1

What Is Non-Ideal Epistemology?

What is epistemology? If you have taken an epistemology class, you were likely
told that epistemology is the ‘theory of knowledge’. (If you have taught an
epistemology class, you perhaps told your students it is the theory of knowledge.)
So understood, epistemology is concerned with the nature of knowledge, which
for many epistemologists really means the conditions under which someone
knows something (‘S knows that p if and only if . . .’).
If your epistemology class (I am assuming you had an epistemology class) was
about this, you might have had the nagging feeling that there was something odd
about the whole thing. You were probably asked to consider what you would say
about imagined scenarios where someone believed something to be true on what
seemed to them like good evidence but, unbeknownst to them, something odd was
going on. Perhaps our imagined individual—let us call them Smith—was in a field.
Smith, you were told, sees something that looks a lot like a sheep. Surely, you
thought, that means Smith knows there is a sheep in the field. But then your
teacher told you the twist. The thing Smith is looking at is really a rock that looks
like a sheep but behind the rock is an actual sheep obscured from view. What now?
Does Smith know that there is a sheep in the field?
It may be that you figured out how epistemologists usually answered these
questions. Maybe you even got good at concocting your own examples to test the
plausibility of this-or-that theory of knowledge. On the other hand, it may be that
you decided the whole thing was not for you. Either way, it was probably clear to
you that, whatever epistemology is, it is an extremely abstract and theoretical
enterprise. You might have asked what relevance it has to your life. You might
have wondered whether it has any applications to ‘real world’ issues and problems.
If you were fortunate enough to have taken your introductory epistemology
class recently, your teacher might have said that things are different now.
Epistemologists do not just talk about lone individuals (or people called Smith).
They also do not just talk about the theory of knowledge. One exciting develop-
ment is that they now talk about social interactions. You might have been asked
what you think about examples featuring two individuals—let us call them Smith
and Jones—who tell each other things. Perhaps Smith tells Jones that the train
leaves in ten minutes. Is Jones justified in believing what Smith has told him
without checking whether Smith is trustworthy first? Maybe you were also asked
to consider scenarios where Smith and Jones disagree with each other. Perhaps
Smith says that the restaurant bill works out at £20 each whereas Jones says it is £21.

Non-Ideal Epistemology. Robin McKenna, Oxford University Press. © Robin McKenna 2023.
DOI: 10.1093/oso/9780192888822.003.0001
2 - 

Should Smith and Jones redo their sums until they come to an agreement, or should
they each conclude that the other has made an arithmetical mistake?
It may be that these developments satisfied you. You might have said that at
least you can see why these questions might matter. We have all been in situations
where we had to decide whether to believe what someone has told us and are not
able to check whether they are trustworthy or not. We have also all had disagree-
ments over the bill. Alternatively, it may be that these developments did not satisfy
you at all. You might have objected that these sorts of scenarios are too simplified
to be of any practical relevance. You might have got frustrated about the lack of
detail. Why would Jones approach a stranger to ask when the train leaves when he
could just use an app on his phone? How does Smith know that Jones is not just
trying to trick him into paying more than he needs to? Why can’t they just
calculate their shares of the bill using an app on their phones?
Let us continue our story. Imagine that, despite your misgivings, you ended up
doing more advanced courses in epistemology. Perhaps you even ended up doing
a PhD on the subject and teach it yourself. If nothing else, this gave you the
vocabulary to articulate the misgivings you had always had about epistemology.
Your first exposure to epistemology was to what epistemologists now often
derisively refer to as ‘Gettierology’, which is a central topic in ‘traditional epis-
temology’. Gettierology—so called because it was sparked by Edmund Gettier’s
1963 paper ‘Is Justified True Belief Knowledge?’—is the project of trying to find
necessary and sufficient conditions for a subject S to know some proposition p. It
turns out that, at least nowadays, many professional epistemologists do not like
Gettierology either.
You were then exposed to what is called ‘social epistemology’, which considers
the epistemological implications of social interaction (Goldman and O’Connor
2021). However, as you now recognize, the problem with social epistemology is
that you can consider the epistemological implications of social interaction while
working with highly idealized pictures of what social interaction, and the creatures
who do the interacting, are like. Any dissatisfaction with your initial exposure to
social epistemology was really with these highly idealized models. Your complaint
was that these idealized models do not tell us much about social interactions and
so, despite its promise, social epistemology is of far less ‘real world’ relevance than
it might initially appear to be.
Before this veers (too far) into autobiography, let me get to the point. This book
is about the idealized models of human beings and the social interactions between
them favoured by many social epistemologists. My central aim is to argue that
serious problems can result from working with these idealized models. This book
is therefore at least in part a defence of the bemused reaction of our imagined
student (ok, me) to their first exposures to epistemology.
I say ‘in part’ because it is not a complete defence. Frustratingly for our
imagined student, I am not going to come down entirely on their side. My basic
   -  ? 3

claim is that what we can call ‘ideal epistemology’—the kind of epistemology that
works with these idealized models—goes wrong in that we sometimes need to
work with less idealized models of human beings and of the social interactions
between them. There is, therefore, a need for what we can call ‘non-ideal epis-
temology’. Where ideal epistemology works with idealized models of humans and
social interaction, non-ideal epistemology works with less idealized, more realistic
models.
But let me be clear. I do not claim that ideal epistemology is fundamentally
misguided, or that it is always a mistake to work with idealized models of
epistemic agents and the interactions between them. Idealization can be a valuable
tool, and it would be a grave mistake to object to idealization per se, both in
general and in epistemology. Two sorts of problems can, however, arise when you
engage in idealization. The first is that you might end up ignoring phenomena that
are of real interest because you work at a level of idealization from which they are
rendered invisible. The second problem arises when you forget that idealization is
a tool or—still worse—when you forget that you are engaging in idealization in the
first place.

