The Source of Epistemic Normativity: Scientific Change as an Explanatory Problem

2.1. Instantiations of (Scientific) Naturalism

One of the most famous and most explicit attempts to form a naturalistic approach to scientific change and scientific knowledge in general is Quine’s (1969) Epistemology Naturalized. Based on the rejection of the two central tenets of logical positivism (Quine 1961), and on the consequent conviction that Carnap’s project was futile, Quine concluded that the only viable option for empiricism is naturalism. Instead of trying to provide a rational reconstruction of the language of science, we should attempt to find out what is the actual relation between scientific beliefs and their evidences; the way to achieve this is the empirical study of the relation in question (Quine 1969, 82-3). According to his view, psychology is the branch of natural science which takes empirical knowledge as its subject matter; hence, epistemology becomes a chapter of empirical psychology. ²

A few years before Quine’s paper, Thomas Kuhn published The Structure of Scientific Revolutions (Kuhn 1962, Structure hereafter) wherein he famously presented scientific change as the revolutionary replacement of an old paradigm by an incommensurable new one. Kuhn never self-consciously endorsed either naturalism or cultural relativism. On the contrary, in his later work, he tried to place his historiographical account into a neo-Kantian philosophical framework (Bird 2002; Kuhn 2000a, 104; 2000b, 264; 1993, 331). Nonetheless, both in Structure and in his other early writings (Kuhn 1970, 20-1) Kuhn did employ a language that justified both the characterization of naturalism ³ and the accusation of relativism. ⁴

Kuhn’s Structure triggered “the historical turn in the philosophy of science” (Bird 2014). The core idea of this turn is that our philosophical accounts of science should be substantially history-informed. As the opening lines of the Structure state, “[h]istory, if viewed as a repository for more than anecdote or chronology, could produce a decisive transformation in the image of science by which we are now possessed” (Kuhn 1996, 1). Despite its attractiveness, this idea is not philosophically unproblematic. According to the traditional conceptions of the philosophy and history of science, philosophy of science is a normative enterprise (Burian 1977; Lakatos 1978c; Pinnick and Gale 2000; Shapin and Schaffer 2011) which attempts to determine the rules that govern or should govern scientists’ belief modification. History of science, on the other hand, is a descriptive enterprise which seeks to recount scientific developments of the past without being concerned with evaluative questions about these same developments. ⁵ If this is the case, then given “Hume’s Law” (Hume 2000, §3.1.2, 302), which entails that we cannot derive a norm from a fact, the historical case studies cannot be relevant to the philosophical framework. ⁶ This problem has been repeatedly noticed in those accounts that attempt to integrate history and philosophy of science (Schickore 2011). Many philosophers and historians of science (e.g. Donovan et al. 1992; Giere 2011; Laudan et al. 1986) chose to overcome this problem by stripping philosophy of science of its normative dimension.

Another influential instantiation of the naturalization of science studies is the so-called the Strong Programme (Bloor 1991) in the sociology of science. As Barry Barnes, David Bloor, and John Henry have eloquently put it: “[f]or the scientist the world is the object of study; for the sociologist it is the scientist-studying-the-world that is the object” (Barnes, Bloor and Henry 1996, 30). Bloor (1991, 4-5) famously suggested that sociology of science should adhere to four tenets: causality, impartiality, symmetry, and reflexion. These tenets should be the backbone of mature sociological research which takes the content of scientific knowledge as its subject matter.

2.2. The Common Naturalistic Root

The perspectives that I have briefly described contain many differences with regard to multiple levels of philosophical analysis. However, I want to remain focused on the common feature of all these perspectives, that is, that they are all an expression of scientific naturalism. Scientific naturalism can be seen as a philosophical combination of two doctrines: (a) the ontological doctrine which implies that “ [t]he world consists of nothing but the entities to which successful scientific explanations commit us” (Macarthur and De Caro 2010, 4); and (b) the methodological doctrine which implies that scientific inquiry is the only genuine source of knowledge (Macarthur and De Caro 2010, 4), and thus philosophical inquiry can be viewed as genuine only if it is “conceived as continuous with science” (Macarthur and De Caro 2004, 3).

In what follows, I want to dwell exclusively on the methodological dimension of scientific naturalism. The above-mentioned perspectives on science are instantiations of scientific naturalism since they presuppose that the only way to make scientific belief modification intelligible is by employing the conceptual tools of one or more empirical sciences. Scientific change can be thoroughly understood by the same patterns of explanation as any other kind of belief modification and actually any other kind of fact in the world. In particular, this means that the explanations that are used to grasp scientific change share or should share the same logical form as the explanations used by the empirical sciences to grasp any other fact in the world.

3.1. Empirical-Scientific Explanations

The response to that question needs to overcome a number of philosophical problems. First, I will refer to the problems I won’t discuss and therefore which positions I will take for granted. Then, given these positions, I will discuss the problem of diversity with respect to the logical form of explanations that are employed by the empirical sciences, and will attempt to show that, despite their plurality, we can single out a cluster of explanations which we may call empirical-scientific explanations (as opposed to another kind of explanation, i.e., the normative explanations). This distinction will help me to elucidate the philosophical problem of scientific change.

The first problem I won’t discuss is the ambiguity of the term “explanation”. As Bradford Skow (2016, ch. 2) recently argued, the term “explanation” may stand for a response to various types of questions. I am not going to discuss the polysemy of the term “explanation”. I will take for granted that explanations, as I mean the term, answer properly to why-questions. I will also take for granted, following Skow (2016, 17), that “[w]hy-questions admit more kinds of answers than other wh-questions.”

The other topic I am not going to discuss is whether there is a substantial gap between the natural sciences on the one side and the social and human sciences ⁷ on the other. I acknowledge that we may detect differences in the explanatory patterns used by the various disciplines but these differences should not be understood as following a clear-cut division between the world of human beings and the rest of the natural world. I agree with perspectives which suggest that “ [t]he types of knowledge gained of the social world are much like the types of knowledge we can claim of the biological world” (Mitchell 2009a, 131; see also Kaiser and Plenge 2014, 2; Khalifa 2019, 287-8; Scholz 2014). I take for granted that there is a plurality in the logical form of explanations employed by the various empirical sciences. However, in contradistinction with the Hermeneutic tradition, this plurality neither entails nor presupposes any kind of predetermined division between ontological territories (the territory of nature and the territory of human minds).

