A Non-Ideal Theory of Knowledge

Abstract

In her article in this volume Linda Martín Alcoff makes the case for a ‘reorientation of the problematic of political epistemology’ that ‘puts power at the centre of analysis’ and considers ‘the connections between those epistemologies that became influential in the mainstream public discourses and broad social injustices’ (Alcoff 2024, p. 67). Her aim is not to replace epistemic concerns with political concerns but to identify some ways in which the epistemic is the political. The result is a conception of political epistemology on which it is analogous to political economy:

The project of political economy was intended to provide an explanatory account at a meta-level, but one that did not eclipse economic theory so much as trace the interrelationship between economic forces and political formations, to note the political constitution of economic roles and forms of agency, and thus to de-naturalize contemporary processes of production and distribution, as well as create a space for imagining alternative configurations and research questions. (Alcoff 2024, p. 69)

On Alcoff’s conception, political epistemology provides an explanatory account that traces the interrelations between ‘epistemics’ (Goldman 1978) and politics, the political constitution of epistemic roles, norms and forms of agency, and denaturalizes, in the sense of historically and socially situating, procedures of knowledge production and distribution. Alcoff pursues this project via a discussion of three twentieth-century thinkers (Horkheimer, Habermas, Foucault) who, she argues, pursued this form of political epistemology, albeit in different ways, and to different ends.

In this paper I pursue a similar project, but within a different tradition, one that grows out of naturalized epistemology. My aim is to sketch a picture of knowledge that does justice to two strands in naturalized epistemology. The first strand, represented historically by W. V. O. Quine and more recently by Hilary Kornblith, emphasizes the importance of psychology. The second strand, represented by a diverse range of thinkers, emphasizes the importance of history, sociology, and the ‘human sciences’ (Foucault 1980; Rorty 1981; Collins 1985; Bloor 1991; Burke 2000, 2012; Canguilhem 2000; Kusch 2002). I argue that, to do justice to the first strand, we need to recognize that some forms of knowledge, such as simple observational knowledge, are best understood through the lens of a psychologistic naturalized epistemology. But, equally, to do justice to the second strand, we need to recognize that other forms of knowledge, such as scientific knowledge, are best understood through a more sociological lens. The result is a form of pluralism that does justice to both strands of naturalized epistemology and is, hopefully, a rapprochement between two different epistemological traditions, one that has its home in ‘continental philosophy’, the other in ‘analytic philosophy’.

I

Naturalized Epistemology Beyond Psychology. What is naturalized epistemology? Kornblith (1994) characterizes naturalized epistemology in terms of the answers it gives to two questions:

• (1) How should we form beliefs?

• (2) How do we form beliefs?

The first question is normative. An answer to it provides a standard against which we can measure beliefs and processes of belief formation. We can say that beliefs that are formed in the right way are justified and that beliefs that are both true and justified (typically) count as knowledge. The second question is descriptive. An answer to it provides insight into how humans end up with their beliefs. The traditional epistemologist views these two questions as distinct, in the sense that an answer to the descriptive question does not constrain the answer to the normative question. Epistemology, as a normative domain, is entirely autonomous from descriptive psychology. The naturalized epistemologist views these questions as related, in the sense that an answer to the descriptive question, in some way or other, constrains the answer to the normative question.

Different forms of naturalized epistemology take different positions on the nature of these constraints. The strongest version of naturalized epistemology, Quine’s, is usually read as replacing the first question with the second. As Quine put it, ‘Epistemology, or something like it, simply falls into place as a chapter of psychology and hence of natural science’ (Quine 1969, pp. 82–3). Many have objected to Quine’s reduction of epistemology to psychology, most influentially Jaegwon Kim (1988). But weaker versions of naturalized epistemology do not simply reduce epistemology to psychology. Kornblith’s view is, put programmatically, that the ways in which we should form beliefs must be a subset of the ways in which we in fact form beliefs. Some ways of forming beliefs are good, from the epistemic point of view, and a theory of justification identifies the good ways and distinguishes them from the bad. Normative epistemology cannot ‘float free’ of descriptive psychology. But it does not reduce to it either.