1.1 Three Aims

Let me get into the specifics. In this book, I set out to achieve three things. First,
I demonstrate the importance of distinguishing between ideal and non-ideal
epistemology. The main obstacle to recognizing the importance of this distinction
is the existing distinction between traditional and social epistemology. What, you
might ask, does the distinction between ideal and non-ideal epistemology give us
that we do not already get from the distinction between traditional and social
epistemology? My answer is that it is possible—indeed common—to pursue social
epistemology in a way that is highly idealized. In the following chapters I argue
that problems can result when we pursue social epistemology in an overly
idealized fashion.
My claim is not that the distinction between ideal and non-ideal epistemology
should replace the distinction between traditional and social epistemology. My
claim is that we need to complicate this distinction by recognizing the differences
between ideal and non-ideal approaches to social epistemology. More generally,
my contention is that we should pause to consider what ties together the various
approaches adopted, and issues considered, within the umbrella of ‘social epis-
temology’. What does the debate between reductionists and anti-reductionists in
the epistemology of testimony have in common with the literature on epistemic
injustice and oppression? What does the debate between conciliationism and the
steadfast view in the epistemology of disagreement have in common with the
(epistemological) literature on group polarization? To say that they are all
4 - 

concerned with the epistemological implications of social interaction is fine, but it

ignores the many differences. These differences have to do with the fact that some
of these debates are predicated on certain idealizations while others are not.
Second, I show that, at least with respect to some issues and problems of (social)
epistemological interest, the non-ideal epistemologist’s approach is preferable to
the ideal epistemologist’s approach. I have already said this earlier, but it bears
emphasis, so I will say it again. I do not argue that non-ideal epistemology is
‘better’ than ideal epistemology (whatever that might mean), or even that it is
usually the right approach to adopt. My claim is just that it is sometimes the right
approach to adopt.
This claim is, I think, modest. I do not view this book as arguing for a
particularly radical claim. However, there is one sense in which it is a little less
modest. I suspect—though have no evidence to back this up—that many epistem-
ologists are tempted by the view that ideal and non-ideal epistemology are about
different things and so cannot come into conflict in the way I think they can. I also
suspect that this attitude is common when it comes to feminist epistemology (for
more on the difference between non-ideal and feminist epistemology, see §1.3).
Feminist epistemologists have identified some important questions and brought
new issues to the foreground of epistemology (for example, epistemic injustice).
But—you might think—the key contribution of feminist epistemology lies in the
ways in which it expands the field of epistemology, not in the ways in which it
critiques central assumptions of traditional (and much social) epistemology.
Whether this attitude is common or not, I think it is misplaced (Ashton and
McKenna 2020; Dotson 2014, 2018; Toole 2019, 2022). In the chapters that follow,
I identify some places where ideal and non-ideal epistemology can—indeed, do—
come into conflict. I therefore do not just intend to show that ideal epistemology
must be supplemented by non-ideal epistemology. I intend to show that, at least in
certain cases, it must be replaced by non-ideal epistemology.
Third, I make progress in several debates in social epistemology by adopting a
non-ideal approach. In the chapters that follow, I defend the following claims, all
of which I take to be characteristic of non-ideal epistemology (I say more about
what ties these claims together in §1.2):

(1) Solving the problem of public ignorance about consequential political and
scientific issues like global warming requires creating a better epistemic
environment and better social institutions (Chapter 3, Chapter 4).
(2) An ‘epistemic’ form of paternalism is (sometimes) justified (Chapter 4,
Chapter 5).
(3) We often should not strive to be intellectually autonomous (Chapter 5).
(4) Some (e.g. John Stuart Mill) think we should all engage with challenges to
our views. But, contra Mill, some of us are under no obligation to engage
with (certain) challenges to our views. Indeed, sometimes we can dismiss
   -  ? 5

challenges without engaging at all. More generally, our obligations as

inquirers depend on and vary with aspects of our social situation such as
our social roles and identities (Chapter 6, Chapter 7).
(5) The core idea behind ‘responsibilist’ virtue epistemology is that epistem-
ically responsible agency involves manifesting the intellectual virtues and
avoiding intellectual vice. But making good on this idea requires a con-
textualized and socialized conception of epistemic agency and responsibil-
ity (Chapter 7).
(6) If we take the empirical literature on political cognition seriously, we are
pushed towards a form of scepticism about whether our beliefs about
political and politically relevant scientific issues are justified (Chapter 8).

This book is emphatically not just an exercise in epistemological method-

ology. Indeed, I spend far more time on first-order epistemological issues and
questions than I do on methodology. While I offer some broad methodological
considerations (particularly in Chapter 2), my aim is primarily to make the case
for non-ideal epistemology by doing it rather than by talking about doing it. In
the process, I hope that the distinctive features of non-ideal epistemology will
become clearer.
It might, however, be helpful if I say a little more at the outset about what non-
ideal epistemology is and what I take its distinctive features to be. I also want to
say something about how my approach in this book differs from other
approaches, especially feminist epistemology, and provide the reader with an
overview of the book’s contents. I start with the task of saying what non-ideal
epistemology is in §1.2, before turning to situating my approach with respect to
other approaches in §1.3 and the overview in §1.4.