Before I proceed with my main line of argument, one last preliminary remark is necessary. My aim is to examine which are the appropriate kind(s) of explanations in order to understand scientific belief modification. Within this context the term “understanding” has nothing to do with any particular sense of understanding. In particular, it has nothing to do with the Hermeneutic notion of Verstehen. It is also not to be equated with scientific understanding. As Khalifa (2019, 284) notes, “‘understanding’ [is] the genus that includes both Verstehen and scientific understanding.” ⁸ In what follows, I use the terms “understanding” and “making intelligible” interchangeably. Both denote the thorough comprehension or knowledge of the phenomenon under examination which is no other than scientific change. In Posterior Analytics (94a 20) Aristotle points out that “[w]e only think that we have knowledge of a thing (ἐπίστασθαι) when we know its cause.” Interestingly enough, Jonathan Barnes translates the same passage as follows: “we think that we understand something when we know its explanation” (Posterior Analytics, 94a 20). In the Aristotelian thought the notion of ἐπίστασθαι ⁹ (“having knowledge” or “understanding” something) is equated with knowing its causes or explaining it. I use the terms “understanding” or “making intelligible” as synonymous with the Aristotelian ἐπίστασθαι. They denote the thoroughest possible comprehension of the phenomenon under scrutiny (scientific change in this case). Of course, there are vast differences between the Aristotelian conception of ἐπίστασθαι and the one I am attempting to sketch here. ¹⁰ Nonetheless, like Aristotle, I link ἐπίστασθαι with the act of explaining.

Let me now move from the genus ἐπίστασθαι (“understanding” or “making intelligible”) to the species “scientific understanding.” As I have already said, we can detect a plurality in the logical form of explanations employed by the various empirical sciences. Admitting this kind of explanatory pluralism brings me to the problem that I want to discuss. The objection against the image I have already sketched could be the following: if there is a plurality in the logical form of explanations among the different special sciences, and if the various perspectives on scientific change that I described disagree about what is the most appropriate discipline in order to make scientific knowledge intelligible, why am I entitled to argue that there is a common naturalistic root to them? If different disciplines like history, sociology, psychology, or anthropology employ different kinds of explanations, and if the perspectives I described (and many others) take different disciplines as the privileged way of understanding scientific change, then what makes them all naturalistic? Instead of talking about a naturalistic turn in the science studies, shouldn”t it be more correct to talk about a historicist, or a psychologicist, or a sociologicist turn, and so on? What makes both historicism and psychologism, for instance, part of a naturalistic trend if we take for granted that history and psychology do not (at least necessarily) employ the same logical forms of explanation? Respectively, are we justified—as the title of this subsection implies—to talk about empirical-scientific explanations in general?

I suggest that, despite the explanatory pluralism, we are justified in talking about empirical-scientific explanations, and respectively about a naturalistic turn, in general. Despite the internal diversity of the concept of “empirical-scientific explanation,” it can be determined in contradistinction to the concept of “normative explanation.”

3.2. Normative Explanations

The way I use the term “normative explanation” is different and in one sense quite opposite from the way it has been recently used (Väyrynen 2013; 2019; 2020). In Pekka Väyrynen’s use of the term, what is explained by a normative explanation is why a thing has the normative feature it has. For instance, why a thing is good or unfair. On the contrary, in the way I use the term, a normative explanation explains a thing in the world such as the epistemic attitudes involved in scientific change or, for instance, why a scientific community endorsed a theory over a rival one. In Väyrynen’s sense, the normative feature is the explanandum while in my sense it is the explanans. A normative explanation can make a belief modification intelligible by showing that this modification conforms to a norm or a set of norms. In this respect, its general form looks like the rational explanations (Pettit 2002, 177), which are based on the assumption that “[r]easons must be capable of explaining actions” (McDowell 1998a, 108). ¹¹

Normative and empirical-scientific explanations can both be used in order to make epistemic attitudes intelligible. Empirical-scientific explanations make belief modifications intelligible by placing it into the causal ¹² order described by one or more empirical sciences (see Forman 1971). Normative explanations, on the other hand, make belief modifications intelligible by showing how they conform to one or more epistemic norms (see Lakatos 1978a, 68). Normative explanations of scientific change presuppose one or more epistemic norms (e.g., simplicity, coherence, accuracy) which serve also as implicit definitions of scientific rationality. In general, they have the form “scientific theories are/have/do X.” These judgements are normative for they imply how someone should modify their beliefs if they want to think as a scientist. Every single one of them can be transformed to the imperative grammatical mood. For instance, “endorse theories which provide accurate predictions.” They also serve as implicit definitions of scientific rationality and imply what is to be considered as scientific rationality and what is not. They do so not by providing a set of mechanical rules in a Cartesian style but by showing how to appraise belief modifications (Lakatos (1978c, 140). Furthermore, based on the principles above we can form normative explanations of past belief modifications. Given a norm (N) or a set of norms (N₁, N₂, . . ., N_n) and a past belief modification (b1→b2), one can explain the acceptance of b2 as an application of N.

3.3. The Distinctive Logical Form

Despite the diversity that may characterize the explanations employed by the empirical sciences, they can all be ranked in a sole category and in contradistinction to the kind of intelligibility which corresponds to normative explanations. My point is that the internal differences among the empirical sciences, on the explanatory level, fade out in light of the difference between placing something in the causal order assumed by one or more empirical sciences and showing how it conforms to one or more norms. The main difference relies on the fact that normative relations presuppose the possibility of failure while causal relations don’t. ¹³ It is possible for a belief-formation or belief-modification to fail to conform to an epistemic norm. This does not necessarily imply something for the validity of the norm in question. On the contrary, a phenomenon which does not adhere to the causal order described by an empirical science always leads to the revision of the assumed causal order in question. ¹⁴

It is important to stress that the concept of failure does not have a statistical dimension. It is not the same as the concept of exception. Exceptions can be part of the logical form of the explanatory patterns employed by some empirical sciences, for example, history or biology. As Schurz (2004, 181) stresses, “most law hypotheses in [. . .] (applied) sciences are not strictly universal but admit of exceptions. Their linguistic form is not “All As are Bs,” formally, ∀x(Ax → Bx), but “As are normally Bs”. Scriven (1959) calls these kinds of laws “normic.” The logical form of normic laws includes a clause that is expressed in the linguistic form of an adverb such as “typically,” “usually,” “naturally,” or “probably” (Scriven 1959, 465). For instance,

The concept of exception in these kinds of laws clearly has a statistical dimension. It would be absurd to claim that “donkeys naturally have four legs” if most of the donkeys were two-legged.