We can distinguish between two versions of this weaker view, Kornblith’s (see Kornblith 1999, 2002) and Alvin Goldman’s (see Goldman 1979, 1986). Kornblith tells an evolutionary story, on which our inbuilt cognitive processes are—typically and within certain limits—reliable, and so the beliefs we form via our inbuilt cognitive processes are justified. Kornblith defends a version of reliabilism (roughly: a belief is justified if and only if it is produced by a reliable process or procedure). Goldman also defends a reliabilist theory of justification, but he arrives at it via philosophical reflection on the concept of justification, and then shows how his theory can be ‘grounded’ in empirical psychology. Either way, the result is a psychologistic epistemology that still has normative aspirations (Kornblith 2020).

While Kornblith’s characterization of naturalized epistemology is helpful, it is limited in several respects. One respect in which it is limited is that it is too individualistic. It focuses on the individual and their processes of belief formation, rather than on groups, or the social institutions, such as science, which serve knowledge-producing functions. We can characterize more social approaches to naturalized epistemology in terms of two further questions:

• (1) How should knowledge-producing social institutions be structured, and what should their procedures for certifying something as knowledge look like?

• (2) How are knowledge-producing social institutions structured, and what do their procedures for certifying something as knowledge look like?

The first question is, again, normative. An answer to it provides a standard against which we can measure existing knowledge-producing social institutions and criteria of certification. The second question is descriptive. An answer to it provides an account of how actual knowledge-producing social institutions work, and of the criteria they use to assess their outputs.

A traditionalist form of social epistemology will view the normative question as entirely independent of the descriptive question. On this view, social epistemology is entirely autonomous from descriptive sociology. It is unclear what motivation there might be for defending this view once one has accepted some form of psychologistic epistemology. This is noteworthy, given that many social epistemologists do not view epistemology as entirely autonomous from psychology, yet few mainstream social epistemologists pay anything like as much attention to sociology as they do to psychology. Indeed, their attitude is often downright hostile (Kusch 2010).

The strongest version of social naturalized epistemology would be a view that is analogous to Quine’s: social epistemology reduces to sociology. On this view, there would be no difference between something being knowledge and it being certified as knowledge by some social institution or procedure. More generally, there would be no space for normative assessment of social institutions or of the epistemic criteria utilized within an epistemic community or discourse. A form of epistemic nihilism results. It is unclear whether anyone really endorses this view. Many (for instance, Boghossian 2006) view sociologists of scientific knowledge such as David Bloor and Barry Barnes as defending it, and some might attribute it to Foucault. But these attributions are contentious (see Kusch 2020 on the sociologists, and Alcoff 2013 on Foucault).

You can have versions of social naturalized epistemology that do not simply reduce social epistemology to sociology. On these views, while social epistemology does not reduce to sociology, it cannot ‘float free’ from it either. That is, some social institutions are structured in such a way that the procedures they employ, including the procedures they use for certifying something as knowledge, are good, from the epistemic point of view. A ‘social theory of justification’ identifies the good institutions and procedures. But it must be grounded in actual practices, not imposed on them ‘from the outside’.

It is helpful to distinguish between two versions of this view. On the first, we look at examples of good organizational structure within an institution, or criteria and procedures that have proven successful in particular fields, and derive norms from these examples of good practice. Helen Longino’s ‘critical contextual empiricism’ is an example of this (see Longino 1990, 2002), as is the feminist empiricist tradition in philosophy of science of which she is a part (see Nelson 1990; Anderson 1995; Goldenberg 2015). On the second version, we arrive at a set of criteria and procedures for assessing knowledge-producing institutions via philosophical reflection, and then try to show that these criteria and procedures are operative within actual institutions, and so represent a sort of ‘gold standard’ to which these institutions can aspire. Philip Kitcher’s work on the philosophy of science is a good example of this (see Kitcher 2001, 2002, 2011). Either way, the result is a form of social epistemology that takes sociology seriously but still has normative aspirations.

