time. When considering a philosophical discipline’s success in reaching widely accepted answers to fundamental questions, it’s reasonable to lose heart. But, when it comes to ordinary questions, things are arguably more auspicious. (An example here would be the disappearance of any allegiance to semantic instrumentalist views of theoretical discourse in science, that is of views treating theoretical terms as meaningless symbols which are either eliminable or merely syntactic devices aiming at prediction and control (see Psillos 1999, chapter 2)) One might say that things aren’t so different in sciences. For example, if we consider fundamental questions such as those about the “origins of the universe, the basic building blocks of matter, and the… genesis of life” (Bengson et al. 2022: 151), a putative feeling of dissatisfaction wouldn’t be so surprising. The point is that, although progress in philosophy is by no means comparable to that in sciences, once we focus on the respective progress achieved regarding ordinary questions, philosophical research programs—like SRD—may not seem as stagnant as it is assumed to be.

The tendency to focus on fundamental questions while evaluating progress is common in SRD. Dissolvers often rely on claims about the failure to find a “global recipe” for realist commitment (Saatsi 2017) or a “silver bullet” type of evidence (Vickers 2023: 74), viz. failure to answer the fundamental question of the Absolutist Received View. However, focusing on the grim prospects of a final response to this question shouldn’t lead us to conclude that SRD is flawed in general. For, as we’ve seen, more modest versions of the Received View—like the Pro Tanto—are possible and actually more in tune with current philosophical practice. And to push things a bit further, even if the fundamental aims of the Pro Tanto View are unfeasible as well, this should not discourage us. For, a response to fundamental questions of philosophical inquiry must be considered as a regulative ideal, rather than a straightforward requirement. Relatedly—and importantly for us—this point has also been made in respect to MRE. As DePaul notes, the state of wide reflective equilibrium itself is “a state which no one could seriously hope to attain” (1993: 22).

It is in philosophy’s continuing work—in the journey, not in the destination—that we should look for the goods. And looking at SRD’s journey, one finds several achievements. Admittedly, these do not take the form of an accumulating body of knowledge like in sciences and, perhaps, as Vickers says, are far from the “firm conclusion[s] that might be held up to outsiders as a noteworthy achievement” (2023: ix). But such a demand would be uncharitable to philosophy, or else too rose-tinted.