The judgements that sustain normative explanations, on the other hand, are characterized by a particular kind of generality. They share the type of generality that characterizes Michael Thompson’s (2008) “natural-historical judgments”: ¹⁵ “A natural-historical judgment may be true though individuals falling under both the subject and predicate concepts are as rare as one likes, statistically speaking” (Thompson 2008, 68, respectively). The logical form of natural-historical judgements makes them closely related to normative judgements given that they can determine the conditions of normality and defectiveness (Thompson 2008, 74). I don’t want to discuss the notion of a natural-historical judgement here. I focus on normative judgements and especially epistemic normative judgments. My point is that their logical form shares the same type of generality as Thompson’s natural-historical judgements. That is why the concept of exception does not play a key role here and cannot be identified with normic laws.

The crucial difference between the logical forms of empirical-scientific and normative explanations of scientific change can be revealed if we take into consideration the notion of direction of fit (DoF hereafter). The notion is usually considered (Blackburn 2016) as deriving from Anscombe’s (2000) Intention. In this influential book, Anscombe (2000, §32) gives the following example:

Let us consider a man going round a town with a shopping list in his hand. Now it is clear that the relation of this list to the things he actually buys is one and the same whether his wife gave him the list or it is his own list; and that there is a different relation when a list is made by a detective following him about. If he made the list itself, it was an expression of intention; if his wife gave it [to] him, it has the role of an order. What then is the identical relation to what happens, in the order and the intention, which is not shared by the record? It is precisely this: if the list and the things that the man actually buys do not agree, and if this and this alone constitutes a mistake, then the mistake is not in the list but in the man’s performance [. . .]; whereas if the detective’s record and what the man actually buys do not agree, then the mistake is in the record.

The example is used to cast light on the differences between practical and theoretical thought, between mental states such as intentions or desires and mental states such as beliefs. The latter follow a mind-to-world and the former a world-to-mind DoF. The distinction has many implications in philosophy of mind and speech act theory which are far from being uncontroversial. I don’t want to delve into these implications since I am not concerned with the distinction between practical and theoretical thought. ¹⁶ However, I suggest that the concept of DoF may clarify the distinction between the logical forms of empirical-scientific and normative explanations of scientific change.

Empirical-scientific explanations are based on causal orders presumed by the empirical sciences. The causal order should fit the actual belief modification that is observed in actual scientific practice. In cases of discrepancy between what the causal order predicts ¹⁷ and the observed episodes of actual belief modification, the considered causal order should be revised (mind-to-world fit). Normative explanations are based on epistemic norms which are expressed in normative judgements. Unlike the presumed causal order, it is possible for a normative judgement to be correct even if the observed episodes of actual belief modification do not conform to what the judgements in question dictate. We can consider an epistemic norm valid even if very few episodes of actual belief modification are conformed to it. In these cases, we think that the actual belief modification should have been adjusted to what the epistemic norm dictates (world-to-mind fit). In the cases I examine, the term “world” stands for the actual belief modifications that we observe. I will say a few more things about the correctness of the epistemic norms in the next two sections. For now, I just want to point out the difference in the logical form between the normative and the empirical-scientific explanations: the latter are based on a presumed causal order which always needs to fit the episodes of actual belief modification, while the former are based on epistemic norms which, as far as they are presumed valid or correct, determine the epistemic content to which the episodes of actual belief modification should be adjusted.

4.1. Scientific Naturalism and the Eliminability of Normativity

Scientific naturalism relies on the premise that accepting a view of the human mind as part of the natural world (“metaphysical naturalism” Risjord 2014, 9) necessarily entails epistemological or methodological naturalism, that is, the position that empirical-scientific understanding is the only legitimate way of making things intelligible in the world. Furthermore, according to the majority of scientific naturalists, metaphysical naturalism also entails meta-philosophical naturalism (Risjord 2016, 2), that is, the view that philosophy is somehow continuous with the empirical sciences or even a branch of them. As far as concerns the logical forms of explanation which could and should be employed in order to make scientific change intelligible, this perspective could be simply defined as the rejection of the genuineness of normative explanations. In short, scientific naturalists are anti-normativists (Turner 2010), for they think that there is nothing “beyond the reach of the [empirical]-scientific understanding” (McDowell 2008, 217) ¹⁸ and therefore, scientific change, just like every other happening in the world (i.e., in nature), should be made intelligible by employing empirical-scientific explanations rather than any sui generis type of explanation like normative ones. Stephen Turner (2010, 30) expresses this idea very clearly: “The naturalistic response [to normativists] starts from the explanation end: if there is no explanation of something to be had within the ordinary stream of explanation, maybe that thing doesn’t exist.” The ordinary stream of explanation is no other than the explanations provided by the empirical sciences. The common aspiration of all the approaches I presented in §2 is to place scientific change into the ordinary stream of explanation.

Scientific naturalism leads to explanatory reductivism and hence it is programmatically eliminative with regard to the normative content. According to every version of scientific naturalism, the complete comprehension of scientific change presupposes the reduction of normative explanations to empirical-scientific ones. Given an actual episode of scientific change, that is, a belief modification b1→b2 and a set of norms (N = N₁, N₂, . . ., N_n), normative explanations make b1→b2 intelligible by showing how the modification conforms to N. Scientific naturalists claim that normative explanations are not genuine since there is always a further explanatory step: one should explain why the actual thinkers (e.g., the scientific community) took the content of N as normatively right and also why they took b1→b2 as an application of N. This further explanatory step can only be performed by placing the belief modification into the causal order presumed by the empirical sciences. Therefore, normative explanations are at best provisional and at worst illusionary ways of understanding scientific change; they need to be reduced to a set of empirical-scientific explanations. Furthermore, no normative content can play a genuine explanatory role with regard to the scientific belief modification, and hence normative explanations are programmatically eliminable.