While this improves on Kornblith’s characterization of naturalized epistemology, it takes a very narrow view of social institutions, procedures of knowledge production, and the environment in which these institutions and procedures operate. What one might (albeit not unproblematically) view as ‘external’ factors, such as the ways in which these institutions might serve to perpetuate existing injustices, patterns of exclusion and forms of oppression, play no role in the story. This is the problem that Charles Mills (2007) highlighted with analytic social epistemology:

[T]he implications of systemic social oppression for his project are not addressed. The picture of ‘society’ he [Mills is talking about Goldman] is working with is one that—with perhaps a few unfortunate exceptions—is inclusive and harmonious. Thus his account offers the equivalent in social epistemology of the mainstream theorizing in political science that frames American sexism and racism as ‘anomalies’: US political culture is conceptualized as essentially egalitarian and inclusive, with the long actual history of systemic gender and racial subordination being relegated to the status of a minor ‘deviation’ from the norm … Obviously, such a starting point crucially handicaps any realistic social epistemology since in effect it turns things upside-down. Sexism and racism, patriarchy and white supremacy, have not been the exception but the norm. (Mills 2007, p. 17)

If—as I have been suggesting—naturalized epistemology places descriptive constraints on normative theorizing, it is hard to see how one could meaningfully pursue normative naturalized epistemology while working with the sort of idealized picture of society, whether American or otherwise, that Mills is criticizing. The driving impetus behind naturalized epistemology is the idea that the procedures, both psychological and social, by which knowledge is produced are important objects of epistemological theorizing. Naturalized epistemology does not simply reduce to the description of these procedures (that was Quine’s view); but it is centrally concerned with them. It makes little sense to be concerned with the procedures yet not concerned with the ways in which they might serve to reinforce existing social hierarchies or to create new ones.

I have sketched a way of thinking about naturalized epistemology that combines three elements: psychology, sociology, and an emphasis on the social and political functions of procedures of knowledge production. You might think that the resulting picture is disjointed and messy (see §ii). We can bring some theoretical unity to it by viewing naturalized epistemology as a form of non-ideal epistemology. Inspired by Mills, McKenna (2023) characterizes non-ideal epistemology as a form of naturalized normative epistemology that recognizes four ‘dimensions of non-ideality’: (i) human cognitive limitations, such as our susceptibility to various forms of cognitive bias; (ii) limitations in our ‘information environment’, such as the prevalence of falsehood and misinformation; (iii) the role that power dynamics, such as differentials in social power and status, play in shaping social interactions; and (iv) the prevalence of systematic social injustices and oppression in human societies. This provides us with a relatively unified framework for pursuing a normative form of naturalized epistemology that is constrained by psychology and sociology, and also foregrounds rather than ignores the various respects in which human societies differ from idealized pictures of society.

II

Messy Epistemology. McKenna (2023) developed some central aspects of non-ideal social epistemology. But I did not say much about the central concepts of traditional epistemology: justification, knowledge, rationality. In the rest of this paper I sketch the outlines of a non-ideal theory of knowledge. Along the way I will comment on points of contact and disagreement between this theory and Alcoff’s own writings on knowledge, in particular her piece in this volume, her discussion of Foucault’s epistemology in Alcoff (2013), and her excellent book Real Knowing (Alcoff 1996). I want to emphasize that this is just an outline: more needs to be said about the areas of the theory I touch on, and much more needs to be said about the ways in which this theory might inform, and be informed by, the sort of political epistemology that Alcoff advocates.