1.2 Three Faces of Non-Ideal Epistemology

My distinction between ideal and non-ideal epistemology is based on Charles

Mills’ work on ideal and non-ideal theory in ethics and political philosophy (Mills
2005, 2007). Following Mills, I understand ideal epistemology as an approach to
epistemological issues and questions that involves certain characteristic idealiza-
tions. Non-ideal epistemology is, then, an approach to epistemological issues and
questions that eschews these sorts of idealizations. The idealizations that are
characteristic of ideal epistemology include:

• Idealizations about the nature and psychology of epistemic agents or

inquirers (e.g. about their cognitive capacities).
• Idealizations about the interactions between inquirers (e.g. about the extent
to which they are influenced by social power differentials).
6 - 

• Idealizations about social institutions (e.g. about their capacity to produce

and disseminate knowledge).
• Idealizations about the environments in which inquirers are embedded (e.g.
about the prevalence of information over misinformation).

I expand on these brief remarks in Chapter 2 but for now let me say two further
things. First, it may often be the case that ideal and non-ideal epistemology deal
with different issues and questions. This a consequence of the fact that there is a
connection between the sorts of idealizations we engage in and the issues and
questions that seem most pressing. If you work with a highly idealized picture of
epistemic agents on which you abstract away from aspects of their social situation
such as their social identity or role, you are hardly going to consider whether there
are interesting epistemological differences between differently situated agents (e.g.
do they have access to different bodies of evidence, as feminist standpoint theorists
suggest?). If you ignore the fact that there are power differentials between epi-
stemic agents, you are hardly going to consider the epistemological consequences
of social power differentials between epistemic agents (e.g. do we afford more
credibility to agents with more social power?). What this tells us is that one way in
which non-ideal epistemology might improve on ideal epistemology is by identi-
fying issues and questions that are of epistemological interest but obscured by the
idealizations that are typical of ideal epistemology.
Second, the reader might find it uninformative to be told that ideal epistemol-
ogy is an approach to epistemology that deals in idealizations while non-ideal
epistemology is an approach that avoids idealizations. But this is no less inform-
ative than standard ways of distinguishing between traditional (or individual)
epistemology and social epistemology. On one way of drawing the distinction,
traditional and social epistemology differ in that traditional epistemology focuses
on socially isolated individuals while social epistemology focuses on individuals
embedded in a social context. This is all very well, but it is not terribly informative.
If you want to better understand the distinction between traditional and social
epistemology, you are best advised to look at concrete examples of social epis-
temological projects and compare them with concrete examples of more trad-
itional epistemological projects. Similarly, if you want to better understand the
distinction between ideal and non-ideal epistemology, you need to look at con-
crete examples of non-ideal epistemology. The chapters that follow supply several
such examples.
Over and above offering a characterization of ideal and non-ideal epistemology,
this book identifies three key aspects or ‘faces’ of non-ideal epistemology. The first
key aspect or face of non-ideal epistemology is a focus on systems of knowledge
production and the social institutions that play a crucial role in these systems.
Now, a focus on systems and institutions is of course also a key aspect of
social epistemology. Consider, for example, what Alvin Goldman (2010a) calls
   -  ? 7

‘systems-oriented social epistemology’ or what Elizabeth Anderson (2006) calls

‘institutional epistemology’. Because these approaches are quite similar, I will
confine my attention to Anderson’s institutional epistemology. For Anderson,
institutional epistemology is a branch of social epistemology that looks at the
epistemic powers of social institutions. It considers questions such as:

• What sorts of knowledge reside within our social institutions, and what sorts
of problems is this knowledge needed to solve?
• What problems should we assign to these institutions?
• How can these institutions be (re)designed to improve their epistemic
powers?

These questions can be answered, and the programme of institutional epistem-

ology can be pursued, in an idealized fashion or a non-idealized fashion. To make
things more concrete, let us focus on the third question. One way of answering
it—a way typical of an ideal approach to institutional epistemology—would be to
consider what the optimal (epistemic) design of a social institution like science
might be. On this approach, a central question would be something like ‘how
might we design science as a social institution so that it produces knowledge of the
things we want to know about and tackles the sorts of problems we want it to
tackle?’ Philip Kitcher’s work on science is a prime example of this sort of
approach to institutional epistemology (Kitcher 2001, 2011).
Another way of answering the third question—a way typical of non-ideal
epistemology—would be to start with the social institution as it currently is and
ask which concrete steps we could take to improve its epistemic design. The
proposed modifications would need to be evidence-based. That is, there would
need to be evidence that the proposed modifications would secure the desired
epistemic improvement. In the first half of this book, particularly in Chapters 3
and 4, I pursue this non-ideal approach to institutional epistemology. I do this in
the context of a pressing social and political problem: science denialism, in
particular the various forms of global warming scepticism. My task will be to
survey the literatures on the causes and psychological drivers of global warming
scepticism and on effective strategies for persuading sceptics to change their mind.
Based on a survey of these literatures, I will make concrete proposals for what
science communicators can do to combat global warming scepticism. The first
half of this book therefore illustrates the institutional face of non-ideal
epistemology.
While non-ideal epistemology pays particular attention to social institutions, it
does not ignore individuals. The second face of non-ideal epistemology is a view of
epistemic agents or inquirers as deeply socially situated. A view of epistemic
agents as in some sense socially situated is of course also a key aspect of social
epistemology. But the crucial question is which aspects of social situation are
8 - 