4.2. Why Liberal Naturalism Is Naturalism

Liberal naturalism rejects the idea that ontological naturalism necessarily leads to methodological or meta-philosophical naturalism. On the explanatory level, this entails that, at least in some cases, normative explanations are indispensable in order to make belief modifications intelligible. Genuine (as opposed to merely seemingly) ¹⁹ normative explanations are not reducible to empirical-scientific ones, and therefore the normative domain is not eliminable. There are several versions of liberal naturalism which attempt to defend the ineliminability of normativity in various ways. I am not going to discuss the various strategies. In the rest of the present subsection I will focus on arguing why the ineliminability of normative explanations can be genuinely naturalistic, and thus compatible with the so-called “Modern Scientific Worldview” (Korsgaard 1996, 18). After that (in §4.3), I will argue in favor of the liberal rather than the scientific version of naturalism, always with respect to scientific change.

Understanding normativity as an explanatory concept and also understanding normative explanations as ineliminable does not entail an appeal to any kind of spooky entities characterized by curious causal efficacy or other mysterious epistemic faculties. ²⁰ It implies that normative explanations make things in nature intelligible by adding a further level of understanding. This further level of understanding includes the relation between the rational subject and the world in a normative way. For example, take the case in which I believe that the traffic light in front of me is red. What constitutes a complete explanation of my belief? In other words, what is a complete answer to the question, “Why do I believe that the traffic light is red?.” According to the scientific naturalist, a complete reply involves the interaction of the optical stimulus and my neurophysiological infrastructure, my psychological condition, the social context of my upbringing, etc. In general, a complete answer would involve placing my belief into the causal order presumed by the empirical sciences. The liberal naturalist, on the other hand, could argue that normative explanations add a further level of understanding. This further understanding includes a subject’s following one or more epistemic rules, ²¹ which entails the subject’s endorsement ²² of the corresponding epistemic principles. This kind of understanding is sensitive to the truth of the belief and actually makes the truth of the belief part of the explanation of belief modification. The successful application of the correct epistemic principles leads to the truth, and truth is part of how we make epistemic facts intelligible. In this sense, normative explanations add a further level of understanding to what facts themselves involve, as in the epistemic fact of my example. The further level of understanding includes the relation between the knowing subject and the world in a normative way. It does not appeal to any kind of spooky entities or epistemic faculties.

The claim that normative explanations can add a further level of understanding to epistemic facts can be revealed if we take into consideration the case in which I believe that there is a red traffic light when actually the traffic light before me is green. Just like in the case above, in which my belief is correct, the scientific naturalist will suggest that the full explanation of my belief merely involves the interaction of the optical stimulus and my neurophysiological infrastructure, my psychological condition, the social context of my upbringing, etc. The only difference will be in the initial conditions of the process: in the second case I may suffer some kind of color blindness or there might be inappropriate lighting conditions which prevent me from acknowledging that the traffic light is actually green. In these two cases the scientific naturalist can only notice two different causal chains. This is not dissimilar from another example involving a set of two cases which also illustrates the point: one in which smoking and excessive alcohol consumption leads to a heart attack in the one case yet relatively healthy heart function in the other. In this sense, the scientific naturalist is blind with respect to an essential aspect of the epistemic facts, that is, the relation between the knowing subject and the world in a normative way. Strictly speaking, the scientific naturalist cannot know whether the belief is correct or not, for correctness precisely means conformity to a norm. But also, the correctness or falsity of a given belief is additional information about the world, that is, about the relation of the knowing subject and the object of knowledge. The liberal naturalist perspective, by preserving the autonomy of the normative explanations, is sensitive to the correctness of the given beliefs and therefore it can grasp the relation between the knowing subject and the world in a normative way.

At this point, the scientific naturalist may claim that her perspective does not preclude all kinds of norms. Instrumental norms can be part of every (or at least most of the) scientific naturalist account(s). ²³ But I think that instrumentalism does not overcome the “blindness” that I attributed to scientific naturalism. The point is not whether or not an account acknowledges the genuineness of norms as hypothetical imperatives. The point is how an account explains the norms as categorical imperatives. Only as categorical imperatives can norms avoid the programmatic reduction to empirical-scientific explanations and play a genuine explanatory role concerning beliefs. Scientific naturalists may acknowledge that R (a methodological rule) is instrumentally the best way to achieve N (a cognitive goal or an epistemic norm) and in this sense there is room for instrumental rationality in their account. However, the crucial issue here is the status of N. According to every version of scientific naturalism, endorsing N needs to be explained by a set of empirical-scientific explanations. In this sense, N is eliminable. On the contrary, in my view, N can sustain a genuine normative explanation of a particular belief modification and therefore is ineliminable. If I am right and if N is a genuine epistemic norm, the normative explanation in question adds a further level of understanding concerning the relation between the knowing subject and the world. ²⁴

Claiming that normative explanations add a further level of understanding to the epistemic facts does not imply any kind of super- or non-naturalism. The normative judgements which sustain the normative explanations cannot violate the possibilities which are determined by the empirical sciences. Just like chemical or biological phenomena (and the corresponding causal order presumed by chemistry or biology) are constrained by the conditions of possibility determined by physics (Mitchell 2009b, 58-64), the normative judgements which sustain normative explanations are constrained by the conditions of possibility determined by the totality of the empirical sciences. ²⁵ In this sense, there is nothing transcendent or supernaturalistic in the idea that we can explain a particular belief modification by appealing (also) to the epistemic relation of a knowing subject and a known object as it is expressed in the act of conforming (or failing to conform) to specific norms.