Let me start with a methodological issue: all theorizing relies on paradigms—cases and examples around which one builds one’s theory. Theories of knowledge therefore rely on paradigms, cases where it is taken as obvious that knowledge is present, and that are presumed to be representative of knowledge ‘in general’. The choice of paradigms matters because it shapes the theory that results. Within analytic epistemology, the paradigms of choice have tended to have two salient features. First, they feature either lone individuals or very minimal social interaction—Smith sees a sheep, Jones disagrees with Smith about their respective shares of the bill, and so on. Second, they are simple—Smith knows that there is a cup of coffee in front of him because he can see it, Jones knows that his hat is on his desk because he remembers leaving it there, and so on. (These features are related. One respect in which the paradigms are simple is that they do not involve much social interaction).

While anyone of a non-sceptical persuasion will agree that these are clear cases of knowledge, it is far from obvious that they are representative of knowledge in general. Here is Alcoff making this point nicely:

Epistemology has too often operated as a private conversation presuming to sit in judgment on the whole expanse of human knowledge yet separating itself from actually existing knowers. Epistemologists pare down the complex variety of real knowing practices to a few supposedly paradigmatic cases (‘Jones owns a Ford’) whose simplicity is purported to allow a close examination of epistemic justification. As a result, philosophers continue to struggle over finely tuned analyses of simple inferences, observational beliefs, and memory while other fields have moved light years ahead in developing more complex and realistic accounts of socially and historically situated belief-formation. (Alcoff 1996, p. 1)

What are we epistemologists to do? Alcoff continues:

[E]pistemologists must begin to reflect on and address recent developments in the analysis of how knowledge and meaning are actually produced, how science truly proceeds, and how certain bodies of theory get to count as knowledge. No longer can we envision an individual knower grappling with nature’s secrets, or a text as having a single, decisive interpretation. (Alcoff 1996, p. 1)

Because the production and certification of knowledge is invariably messy, the resulting epistemology will be messy too. We need a messy epistemology. The naturalized epistemological tradition, at least as I have reconstructed it, is one way of making a mess. For the naturalized epistemologist, a theory of knowledge must be based on a psychologically and sociologically informed picture of how knowledge is produced and of how it is certified as such. On the psychology side, this will require getting into the details of how humans think about the world around them, including the influence of the external environment (including other humans) on cognition. On the sociology side, this will require getting into the details of how what we call knowledge is certified as such, which means the political dimensions of knowledge cannot be ignored.

While it is a mistake to fixate on simple inferential, memory, or observational knowledge, I also think it would be a mistake to deny the importance, or even the existence, of these paradigms (more guardedly, of actual cases that conform, at least in their essentials, to the paradigms). Alcoff and I may differ on this point. Where Alcoff suggests that even these cases are not as simple as they might appear, I instead want to say that the problem is more with viewing these cases as representative of knowledge in general, and as a place from which we can construct a unified theory of knowledge.

There clearly is a sense in which any individual knower is dependent on others: we cannot grapple with nature’s secrets by ourselves. We are dependent on others for the concepts we use to frame our thoughts, and for the cognitive scaffolding that allows us to think and do whatever we think and do, including the information we have about the world beyond our immediate experience. But I see no good reason to deny that, with this support in place, individuals can gain certain kinds of knowledge on their own. An adult human may need a lot of help to get to the stage where they can look at a cup of coffee and know they are looking at a cup of coffee. But once they have reached this stage, it is unclear why we need to refer to anyone other than the individual in making sense of how they know the cup is in front of them.