viewed as of epistemological importance. It is one thing to acknowledge that

epistemic agents are socially situated in the minimal sense that they depend on
each other for information. It is quite another to hold that aspects of social
situation such as social identity and role are relevant to our epistemic obligations
and responsibilities. It is the second, deeper sense of social situatedness that is
characteristic of non-ideal epistemology.
This point can be amplified by considering the literature on feminist epistem-
ology (for more on feminist epistemology, see §1.3). One of the core ideas of
feminist epistemologies is that epistemic agents are socially situated in precisely
this deeper sense. It is striking that, while many of the classic texts in feminist
epistemology take issue with traditional epistemology for ignoring our ‘social
situatedness’, the same critique can be levelled at much social epistemology too
(see Mills 2007). Take, for example, Lorraine Code’s (1991) classic What Can She
Know? Code’s explicit target is what she calls ‘S knows that p epistemologies’:
theories of knowledge that take the form ‘S knows that p iff . . .’ where ‘S’ stands for
all knowers. Code thinks these ‘S knows that p epistemologies’ are mistaken
because they ignore the fact that differently socially situated knowers have access
to different bodies and types of evidence, have different cognitive capacities, and
have different epistemic obligations and responsibilities.
If this is a good criticism of ‘S knows that p epistemologies’, it is also a good
criticism of much of the literature on the epistemology of testimony, the
epistemology of disagreement, or indeed many of the central topics in contem-
porary social epistemology. If we need to consider social situation, we cannot
ask what the rational response to testimony or disagreement is without speci-
fying the social situations of the relevant parties (does one occupy a more
socially powerful position than the other does?). Code’s objection is therefore
not just an objection to ‘S knows that p epistemologies’. It is an objection to an
approach to epistemology that abstracts away from aspects of our social situat-
edness that may be epistemologically relevant. It is an objection to ideal
epistemology.
In the second half of this book, particularly Chapters 6 and 7, I develop this
objection. I start (in Chapter 6) by looking at John Stuart Mill’s famous argument
in On Liberty that we all have an obligation to engage with challenges to our
beliefs. I argue that we can explain both why many find Mill’s argument attractive
and why it is mistaken if we understand his argument as driven by certain
idealizations that are typical of ideal epistemology. Further, I argue that, once
we abandon these idealizations, a different picture of our epistemic obligations
and responsibilities emerges. On this picture, the nature and extent of your
epistemic obligations and responsibilities depend on whether you can expect
other inquirers to satisfy their epistemic obligations to you. Inquirers who cannot
expect other inquirers to satisfy their obligations towards them may have different
obligations than inquirers who can expect these things.
   -  ? 9

I then (in Chapter 7) develop the view that our epistemic obligations and
responsibilities are socially situated using the framework of ‘liberatory’ virtue
and vice epistemology.¹ In the process I address a (if not the) central challenge
for liberatory virtue (and vice) epistemology. The liberatory virtue epistemologist
emphasizes the extent to which our characters are the product of our social
situation. But (you might think) a character trait only qualifies as a virtue (or
vice) if the possessor is responsible for having it, and (you might also think) if
our characters are the product of our social situation then we cannot be
responsible for them. I argue that the version of liberatory virtue epistemology
defended by José Medina in his 2012 book The Epistemology of Resistance has the
resources to deal with this challenge. In the process, I reconstruct Medina’s
account of epistemic responsibility, which is distinctive in that it socially situates
epistemic agency and responsibility. It is, therefore, a non-ideal account of
epistemic responsibility.
The third face of non-ideal epistemology runs through the book and is more
an argumentative strategy than a claim or thesis. Several of the chapters that
follow develop a general objection to ideal epistemology. Rather than stating the
objection in the abstract, let me give a compelling example of it. (I do not take
up this example in the book; I have nothing to add to what has been said
about it.)
One strand running through feminist epistemology and the philosophy
of science is an objection to a common way of thinking about objectivity
(Anderson 1995, 2017; Harding 1995; Longino 1997). On this way of thinking,
being objective is a matter of being detached, disinterested, and not emotionally
invested in the outcome of your inquiries. An inquiry is then objective to
the extent that the inquirers are objective: they keep their personal views,
feelings, and value commitments out of their inquiries. This way of thinking
about objectivity expresses an intellectual ideal that I will call ‘objectivity as
detachment’. This ideal was—and in some circles still is—common in the
philosophy of science. The thought is that what distinguishes scientific from
other forms of inquiry is its commitment to the ideal of objectivity as detach-
ment (Lacey 1999).
The objection to this way of thinking of objectivity developed by feminist
philosophers of science combines two claims. The first, which is descriptive, is
that the ideal of objectivity as detachment is unattainable. It is extremely hard—
perhaps impossible—to genuinely not be emotionally invested in the outcome
of your inquiries, or to keep your personal views and values out of it. Worse,