Finally, defending the ineliminability of normative explanations does not entail anything non-naturalistic even on the methodological level. It does not imply that we should appeal exclusively to “armchair philosophy” or “a priori analysis” in order to determine the content of the normative judgements which sustain the normative explanations. As I said in §2, the key idea of the historical turn in the philosophy of science was that the philosophical accounts of science should be substantially history-informed. This means that historical research about the actual belief modification of scientists in the past should be taken into account in the process of determining the rational principles that govern scientific belief modification. Determining the content of the epistemic norms which are involved in the explanation of scientific belief modification entails empirically acquired knowledge of the actual belief modification of the scientist in the past. This idea should not be construed as necessarily implying the transformation of philosophy into an empirical theory of science by stripping away its normative content. There is an intermediate logical space between normative absolutism, that is, the position that normative standards of scientific rationality are a-historical and exclusively determinable by a priori philosophical analysis, and anti-normativist historicism, that is, the position that the standards of scientific rationality are strictly idiosyncratic and lack any normative dimension. The intermediate space is occupied by a liberal naturalist view which retains the normative character of philosophy and, at the same time, acknowledges that it should be substantially history-informed. ²⁶ I cannot describe the details of this view here. ²⁷ Nonetheless, I should say that this view, unlike normative absolutism and like historicism, entails that the (historical) study of the actual episodes of scientific change may drastically shape our conception about the standards of scientific rationality. That is why it can be understood as naturalistic. However, unlike anti-normativist historicism, this view does not equate the process of shaping our normative accounts of scientific rationality (by studying the actual episodes of scientific change) with the empirical testing that leads to the revision of empirical-scientific theories. That is why it is not a version of scientific naturalism.

To sum up, defending the ineliminability of normative explanations does not necessarily violate naturalism. The ineliminability of normative explanations is compatible with metaphysical naturalism since it does not presuppose any kind of spooky entities or epistemic faculties. It is also compatible with a moderate version of methodological naturalism since (a) the content of normative judgements is constrained by the possibilities defined by the empirical sciences, and (b) the content of normative epistemic judgements can essentially be informed by the study of the actual episodes of scientific change. The ineliminability of normative explanations violates only the scientific version of naturalism, not naturalism per se.

4.3. Liberal Versus Scientific Naturalism and Scientific Change

This brings us immediately to the following question: What (if any) is the ground of the ineliminability of normative explanations or what (if any) is the source of epistemic normativity? The reply to this question is crucial in defending the liberal over the scientific version of naturalism. Part of the reply has already been sketched in the previous subsection but a few more complementary things need to be said. Normative explanations add a further level of understanding to the epistemic facts: the relation between the rational subject and the world in a normative way. Now it may seem that it is just circular to claim that normative explanations add a further level of understanding which concerns the relation between the rational subject and the world in a normative way. This is exactly what scientific naturalism rejects, the idea that there are genuine normative relations, that is, normative relations which cannot be reconstructed out of conceptual materials that belong to the empirical-scientific understanding. However, I suggest that upon a closer examination this remark may prove more meaningful than a mere circularity, for it can reveal the deep philosophical root of the difference between the scientific and liberal versions of naturalism with regard to epistemic normativity.

The main difference relies on whether or not it is considered that the truth-value of a belief is or can be part of the explanation of its modification. Scientific naturalism rejects—implicitly or explicitly—the idea that truth plays an explanatory role in belief modification processes. As Ian Hacking (1992, 14) stresses, “[t]he truth of a proposition in no way explains our discovery of it, or its acceptance by a [knowing subject], or its staying in place as a standard item of knowledge.” This is also echoed in the Strong Programme’s impartiality and symmetry tenets. Elsewhere, Bloor (1999, 89) is even clearer on the issue: ‘“corresponding,” or “not corresponding” to reality, are not causal relationships that bodies of belief bear to their referent.” I suggest that this idea expressed by Hacking and Bloor is not characteristic of simply their version of scientific naturalism but of scientific naturalism ²⁸ per se. In other words, I suggest that the eliminative version of naturalism is at odds with every version of scientific realism and, consequently, leads necessarily to epistemic relativism (Turner 2020). As Psillos and Shaw (2020, 411) point out:

‘Scientific realism is a view of science which [. . .] claims that the truth of scientific theories (approximate truth, or partial truth, but still truth, viz., getting the world by and large right), is the best explanation of their empirical success and that the reason for being optimistic about science is precisely that past theories build on the successes of their predecessors. That’s precisely the view that SSK ²⁹ denies.’

The philosophical root of rejecting the genuineness of the domain of normative explanations altogether lies on the assumption that truth plays no role in the explanation of belief modification; normative relativism ³⁰ is then inevitable. ‘The relativist crux [. . .] is that there is no distinction between being-taken-to-be-justified (by a community) and being-justified’ (Psillos and Shaw 2020, 408). This is equivalent to my abovementioned claim (§4.2) that the scientific naturalist is blind to the relation between the knowing subject and the object of knowledge.

The ineliminability of the domain of normative explanations can be sustained if we seek the source of epistemic normativity in the essential role that truth plays in explaining belief modifications. ³¹ Truth can play this role only through epistemic norms. The epistemic norms are relatively stable (even if not infallible ³² ) ways of grasping truth. If an epistemic norm is valid and if its application is correct, then most of the time the application in question leads to the apprehension of a truth. Respectively, the normative explanations reveal the relation between the rational subject and the world in a normative way. They make the difference between being-taken-to-be-justified (by a community) and being-justified. The domain of normative explanations is essential in order to grasp the epistemic fact as such (i.e., as epistemic). In this sense, the normative explanations add a further level of understanding to what facts involve, a level to which the empirical-scientific understanding is blind.