Alcoff pushes back here. In her piece in this volume, she quotes a passage from Max Horkheimer:

[H]uman action unconsciously determines not only the subjective side of perception [through, for instance, production of certain kinds of perceptual and measuring tools] but in larger degree the object as well. The sensible world which a member of industrial society sees about him every day bears the marks of deliberate work: tenement houses, factories, cotton, cattle for slaughter, men, and in addition, not only objects such as subway trains, delivery trucks, autos, and airplanes, but the movements in the course of which they are perceived. The distinction within this complex totality between what belongs to unconscious nature and what to the action of man in society cannot be drawn in concrete detail. (Horkheimer 1975, p. 201; quoted in Alcoff 2024, p. 71)

For Horkheimer, human action shapes the object of perception as well as the ‘subjective side’ of perception (the concepts we use, the assumptions and prejudices we bring to our interpretations of what we see). The objects of human perception are, we are told, the products of deliberate work and, as a result, the distinction between what belongs to human action and what belongs to unconscious nature cannot be maintained.

There is something to this. But Horkheimer glosses over two important points. First, as far as the knowing subject is concerned, it is important to distinguish between the somewhat minimal role that human action—and more generically social factors—plays in any sort of perception and the very extensive role that social factors play in, for example, the production of scientific knowledge. To treat these things as equivalent because there is a sense in which any form of perception is shaped by social factors is to ignore the crucial differences between them.

Second, as far as the object of knowledge is concerned, valid distinctions admit of unclear cases. Indeed, most of the examples Horkheimer cites are not unclear—houses, factories, trains, trucks, cars and planes are all direct products of human action. But other things are products of human action in a more attenuated sense (a beach, say), and some things are not products of human action at all (the sun, the moon). Neither the fact that it is unclear whether certain things are the product of human action nor the fact that some things clearly are the product of human action shows that there is no distinction to be drawn between what belongs to human action and what does not.

Even if there is a ‘thin’ sense in which the objects of human knowledge are always shaped by human action (we can only know what we have interacted with, however indirectly), this is very different from the far richer sense in which, say, the objects of social scientific knowledge are shaped by human action. The social sciences traffic in kinds of knowledge that are social in ways that go far beyond the ways in which, say, simple observational knowledge is social. Most obviously, the social sciences deal with what Ian Hacking (1995) calls ‘human kinds’ (roughly: categories of people, like asylum seekers or immigrants). The classification and description of human kinds leads to ‘feedback effects’ because the objects of classification can respond to being classified in ways that cups of coffee simply cannot. To capture this, we need a richer sense in which certain kinds of knowledge are social, a sense that excludes simple inferential, memory, or observational knowledge.

The conclusion I draw from this is not that we should go back to constructing a theory of knowledge around the paradigms of analytic epistemology. We need to consider paradigms of knowledge that are social in the richer senses in which scientific knowledge is social. We need a broader set of paradigms. I therefore thoroughly agree with Alcoff when she writes, in an illuminating discussion of the value of Foucault’s case study-driven approach to epistemology, that:

The dominant view in the Anglo-American tradition remains that the diversity of knowledge forms can be distilled into a set of essential features, and thus epistemology can justifiably pursue a single, unified account [of knowledge]. The unexamined possibility here is that there are multiple forms of knowledge needing multiple analyses, in which case Foucault’s analysis might have a restricted domain and yet still contribute to an overall theory of knowledge. I would argue that a commitment to a universal treatment of knowledge is not a necessary requirement for an epistemology (Alcoff 2013, p. 223).

In the next section I will sketch my own version of the view that there are multiple forms of knowledge requiring multiple analyses. The result is a theory of knowledge that is messy but does justice to the complexity of the underlying phenomena.

III

A Messy Theory of Knowledge. In his book Knowledge and Its Place in Nature, Kornblith says this:

The idea that philosophy consists in, or, at a minimum, must begin with an understanding and investigation of our concepts is, I believe, both natural and very attractive. It is also, I believe, deeply mistaken. On my view, the subject matter of ethics is the right and the good, not our concepts of them. The subject matter of philosophy of mind is the mind itself, not our concept of it. And the subject matter of epistemology is knowledge itself, not our concept of knowledge (Kornblith 2002, p. 2).