¹ I use the term ‘liberatory epistemology’ to cover approaches to epistemology that are methodo-
logically similar to feminist epistemology but focused on aspects of social identity beyond gender.
Charles Mills’ work will be particularly important in what follows (especially Mills 2007), as will José
Medina’s (especially Medina 2012).
10 - 

we are often blind to the fact that we are emotionally invested in the outcome
of our inquiries, and to the ways in which our views and values influence
their direction. This is the point of several influential works of feminist
science criticism that emphasize the extent to which scientists were blind to
the impact of sexist assumptions and biases on their work (see e.g. Fine 2010
and Keller 1985).
The second claim, which is normative, is that scientific inquiry is not improved
by trying to approximate the ideal of objectivity as detachment. You might hold
that, even though it is not possible to remain completely detached, or to keep all
your personal views and values out of inquiry, the ideal of objectivity as detach-
ment can still function as a regulative ideal. It is an ideal that we should try to
approximate even though we are unlikely to reach it.²
The feminist rejoinder is that we should not try to approximate the ideal of
objectivity as detachment because in doing so we are liable to worsen rather than
improve our inquiries. The thought is that we improve inquiry—make it more
likely to produce knowledge—not by trying to keep as many values out of it as we
can but by making sure it is informed by the right values. Many works of feminist
science criticism emphasize that, in cases where feminist political values guided
scientific interventions, the result was better science. For instance, in her work
Elizabeth Anderson has argued that social scientific research is better if it is
informed by feminist political values (e.g. Anderson 2004).
The following chapters develop an objection to ideal epistemology that has a
similar structure to this objection. That is, they argue that the ideal epistemologist
proposes intellectual ideals and norms of inquiry that are not only unattainable
(the descriptive claim) but also such that, in trying to attain them, we run the risk
of doing worse epistemically than we would if we did not try to attain them (the
normative claim). The claim is therefore that the ideals and norms proposed by
the ideal epistemologist are often not regulative ideals. We should not try to
approximate them at all. For example, in Chapter 5 I argue that, in striving to
be intellectually autonomous, we run the risk of sacrificing other, more valuable,
intellectual goals. The lesson is that it is often better not to strive to be intellec-
tually autonomous because doing so will often worsen rather than improve our
epistemic situation.
This should suffice for now as an explanation of what I mean by ideal and non-
ideal epistemology and of the key claims I make in this book. I now want to turn
to situating my project with respect to three others that might seem similar to
my own.

² For interesting discussions of regulative ideals, see Emmet (1994) and Rescher (1987). Note that
neither Emmet nor Rescher suggests that this way of thinking about objectivity is a regulative ideal.
Their point is that there can be regulative ideals—ideals that are unattainable but still serve to regulate
our practices. I agree that there can be regulative ideals but in the following chapters I try to show that
many of the ideals proposed by ideal epistemologists are not regulative ideals.
   -  ? 11

1.3 Non-Ideal Epistemology and Feminist Epistemology

The first project I want to briefly situate my own with respect to is the sociology of
scientific knowledge, particularly as developed by the ‘Strong Programme’. Key
figures in the Strong Programme include Barry Barnes, David Bloor, and Harry
Collins (see Barnes 1977; Barnes and Bloor 1982; Bloor 1976; Collins 1985). The
Strong Programme is not centrally concerned with epistemology or idealization
within it (the focus is more on philosophy of science). But it is certainly true that
its proponents view traditional epistemology as one of its targets, and part of the
reason why it is among the targets is that traditionally epistemologists were not
interested in the social causes of belief (Kusch 2010).
One of the central principles of the Strong Programme is the ‘symmetry
principle’, which says that we should ask about the causes of belief without any
regard to whether the beliefs in question are true or false, or rational or irrational
(Barnes and Bloor 1982). If we do this, and foreground sociological and psycho-
logical factors in our explanations of beliefs, then the sorts of factors epistemolo-
gists like to cite (reasons, evidence) just will not feature.
Many epistemologists view the symmetry principle and the broader meth-
odology it embodies with suspicion because they seem to amount to an attempt
to debunk the authority of knowledge by reducing it to social interests
and power (Boghossian 2006). Whatever the merits of this criticism, the
proponents of the Strong Programme are clearly not concerned with the
normative assessment of belief or processes of belief-formation. If we are not
going to assess beliefs for truth or falsity, we are not going to be engaging
in normative assessment of beliefs or believers. We are going to focus on the
descriptive question of why people believe what they do, or why they work
with the epistemic norms that they do, rather than the normative question
of what they should believe or how they should conduct their inquiries. My
interest in this book is in both the descriptive and the normative questions.
While I think that any serious attempt to answer the normative question needs
to start with the descriptive question (see Chapter 4), I do not think that the
descriptive question replaces the normative question or makes it obsolete. My
concerns are therefore different from those of ‘Strong Programmers’ and closer
to those of traditional epistemologists.
The second project I want to situate my own with respect to is really another
way of drawing the distinction between ideal and non-ideal epistemology. My
initial gloss of the distinction was that, where ideal epistemology involves certain
idealizations, non-ideal epistemology avoids these idealizations. But, as I will go
on to detail in Chapter 2, there are many ways in which you might idealize.
I highlight idealizations about human cognitive capacities, social institutions,
social interactions, and the social environment. Another way of drawing the
distinction between ideal and non-ideal epistemology focuses primarily on
12 - 

idealizations concerning our cognitive capacities (Carr forthcoming). On this way