One way to understand this kind of blindness is by moving into the territory of the empirical sciences. When a biologist attempts to explain the motion of a set of cancer cells, she doesn’t think she has to do it in terms of the chemical constituents of the cell. Likewise, when a chemist explains a chemical reaction, she need not to do it in terms of the elementary particles that constitute the chemical elements. The biologist explains the behaviour of the chemical constituents of the cells in terms of the biological system they form. Respectively, the chemist explains the behaviour of the physical constituents of the chemical elements in terms of the chemical system they form. Neither would the biologist deny that the cell consists of chemical constituents, nor the chemist that the chemical elements consist of physical particles. It is not the case that without the biological conceptual system the chemist would not understand what happens at all. She can observe the behaviour of chemical compounds and chemical reactions. But observing exclusively the chemical processes keeps her blind to phenomena such as metastasis. In this sense, the addition of the biological explanatory level enriches our understanding. ³³

In an analogous way, it is not that without the normative explanations we would not understand at all what happens in cases of scientific change. A lot of things could be understood about the sociological or psychological factors that are involved in any episode of scientific change. However, the addition of the normative explanatory level enriches our understanding by revealing the relation between the rational subject and the world in a normative way. Of course, from the scientific naturalist’s standpoint, there is nothing in the epistemic fact to which she is blind. The so-called normative dimension of the fact is a mere epiphenomenon which should be reconstructed out of the conceptual tools of psychology, sociology, and the other empirical sciences. Oppositely, according to the normativist standpoint, the sociological, psychological, or biological constituents of the fact cannot fully explain it. The full explanation—and therefore the full understanding (ἐπίστασθαι) ³⁴ —of the epistemic facts also requires normative explanations. This kind of explanation adds a further level of understanding concerning the relation of the knowing subject and the object of knowledge in a normative way. Moreover, assuming the genuineness of the normative explanation both presupposes and entails the tenability of scientific realism, that is, the tenability of the idea that the truth-value of the belief plays some role in the explanation of its formation or modification.

I don’t have a knock-down argument in favor of the normativist and therefore the realist standpoint here. There are too many interconnected philosophical issues that should be separately examined in order to defend the liberal over the scientific version of naturalism. However, my aim is not to defend liberal naturalism here. My aim is to detect the philosophical roots of the difference between the eliminative and non-eliminative versions of naturalism, and furthermore, to show that defending the autonomy of epistemic normativity can take place within a naturalistic framework. Finally, I aim to stress that liberal naturalism and the consequent ineliminability of normative explanations presuppose the defense of scientific realism and lead to the defense of scientific rationality. In my view, the combined defense of scientific rationality and scientific realism makes the framework of liberal naturalism, in principle, preferable.

4.4. Naturalism and Scientific Realism

At this point, I need to provide two brief elucidating remarks on scientific realism and on the relation between scientific realism and the two versions of naturalism. While it is true that there is little consensus in the relevant literature about what exactly scientific realism amounts to, we can agree with Boyd (1980, 613) that generally speaking, “[b]y” “scientific realism” philosophers ordinarily mean the doctrine that non-observational terms in scientific theories should typically be interpreted as putative referring expressions, and that when the semantics of theories is understood that way (“realistically”), scientific theories embody the sorts of propositions whose (approximate) truth can be confirmed by the ordinary experimental methods which scientists employ’. I take this doctrine to have a crucial impact on the way we explain scientific belief modification and scientific belief endorsement in general. As Bas van Fraassen (Chakravartty and van Fraassen 2018, 18) points out, the difference between a realist and an anti-realist is “not in what they believe about the natural world, but what they say about this belief.” In the terms that I have used throughout this paper, this means that the difference between realists and anti-realists lies in how they explain scientific beliefs. Scientific realists take the belief in unobservable aspects of reality, which are implied by the successes of certain scientific theories, to be rationally compelled. ³⁵

If this is the case, then from a naturalistic viewpoint, the genuineness of normative explanations as a distinctive kind of explanation can be sustained by the framework of scientific realism. Norms should not be taken as spooky entities or as the result of spooky rational processes. Following epistemic norms is a relatively stable (but not infallible) way of assaying both observable and unobservable aspects of the world. Respectively, normative explanations of belief modification reveal specific aspects about the relation between the knowing subjects and both the observable and unobservable parts of the world.

This takes me to the second remark. I have already suggested (§4.3) that the eliminative version of naturalism is at odds with every version of scientific realism. This may look bizarre since it is a fact that prominent defenders of scientific realism (e.g. Richard Boyd, David Papineau, Stathis Psillos, Phillip Kitcher and Ronald Giere) have also explicitly endorsed naturalistic epistemology. I think that this fact does not undermine my conclusion for two main reasons. First, the scientific realists who have also adopted naturalism have not, to the best of my knowledge, been concerned with the differences between the scientific and the liberal versions of naturalism. Hence, Boyd (1980, 625) for instance, describes the two central elements of naturalistic epistemology as follows: “the abandonment of a priori standards for the evaluation of inductive strategies and the employment of unreduced causal notions in the analysis of knowledge.” It is clear that the liberal version of naturalism as I presented it here is compatible with those two elements. Liberal naturalism does not preclude the employment of empirical-scientific explanations in order to make scientific change intelligible. It merely implies that there is also another kind of explanation which can and should be employed. If naturalistic epistemology is construed generally enough to include both versions of naturalism that I have described, then my claim that scientific realism is at odds with the scientific version of naturalism is not put at stake by the fact that there exist accounts which combine naturalism and scientific realism.

The second reason is that the accounts of some philosophers which attempt to combine scientific realism and naturalism are, in my view, incoherent. For instance, Ronald Giere (1988, 7-8) argues that both rationalists (those who think “that there are rational principles for the evaluation of theories”) and naturalists (those who think “that theories come to be accepted (or not) through a natural process involving both individual judgment and social interaction”) can be either realists or anti-realists. Among the naturalists, the proponents of social constructivism are anti-realists. Giere suggests that the cognitive study of science can help us construct a viewpoint which combines scientific realism and naturalism. “The most status a philosopher of science could claim is that of the theoretician of a developing cognitive science of science” (Giere 1988, 12). Giere’s version of naturalism is clearly scientific, for he explicitly suggests the replacement of traditional philosophy of science with an empirical discipline, that is, cognitive science. Furthermore, he chooses an “interventionist’ strategy in order to defend scientific realism akin to the one followed by Hacking (1983). ³⁶ He argues that the observation of scientists’ ability to manipulate protons in so many different ways is the ground for accepting that their realistic attitude towards protons is correct (Giere 1988, 125). He finally suggests that it is not enough to focus on the scientists’ beliefs but we should also explain what they do (Giere 1988, 125). In my view, Giere’s account is incoherent. The main reason for this is that judging scientists’ realistic attitude as justified requires also a normative rather than merely an empirical-scientific explanation of this attitude. Cognitive scientific explanations of either the belief modification or the behavior of scientists are no more sensitive to the relation between the knowing subject and the object of knowledge in a normative way than the sociological explanations. Both are blind to what counts as real manipulation of real entities as opposed to something-taken-to-be-manipulation of real entities. Therefore, the scientific version of naturalism in this cognitivist form is incompatible with scientific realism even in this interventionist form. It is not a matter of picking up the correct scientific discipline (cognitive science rather than sociology) for explaining scientific change in order to provide an account that is compatible with scientific realism. It is a matter of acknowledging the indispensability of a distinct kind of explanations (i.e., the normative ones). However, this move by definition violates the presuppositions of scientific naturalism.