For Kornblith, you go about studying knowledge in the same way that you go about studying any natural phenomenon: identify paradigm instances of the phenomenon, and then try to identify their unifying features. He focuses on paradigms of knowledge provided by cognitive ethology, a branch of biology concerned with animal behaviour. He argues that, to explain how animals manage to successfully navigate their environment (get food, procreate, survive), we need to attribute to them a capacity for reliably producing true beliefs—a capacity for knowledge.

Many have voiced doubts about this story (see, for example, Kusch 2005; Weinberg 2006; Pernu 2009; Brown 2012; Horvath 2016). To my mind, though, the main mistake Kornblith makes is that he focuses on a narrow set of paradigms: simple observational knowledge in non-human animals. As a result, his theory of knowledge is based on a limited set of samples. Even if he is correct about what unifies his paradigms, he is hardly justified in drawing general conclusions about knowledge. It is instructive to consider how different the results of Kornblith’s inquiry would be if we started with a different set of paradigms (see Kusch and McKenna forthcoming). I will consider two different paradigms, cognitive neuroscience and the sociology of scientific knowledge (ssk), though one could include many others, including Foucault’s own case studies in the human sciences (see Alcoff 2013).

Let us start with cognitive neuroscience. Andreas Stephens (2016) argues that what cognitive neuroscientists call ‘semantic memory’ roughly corresponds to what epistemologists call propositional knowledge. Semantic memory includes knowledge of people, knowledge of places, and knowledge of the meanings of objects and words (Gazzaniga, Ivry and Mangun 2002). It is different from episodic memory; knowing that Paris is the capital of France is semantic memory, whereas remembering visiting Paris, or remembering being taught that Paris is the capital of France, is episodic memory (Ward 2010, p. 186). Stephens suggests that semantic memory may be a functional but—within certain limits—not entirely accurate way of producing or retrieving information about the world. It provides us with a model of the world that is skewed by ubiquitous cognitive biases but, for the most part anyway, enables us to function well in the environments in which we find ourselves (Stephens 2016, p. 897). To my mind, the reasons Stephens cites aren’t entirely persuasive, but Adam Bricker has recently provided a good deal more empirical backing for the view that knowledge, understood as a cognitive neuroscientific kind, isn’t factive (see Bricker 2018, 2022).

Turning to ssk, one finds a very different approach to thinking about knowledge. David Bloor (an influential author in the field) puts the central idea like this:

The sociologist is concerned with knowledge, including scientific knowledge, purely as a natural phenomenon. … [K]nowledge for the sociologist is whatever people take to be knowledge. It consists of those beliefs which people confidently hold to and live by. … Of course, knowledge must be distinguished from mere belief. This can be done by reserving the word ‘knowledge’ for what is collectively endorsed, leaving the individual and idiosyncratic to count as mere belief. (Bloor 1991, p. 5)

ssk practitioners like Bloor want to explain central episodes in the history of science, both the formation of scientific consensus, and so of ‘scientific knowledge’, and the persistence of dissensus. To this end, they develop theories that identify and explain ‘the regularities and general principles or processes’ that concern the production, preservation and dissemination of true and false beliefs (Bloor 1991, p. 7). This involves placing knowledge in its social and cultural context. As Barry Barnes (another influential author in the field) puts it:

Knowledge is not produced by passively perceiving individuals, but by interacting social groups engaged in particular activities. And it is evaluated communally and not by isolated individual judgments. Its generation cannot be understood in terms of psychology, but must be accounted for by reference to the social and cultural context in which it arises. Its maintenance is not just a matter of how it relates to reality, but also of how it relates to the objectives and interests a society possesses by virtue of its historical development. (Barnes 1977, p. 2).