of drawing the distinction, ideal epistemologists are concerned with what perfectly
rational, cognitively unlimited agents would believe while non-ideal epistemolo-
gists are concerned with the norms governing humans with their many cognitive
limitations.
I have no objection to this way of drawing the distinction. Distinctions can be
drawn in many ways. But you can always ask, of a particular way of drawing a
distinction, whether it highlights or obscures the things you are interested in. This
way of drawing the distinction (in terms of cognitive capacities) obscures the
fact that you can recognize our cognitive limitations while still making all sorts
of other idealizations. For instance, you might acknowledge our cognitive limita-
tions yet still work with idealized pictures of the social interactions between
cognitively limited agents, and about the environment in which they do the
interacting. Because my aim is to identify an idealizing tendency in social epis-
temology, I need a way of distinguishing between ideal and non-ideal epistemol-
ogy that goes beyond our cognitive capacities. Consequently, in the next chapter
I foreground idealizations about social interactions, social institutions, and our
social environment as well as idealizations about our cognitive capacities.
The result is a conception of non-ideal epistemology that is explicitly ethical
and political in that it brings phenomena like injustice and oppression into the
purview of epistemology. I do not claim that no other conceptions are available.
But I do claim that my conception ties epistemology up with debates in social and
political philosophy. Given that deeper engagement with social and political
philosophy is a clear trend in twenty-first-century epistemology, my conception
of non-ideal epistemology is well-placed to make sense of—and contribute to—
these developments.
The third and final project I want to situate mine with respect to is feminist
epistemology. Because there are many similarities, I will go into some detail. As
what I said earlier should already make clear, in this book I make use of several
insights from the feminist (and liberatory) epistemological literature. In places,
I also try to contribute to this literature (e.g. Chapter 7, where I address a key
challenge for liberatory virtue epistemology). It is, however, important to empha-
size that my distinction between non-ideal and ideal approaches to epistemology
is not equivalent to the distinction between feminist and traditional or main-
stream epistemology. This is for two reasons.
The first reason is that, in some respects, my distinction is broader than that
between feminist and traditional or ‘non-feminist’ epistemology. Recall that non-
ideal epistemology has three aspects or faces: a focus on institutions, a view of
inquirers as deeply socially situated, and a distinctive argumentative strategy. Of
these faces, only the second is particularly characteristic of feminist epistemology.
Non-ideal institutional epistemology can be specifically feminist, as when social
institutions are analysed through a ‘feminist lens’, or when an epistemological
   -  ? 13

analysis of institutions serves feminist political goals. But in the first half of the
book (Chapters 3, 4, and 5) I tackle the problems raised by ‘science denialism’,
particularly global warming denialism, without adopting a specifically feminist
approach to them.
The second reason is that, in other respects, my concerns are narrower than the
concerns of feminist epistemologists. One important strand in feminist epistem-
ology is what we might call the ‘feminist theory of knowledge’; the feminist
alternative to Code’s ‘S knows that p epistemologies’. In previously published
work I have argued that the core feminist critique of ‘S knows that p epistemolo-
gies’ is best understood as methodological rather than as allied to any substantive
claims about the nature of knowledge (Ashton and McKenna 2020; McKenna
2020).
Let me briefly rehearse this argument. The epistemological tradition treats
certain kinds of knowledge—those amenable to the ‘S knows that p’ analysis—as
paradigmatic. Feminist epistemologists invite us to consider what a theory of
knowledge would look like if it were constructed around different paradigms. As
Anderson puts it in her excellent overview of feminist epistemology:

Mainstream epistemology takes as paradigms of knowledge simple propositional

knowledge about matters in principle equally accessible to anyone with basic
cognitive and sensory apparatus: “2 + 2=4”; “grass is green”; “water quenches
thirst.” Feminist epistemology does not claim that such knowledge is gendered.
Paying attention to gender-situated knowledge enables questions to be addressed
that are difficult to frame in epistemologies that assume that gender and other
social situations of the knower are irrelevant to knowledge. Are certain perspec-
tives epistemically privileged? Can a more objective perspective be constructed
from differently gendered perspectives? (Anderson 2017)

This is a critique of a kind of ideal epistemology that takes the paradigm cases of
knowledge to be the simple ones—the ones where social situation seems irrele-
vant. In contrast, the feminist theorist of knowledge takes the paradigm cases of
knowledge to be the ‘messy’ ones—gendered and more broadly socially situated
forms of knowledge. Anderson’s point is that your choice of paradigms will
inform your theory of knowledge. If we view simple cases as the paradigm, we
will construct a theory of knowledge designed to accommodate simple cases and
then try to extend it to messier ones. (Most likely, we will get stuck on the simple
cases and never get to the messy ones.) If we view messier cases as the paradigm,
we will end up with a messy theory of knowledge, or even no ‘theory’ of knowledge
at all.
While I am sympathetic to this critique, neither it nor the kind of ideal
epistemology it targets is my focus in this book. More generally, my target in
this book is not the epistemologist who is interested in the theory of knowledge,
14 - 