5.1. Fallibilism and Demarcation

Traditionally, the idea that science is a rational enterprise is construed as entailing the existence of a criterion (or a set of criteria) which can demarcate the cases of genuine expression of rationality from the cases in which scientists only think they have made a rational choice. For years, it was considered that the main task of the philosophy of science is to determine such a demarcation criterion. The efforts in that direction faced various difficulties, one of which is the persistent counterexamples from the history of science. Laudan (1983), declared the demise of the demarcation problem and philosophers of science since then have given up, to a great extent, the effort to rationally demarcate science from other cognitive enterprises. ³⁷ Giving up the aspiration of demarcating the cases of genuine expression of scientific rationality left room for anti-normativist naturalist approaches. If there is no reliable criterion for demarcating the genuine expressions of scientific rationality, then there is no distinction between being-taken-to-be-justified and being-justified. Therefore, our only option is to explain why particular belief modifications were-taken-to-be-justified by particular communities. As McDowell (2010, 11) points out, “if what one [thinks] ³⁸ on is something that merely seems to one to be a reason, what ultimately explains one’s thought or action is whatever explains its seeming to one to be a reason”. This can be explained by appealing to various psychological or sociological factors.

However, demarcationism and anti-normativist naturalism are not our only options. As I already said, epistemic norms are stable but not infallible ways of grasping the truth. They spring from our long-lasting epistemic exchange with the world. The fact that by following them we are not guaranteed that we are going to grasp the truth does mean that we never grasp the truth by following them. The possibility of failure does not preclude the possibility of success. On the contrary, the possibility of failure implies that there is also the possibility of success. When we succeed to grasp the truth, part of the explanation of the belief we hold rests on the norms we followed and this cannot be revealed without appealing to normative explanations. Acknowledging that “there is no criterion, if by that we mean some general formula that it might be possible to apply to mark off genuine reasons from impostors” (McDowell 2010, 12), does not entail that there are no cases of genuine expression of scientific rationality. And those cases cannot be fully made intelligible without appealing to normative explanations. It just means that singling out such cases is a more complex process than a mere application of a demarcation criterion. It also means that normative explanations participate in the competition for explanatory adequacy with other kinds of explanations. They must prove their adequacy in each case that requires explanation. Nothing excludes a priori the possibility of reducing a (seemingly) normative explanation to a set of empirical-scientific explanations.

At this point, given what I have already claimed: (a) the difference between the logical form of empirical-scientific and normative explanations of scientific change, that is, that the judgements which sustain normative explanations can be correct despite the counterexamples that can be observed; (b) that the philosophical activity of determining the content of the epistemic norms which are involved in the explanation of scientific belief modification entails the empirically acquired knowledge of the actual belief modification of the scientist of the past; and (c) that there is no demarcation criterion for the cases of genuine expressions of scientific rationality, my account may seem incoherent. The problem is the following: given a judgement of the form “scientific theories are/have/do X,” and given an individual case that violates what this judgement dictates, how can we decide if this case is defective (unscientific) or if it is a genuine counterexample to our judgement? It seems that my perspective comes to an impasse. I can’t say too much here ³⁹ but I would like to argue that even if there is no criterion in the sense of a general formula that can predetermine what counts as a genuine counterexample, thereby forcing a revision of the judgement and what counts as a mere defective case, the problem is not unresolvable. The constant feedback loop between the individual cases and the judgements which express the content of epistemic norms can result in reflective equilibrium. We arrive at this equilibrium after a process of revision and sophistication of the judgements in light of the particular cases they concern. Epistemic norms cannot be characterized by the old Cartesian certainty. Their content should be considered as fallible and their revision should involve taking into consideration the empirically acquired knowledge of the individual cases that refer to the corresponding principles. In short, epistemic norms are subject to revision. History of science, in both senses of the term “history”—that is, as the historical study of past scientific episodes and as the historical course of science—may result in their revision.

5.2. Internalism and Methodological Individualism

Another aspect of the traditional conception of rationality ⁴⁰ is that the cases of genuine expression of rationality presuppose an individual subject’s own recognition of the right rule (epistemic norm). In Nicomachean Ethics, Aristotle (1144b25) states that the expression of rationality is not found in actions that are performed solely “in accordance with good reason” (κατὰ τὸν ὀρθὸν λόγον) but in actions that are performed out of a regard for the right reason (μετὰ τοῦ ὀρθοῦ λόγου). An analogous idea can be found in Kant (1998, 4:412): “Only a rational being has the capacity to act in accordance with the representation of laws, that is, in accordance with principles, or has a will.” Aristotle and Kant talk about practical reason in these passages. Nonetheless, the same goes for the cases of genuine expression of epistemic rationality. This is what is called internalism in epistemology: To characterize something as an instantiation of rational expression necessarily involves a subject’s apprehension of the reasons that justify her belief.