When Barnes says that the generation of knowledge ‘cannot be understood in terms of psychology’ and so ‘must be accounted for by reference to the social and cultural context in which it arises’ he is talking about the paradigms of knowledge of interest to ssk. If we want to understand scientific knowledge, we need to place it in its social and cultural context, But ssk practitioners acknowledge that some forms of knowledge are amenable to ‘purely psychological’ explanations, and don’t need to be explained in terms of ‘traditions, cultures, conventions and interests’ (Bloor 1992, pp. 135–7). A rat can know where the food is in their maze without any sort of social and cultural context. To the extent that humans know some things in the way in which rats know things, the same goes for certain kinds of human knowledge.

My point is not that the accounts of knowledge proposed by cognitive neuroscientists or ssk practitioners should replace Kornblith’s cognitive ethology-inspired account. My point is that viewing knowledge as a natural phenomenon, amenable to empirical investigation, requires acknowledging the diverse range of ways in which knowledge is understood within different disciplines (see Kusch and McKenna forthcoming). An analytically inclined epistemologist will likely want to try and reduce this diversity, either by arguing that we really have several different phenomena, rather than a diverse range of manifestations of a single phenomenon, or by arguing that what cognitive neuroscientists or ssk practitioners call knowledge isn’t knowledge, at least in the sense in which epistemologists mean it.

Let me finish by saying two things in response to this. First, these objections are hard to motivate within a naturalized approach to epistemology. On the naturalized approach, knowledge is viewed as a natural (which might mean social) phenomenon. If we study knowledge as a natural phenomenon, we need to be open to the possibility that the result of our study will be that we are dealing with a heterogeneous phenomenon. Unity is not to be assumed at the outset but must rather—if there is any unity at all—be the outcome of empirical investigation into knowledge as a natural phenomenon. While we can identify some unifying features (we are dealing with a mental state, one that differs from mere belief), we should not prejudge the results of our inquiry by expecting to find a simple state underlying the diverse range of ‘knowledge practices’. Nor should we expect to find that the object(s) of our study have all the properties that epistemologists standardly view knowledge as having. While knowledge in the epistemologist’s sense is perhaps different from knowledge in the cognitive neuroscientist’s sense or in the ssk practitioner’s sense, this is not itself an objection, at least from the standpoint of naturalized epistemology.

Second, one might object that this is why we need more than naturalized epistemology. Knowledge, one might say, is a normative notion like goodness or justice. Just as it would be a mistake to investigate the nature of goodness or justice simply by investigating how particular groups of people (lawyers, politicians) talk about goodness or justice, it is a mistake to investigate the nature of knowledge simply by investigating how particular groups of people (cognitive neuroscientists, sociologists of knowledge) talk about knowledge.

If this is a good objection, it is also a good objection to investigating the nature of knowledge by investigating how analytic epistemologists talk about knowledge (Hazlett 2010). Setting that aside, I agree that we need something more than empirical descriptions of the psychological or social procedures by which beliefs are formed and consensuses are arrived at. We need normative criteria against which to assess these procedures and their outputs (recall §i). But rather than impose these criteria ‘from the outside’, we need to find a way of extracting these criteria from existing practices. This is something that Helen Longino has tried to do in her work on scientific knowledge:

The (justificatory) activities of knowledge construction as distinct from belief formation (a generative process) are the activities of individuals in interaction, of individuals in certain relations (of criticism and response) with others. The cognitive abilities of individuals are both necessary and sufficient for the generation of ideas, but idea or belief generation is not the same as knowledge production, which involves processes of validation as well as generation. The account of cognitive agency as involving interdependence means that individuals know to the extent they interact critically with others in cognitive communities. (Longino 2002, pp. 122–3)

For Longino, scientific knowledge is constructed from ‘critical interactions’ between individuals in ‘cognitive communities’. Her idea is not that any old interaction produces knowledge. There is, for Longino, a difference between something being knowledge and it being viewed as such by the participants in an interaction. Interactions need to be ‘epistemologically effective’ to produce knowledge, and Longino ventures some suggestions about which sorts of interactions are epistemologically effective (see Longino 2002, pp. 129–31).