the theory of justification, norms of belief and assertion, or anything of that sort.
My target is rather the epistemologist who is engaged in an overly idealized form
of what Quassim Cassam (2016) calls ‘inquiry epistemology’. The ‘inquiry epis-
temologist’ is interested in the activity of inquiry—in how we go about extending
our knowledge. They ask questions about the mechanics of inquiry (how do we
inquire?). They ask normative questions about inquiry (what is a responsible
inquiry like?) and inquirers (what is a responsible inquirer like?). But they are
also interested in improving inquiry. They want to make it more effective (better
at reaching truth) and more responsible (better at reaching truth in the right
kind of way).³
However, it is possible to do inquiry epistemology in an overly idealized way.
For example, the inquiry epistemologist might make use of idealizations about
what inquirers are like, the environment in which they conduct their inquiries,
how they interact with each other, or the social institutions which aide or hinder
their inquiries. As a result, their proposals for improving inquiry might be such
that trying to put them into practice would be likely to achieve the opposite
result—to make us less likely to achieve our epistemic goals. In the following
chapters, I argue that some inquiry epistemologists (including Cassam) fall into
precisely this trap. In making this case I will often draw on work from feminist
epistemology that, in my terms, urges the importance of adopting a non-ideal
approach to inquiry epistemology (see Chapters 6 and 7).
You might object that what I am calling ‘inquiry epistemology’ has its proper
home in ethics rather than epistemology. More modestly, you might worry that
inquiry epistemology blurs the distinction between epistemology and ethics in a
way that is confusing and perhaps even problematic. If the reader has these
concerns, be forewarned. In the following chapters, I spend little time on whether
what I am doing is best understood as epistemology, ethics, or some combination
of them both. I also do not make much effort at keeping epistemological questions
separate from ethical ones. But this lack of care is, I like to think, principled. I do
not think we can neatly distinguish between epistemology and ethics, nor do
I think that we should try. I am interested in norms of inquiry and if we need to
blur the divide between epistemology and ethics to profitably think about them

³ A few remarks about ‘inquiry epistemology’ that do not fit into the main body of the text. First,
Cassam credits Alfano (2012) with coining the term and Hookway (2003) with articulating the idea in a
particularly clear fashion. But the general approach is far from new. Indeed, it is arguably what some of
the founding figures of epistemology as we know it today (Descartes, Locke) were up to (Pasnau 2017;
Wolterstorff, 1996). Second, by ‘inquiry epistemology’ I mean something like what Nathan Ballantyne
(2019) calls ‘regulative epistemology’. I prefer ‘inquiry epistemology’ to ‘regulative epistemology’
because, while I applaud Ballantyne’s aspiration to improve inquiry rather than just describe it, it is
important not to downplay the importance of arriving at a proper understanding of how inquiry works
(see §4.1). Third, the relationship between inquiry epistemology and the literature on epistemic
normativity (norms of action, assertion, and belief ) is a little tricky. I briefly touch on this in the
next footnote, in §2.4, and again at the beginning of Chapter 6. For relevant discussion, see Friedman
(2020).
   -  ? 15

then, as far as I am concerned, the divide can be blurred. I return to this issue
(if only briefly) in Chapters 4 and 6.

1.4 Overview

Finally—and apologies for taking so long to get there—here is the overview. As

already advertised, Chapter 2 (‘Ideal and Non-Ideal Theory’) develops my initial
characterizations of ideal and non-ideal epistemology. I start by looking at the
debate between ideal and non-ideal theory in ethics and political philosophy.
I then consider whether we can construct epistemological analogues of the ways in
which the distinction between ideal and non-ideal theory has been understood in
this debate. I look at three ways in which the distinction has been drawn: John
Rawls’ distinction between full and partial compliance theory (Rawls 1971), a
distinction Laura Valentini has drawn between utopian and realistic theory
(Valentini 2012) and Mills’ list of the sorts of idealizations he takes to be typical
of ideal theory (Mills 2005). I argue that Mills’ characterization is more useful than
Rawls’ because it is more general. Further, I argue that realistic theory in
Valentini’s sense encompasses both ideal and non-ideal epistemology and so
does little to advance our understanding of the differences between them. I also
explain why ideal and non-ideal epistemology are not necessarily opposed to each
other, further situate my project with respect to social epistemology, and address
several objections to my project.
Chapter 3 (‘Anderson and Goldman on Identifying Experts’) illustrates the
contrast between ideal and non-ideal epistemology via a detailed case study.
I focus on the problem of the identification of expertise and on the contrasting
approaches taken to this problem by Alvin Goldman (2001) and Elizabeth
Anderson (2011). I argue that Goldman’s approach is a clear example of ideal
epistemology while Anderson’s is closer to non-ideal epistemology. I also argue
that Anderson’s approach to the problem is preferable to Goldman’s. Where
Goldman sees the problem as one that arises for individual inquirers, Anderson
sees the problem in more institutional and political terms. She is interested in how
we might construct better systems of knowledge production and a better epistemic
environment. She is, therefore, pursuing a non-ideal form of institutional
epistemology.
The two chapters that follow address two tasks that need to be carried out to
properly defend a non-ideal institutional epistemology like Anderson’s. The first is
to say more about how we might construct better systems of knowledge produc-
tion and dissemination and about how we might construct a better epistemic
environment. Doing this requires looking in some detail at the empirical literature
on how to ‘market’ science to secure the maximum possible public uptake. The
second task is to deal with a fundamental objection to Anderson’s approach,
16 - 