I would like to argue that although this conception may look plausible for simple cases of knowledge, it does not look equally plausible for more complex cases of belief modification. In the history of science, we can find numerous cases of brilliant scientists who offered major contributions to the evolution of their disciplines but in which they had a wrong (or even no) understanding of the reasons underlying their own belief modification. A characteristic example is Isaac Newton and his famous claim that he “frames no hypotheses” (hypotheses non fingo). Contrary to this claim, it is widely admitted that the three laws of mechanics could not have been deduced from phenomena (see Friedman 2001, 26). The number of similar examples is vast. This indicates that practicing science, that is, achieving successful belief modification, does not necessarily presuppose the apprehension of the right reasons for belief modification. In short, being a good scientist does not presuppose being a good philosopher of science. ⁴¹ But if this is correct, and given that we take science as a paradigmatic form of rational expression, it seems that the traditional requirement for an expression of rationality is violated. If scientists, and especially the most successful among them, do not necessarily have access to the right reasons for their own belief modification, and if we take science as a rational enterprise, then their thinking is only in accordance with right reason (κατὰ τὸν ὀρθὸν λόγον) and not due to an appreciation of good reason (μετὰ τοῦ ὀρθοῦ λόγου).

I suggest that we have to reconsider what constitutes a case of genuine expression of rationality. In particular, we have to reconsider the demand that an individual subject possesses an appreciation of good reasons from her internal standpoint. I claim that in some cases, thinking just in accordance with good reason (κατὰ τὸν ὀρθὸν λόγον) can be a sufficient condition for understanding an event as an expression of rationality. In other words, if we can adequately explain normatively an episode, then we are entitled to assert that it is an instantiation of rational expression. But what is a normative explanation in this context? In the history of science, this can be equated with what Imre Lakatos (Lakatos 1978b, 117) described as a rational reconstruction which “produces some characteristic pattern of rational growth of scientific knowledge.” A rational reconstruction of a historical episode makes this episode intelligible by employing one or more epistemic norms which adequately explain the belief modification in this particular case. As in any case of normative explanation, the rational reconstruction adds a further level of understanding the situation. This further understanding consists in that the scientific episode is made intelligible as such, that is, as scientific. In other words, the epistemic norms in question provide the conceptual tools for identifying scientific belief modification as such, that is, as scientific. To explain it (also) in normative terms, the scientific belief modification renders its identification as scientific, that is, as substantially different from other kinds of belief modification (ideological, religious, artistic, etc.).

Now, it may look as if the account I propose endorses a version of epistemic externalism since it does not necessarily include the thinker’s internal standpoint ⁴² in explaining her belief modification. However, the defence of externalism is not my purpose. My aim here is the rejection of methodological individualism rather than epistemic internalism. Most of the contemporary analytic philosophers (and not only them) are prone to accept methodological individualism and reject any kind of group or collective mind as spooky. John Searle’s (1990, 404) following quote is characteristic of this inclination:

I think most philosophers would agree that collective behavior is a genuine phenomenon; the disagreement comes in how to analyze it. One tradition is willing to talk about group minds, the collective unconscious, and so on. I find this talk at best mysterious and at worst incoherent

The ontological rejection of group minds often goes hand in hand with the implicit or explicit admission that rational behavior or thinking can be only attributed to individual agents or thinkers. This entails that normative explanations can only be applied to individual subjects. This is what I call methodological individualism and I would like to cast doubt on this.

Of course, casting doubt on methodological individualism does not entail that I want to make any ontological commitment to group minds or anything similar. In fact, I do not want to make any ontological commitment at all. My point is that we do not have to restrict the application of normative explanations to individual thinkers. The study of science can be illuminating on this point; twentieth-century history and philosophy of science, especially in its post-Kuhnian era, has persuasively shown that the scientific enterprise cannot be made intelligible by focusing on the individual scientists. The unit of practicing science is a scientific community. If this is correct, it follows that the individual thinker’s internal standpoint should not necessarily be considered as part of a normative explanation. Understanding a given episode as an expression of rationality does not always presuppose that any single individual thinker involved is operating from an appreciation of good reason (μετὰ τοῦ ὀρθοῦ λόγου). Rather, it sometimes only suggests that the beliefs of a community have been modified in accordance with good reason (κατὰ τὸν ὀρθὸν λόγον). In other words, we only need to be able to reconstruct the belief modification of a community in terms of an individual thinker, as one cogitating with an appreciation of good reason (μετὰ τοῦ ὀρθοῦ λόγου), in order to characterize an episode as rational. The paradigmatic expression of rationality, that is, the one of an individual thinker operating from an appreciation of good reason, should not be considered as a necessary condition or as criterion for ranking an episode as an expression of rationality. In my view, this both presupposes and entails the rejection of methodological individualism.

5.3. Causes and Conditions of Possibilities

Finally, another aspect of the traditional conception of scientific rationality is that empirical-scientific explanations may be employed only in cases of irrationality or error. As Laudan (1977, 202) says, “the sociology [and every other empirical discipline we may add] of knowledge may step in to explain beliefs if and only if those beliefs cannot be explained in terms of their rational merits”. I don’t think that this is correct. Even in cases of rational belief modifications the empirical-scientific explanations may contribute to the deeper understanding of the belief modification in question. I am not implying the endorsement of any version of the symmetry and impartiality principles here. I merely claim that even in cases of genuine expression of rationality there are many psychological or sociological factors which create the conditions of possibility for rational belief modification. Taking into consideration these factors may enrich the comprehension even of the episodes of scientific change we consider as rational. Contrary to the symmetry and impartiality principles though, the factors in question are understood as conditions of possibility rather than as causes of the belief modification. In short, even the episodes of belief formation which count as successful, that is, as rational, are susceptible to empirical investigation. Contrary to the symmetry principle though, we should acknowledge that for them there is an additional level of understanding to be ascertained, and this consists in the conformity to genuine epistemic norms. Thus, empirical investigation can reveal the conditions of possibility rather than strictly the causes of those particular episodes.

To sum up, in cases of successful scientific belief modification, normative scientific explanations are indispensable in order to make the episode intelligible. However, this does not mean that this is the only kind of explanation we can employ in these cases. Empirical-scientific explanations may also be applied in order to grasp the conditions of possibility of the particular belief modification. On the contrary, in cases of failure, the empirical-scientific explanations can be used to reveal the causes of unsuccessful belief modification. In such instances, the normative content does not have a genuine explanatory role. It is only useful to take into consideration that these episodes fail to conform to genuine epistemic norms.