Rather than look at these suggestions, I want to focus on the broader shape of Longino’s account. It is an example of what Alcoff (2013) calls a ‘historicist’ theory of justification. Historicist theories contrast with ‘proceduralist’ theories: proceduralists seek ‘procedures of justification that would transcend all historical and cultural contingencies’ whereas historicists hold that, even if there are universal procedures, ‘the judgment of adequacy, consistency, relevance, and the appropriate forms of testability is dependent on the qualitative determinations specific to historical context’ (Alcoff 2013, pp. 212–13).

To see what this distinction amounts to, consider reliabilist epistemology, and in particular how Philip Kitcher applies it to evaluating scientific institutions and practices (see, for example, Kitcher 2002). His idea is that we can evaluate scientific institutions and practices using a simple procedure: do they typically produce true (or accurate) scientific beliefs and theories? Kitcher views this reliabilist procedure as providing a normative criterion that ‘stands outside’ scientific practices. In contrast, the historicist would say that applying this procedure requires ‘qualitative and comparative determinations of relevance, scope and adequacy’ (Alcoff 2013, p. 213). In the case of the reliabilist procedure, applying it requires, inter alia, determining the degree of reliability that we are looking for (Williams 2015), as well as answering the complicated question of how to determine the reliability of procedures for producing scientific beliefs and theories, given that any particular token procedure is a token of many different types (see Conee and Feldman 1998 on the ‘generality problem’), and any type may be reliable in some environments but not in others (see Brandom 1998 on the ‘reference class problem’).

Why should we prefer a historicist theory such as Longino’s to a proceduralist theory such as reliabilism? Here is Longino:

Reliabilism is of little use in the scientific contexts that involve the adoption of frameworks of inquiry, background assumptions, experimental strategies, instruments and the like, where precisely what we are determining is that these are reliable. Of course, in the end we want processes and practices that are reliable, but our judgment that they are reliable is secured through the material and discursive interactions discussed earlier. (Longino 2002, p. 164)

There are two different points Longino might be making here. The first, which I do not think tells against the proceduralist/reliabilist, is that the problem with the reliabilist procedure, or indeed with any supposedly universal procedure, is that it does not provide adequate guidance. The reliabilist tells us to look for procedures that reliably produce true beliefs. But how are we to do that? The reliabilist procedure provides little guidance on this point. This only tells against the proceduralist if it is the job of a theory of justification to provide this sort of guidance, and it is unclear that this is the job of a theory of justification, especially a reliabilist theory of justification (Hughes 2021). Reliabilism is a theory about what justifies our procedures, not a theory about how to choose a procedure in the first place.

The second point, which I do think tells against the proceduralist, is that any supposedly universal procedure needs to be operationalized. Operationalizing the reliabilist procedure in a particular context will, as above, involve determining the appropriate degree of reliability in that context, as well as resolving the generality and reference class problems. Moreover, given that we are likely going to need further procedures (for example, procedures for determining empirical adequacy, predictive power, testability, and so on), we also need to decide how to balance these procedures against each other, given that they might point in different directions. No matter how these decisions get made in any particular context, they are going to shape the direction and outcome of inquiries in that context. If the outcomes of our inquiries are to be justified, they need to be justified given the way in which our procedures have been operationalized; it is the procedures, as they are operationalized in a particular context, that confer justification in that context. But this means we need more than just the universal procedures. We need to consider the ways in which these procedures have been operationalized in particular contexts.

The upshot is that, while we can—if we squint a little—abstract general procedures from a survey of epistemic practices, this does not vindicate the proceduralist theory, because no scientific belief or theory is justified simply as a result of the general procedure. It is justified because of the judgements and decisions that are made about how to operationalize that procedure in a particular context. It is these judgements and decisions that, as Alcoff (2013, p. 213) puts it, ‘comprise the variable, historically situated form of rationality in a given place and time’.

References