which is that what she proposes involves interfering with our intellectual
autonomy (our right to make up our own minds on issues of importance to us).
Chapter 4 (‘Persuasion and Paternalism’) takes up both tasks. I start by iden-
tifying a set of ‘science marketing’ strategies that we have reason to believe would
be effective in constructing a better epistemic environment—an environment in
which individuals are more likely to form true beliefs about scientific issues, like
global warming, which are relevant to public policy. I then argue that we need to
take seriously the worry that, even if these strategies are effective, they infringe on
our intellectual autonomy. I try to defuse this worry by distinguishing between
interventions on our inquiries that bypass our critical faculties but are trying to
enhance our intellectual autonomy and interventions that undermine our very
capacity for autonomous thinking and deliberation. The sorts of ‘marketing
strategies’ I discuss in this chapter fall into the former category and so need not
infringe on intellectual autonomy.
Chapter 5 (‘Intellectual Autonomy’) challenges the assumption that intellectual
autonomy is valuable in the first place. My claim is that intellectual autonomy is
an intellectual ideal or goal that many of us frequently and predictably fall short of.
Moreover, it is an intellectual goal that often frustrates our other goals, in
particular the goal of arriving at true beliefs about matters of importance to us.
The chapter concludes that, in the absence of reasons for thinking that intellectual
autonomy is more important than other intellectual goals, we often do better not
to strive for it. The chapter finishes by addressing the implications for the
argument of the previous chapter. If intellectual autonomy is less important
than many suppose, then it is unclear why the fact (if it is a fact) that some
form of intervention with someone’s inquiries interferes with their intellectual
autonomy is even a prima facie reason against that intervention. This chapter
therefore amplifies the argument of the previous chapter by undercutting the
original objection against ‘science marketing’.
Chapters 3, 4, and 5 focus on the first, institutional, aspect or face of non-ideal
epistemology. Chapters 6 and 7 turn to the second aspect or face, which is a view
of epistemic agents or inquirers as deeply socially situated and of their epistemic
obligations and responsibilities as depending on aspects of their social situation.
Chapter 6 (‘The Obligation to Engage’) focuses on one obligation in particular: the
(supposed) obligation to engage with challenges to our beliefs. I start by outlining
the best-known defence of the view that we have such an obligation, which is that
given by John Stuart Mill in On Liberty. I then look at a rather different defence of
the same view, which can be found in Quassim Cassam’s (2019) book Vices of the
Mind.
Where Mill’s view is that we all have an obligation to engage with challenges to
our beliefs because this is the best way of securing certain epistemic benefits,
Cassam’s view is that it is only by engaging with challenges to our beliefs that we
earn the right to them. I argue, against both Mill and Cassam, that we do not all
   -  ? 17

have this obligation, at least with respect to all our beliefs. As I argue, the problem
with both Mill’s and Cassam’s arguments is that they are based on assumptions
that are typical of ideal epistemology. These assumptions are: (i) that inquirers will
satisfy their obligations, and (ii) that our epistemic environment is (relatively)
hospitable.
Once we recognize that these assumptions are false, a different picture of our
epistemic obligations and responsibilities emerges. On this picture, the nature and
extent of your epistemic obligations and responsibilities depends on whether you
can expect other inquirers to satisfy their epistemic obligations to you and on how
hospitable the epistemic environment is for you. Inquirers who cannot expect
other inquirers to satisfy their obligations towards them, or for whom the epi-
stemic environment is not hospitable, may have different obligations and respon-
sibilities than inquirers who can expect these things and for whom the epistemic
environment is (relatively) hospitable.
Chapter 7 (‘Liberatory Virtue and Vice Epistemology’) develops the idea that
our epistemic obligations and responsibilities depend on our social identities and
situations using the framework of liberatory virtue (and vice) epistemology.
Specifically, it develops this idea through the version of liberatory virtue epistem-
ology defended by José Medina in his 2012 book The Epistemology of Resistance.
I argue that Medina’s virtue epistemology is distinctive in that it socially situates
epistemic agency and responsibility. It is therefore a recognizably non-ideal form
of virtue epistemology.
Chapter 8 is a little different from the earlier chapters. It considers the epis-
temological implications of the empirical literature on motivated reasoning, which
I draw on in earlier chapters. As this literature documents, the way we think about
issues that matter to us (especially political issues) is quite different from how the
epistemologist might imagine it to be. We often engage in so-called ‘motivated
reasoning’: we gather and assess evidence in ways that serve our goals. For
example, when presented with evidence that conflicts with our goal of thinking
well of ourselves (thinking of ourselves as healthy, kind, compassionate, etc.), we
tend to find ways of discounting that evidence. When gathering evidence about an
important issue (e.g. global warming), we tend to gather evidence in ways that
support a stance on the issue that serves our political goals (e.g. looking for
evidence that humans are not responsible as a way of preventing political action).
I argue that this provides the basis for an empirically driven argument for a form
of scepticism. Specifically, it supports the conclusion that many of us do not have
justified beliefs about political issues and scientific issues that have become
politically contentious, such as global warming.
I finish the chapter and the book by considering whether we can avoid this
sceptical conclusion by making a move you might think is like the move I have
made in earlier chapters. Can we argue that the fact (if I am right) that a kind of
justification for our political and (some) scientific beliefs is unattainable provides a
18 - 

reason to revise our understanding of what having justification for these beliefs
requires? I argue that we cannot because, unlike in the earlier chapters, it is not the
case that striving for this kind of justification worsens our (or anyone else’s)
epistemic situation. This chapter therefore serves not only to illustrate the ways
in which non-ideal epistemology might contribute to more traditional epistemo-
logical debates but also to clarify my more general objection to ideal epistemology.
Ideal epistemology is problematic only to the extent that the goals and norms it
proposes run the risk of worsening rather than improving our epistemic situation.
The non-ideal epistemologist is not committed to the implausible claim that the
mere fact that a goal is unattainable or a norm hard to follow is itself a reason to
not try to attain or strive for it.