MA Philosophy
Master Thesis
What is the Best Way to Respond to the Bias
Paradox?
By
Mary McGuire
11314095
August 2017
Supervisor:
Table of Contents
1. Introduction
p.3
2. Bias and impartiality
p.3
a. What is bias?
p.3
b. What is impartiality?
p.5
c. What is the bias paradox?
p.9
3. Responding to the bias paradox
p.10
a. Antony’s naturalised approach
p.10
b. Empirical assessment of bias and empirical truth
p.11
c. Anderson on contextual truth
p.16
4. What makes a theory epistemically justified?
p.20
a. The social approach
p.20
b. Diversity as an epistemic ideal
p.22
c. Integrating moral, political and epistemic values
p.26
5. Conclusion
p.29
a. Summarising bias
p.29
6. References
p.31
Introduction
In this paper I will be considering the questions, ‘how can we respond to the problem highlighted by the bias paradox?’ and ‘what can the bias paradox tell us about truth?’. In outlining a concept of bias I will consider both ethical and epistemic issues of justice, ultimately arguing that bias must be understood contextually. By considering the paradox itself and the problem with assessing bias, I will highlight the normative aspect to our notion of truth which leads me to argue in favour of an integration of moral and epistemic values in understanding justification.
To introduce the bias paradox, I will outline a general conception of bias drawn from ideas discussed by Louise Antony, Elizabeth Anderson, Deborah Heikes, and Miranda Fricker, and give few illustrative examples of bias before considering impartiality as an epistemic and ethical ideal as one approach to explaining the problem with bias. As we shall see, it is unclear how impartiality can be understood in a way that is epistemically beneficial, and I argue that it has been shown by Antony to be untenable as an epistemic ideal given the necessity of many forms of partiality.
I will then move on to consider Antony’s response to the bias paradox in more detail, and through comparison with Anderson’s explanation of the role of value in inquiry, I will highlight potentially difficult aspects of Antony’s approach, such as her notion of truth and a risk of circularity in relying on purely empirical understandings of justification. Specifically, I suggest Antony does not give enough consideration to the contextual nature of truth in theory building, to the difficulties in recognising truth, and to the significance of values in understanding not only bias but our concept of truth. I argue with Anderson that these values can often incorporate moral and political judgements which act epistemically. Through a close reading of Anderson and Antony’s texts I suggest that they can be used to support a contextual notion of truth, and that the context must be understood as normative.
After summarising conclusions drawn from the above discussions, I will outline my response to the bias paradox by considering epistemic justification, acknowledging a normative aspect to truth, and considering the nature and meaning of empirical assessment in relation to contextual values. In forming my summary of bias and value-driven approach in responding to the bias paradox, I will use insights from the authors mentioned, as well as from Kristen Intemann who discusses the epistemic significance social diversity and standpoints, and Naomi Scheman whose argues for what she terms sustainable epistemology and an integration of moral, political, and epistemic values.
Overall I see my arguments as being useful in situations involving knowledge which is socially or morally significant, and along with authors considered, I will be motivated by issues relating bias to moral and social justice, as well as wider notions of epistemic justice. I hope to show that the concept of bias arises because notions of epistemic truth and moral and socio-political justice often overlap, and that both must be understood in the context of background beliefs and normative assumptions about our overall aims.
Bias and Impartiality
What is bias?Heikes introduces her discussion of bias by stating that “broadly speaking, bias is an inclination that interferes with impartial judgement” (2004, p.316). I will now expand on this general statement of bias and discuss how bias, partiality, and impartiality are linked by considering arguments made by Antony and Anderson. Overall we will see that further work is needed to understand the ways in
which bias operates problematically in particular, and not just in contrast to impartiality. I will also use insights from Fricker’s discussion of prejudice, understanding this as a form of bias.
Bias can be most generally conceived as an interest or (set of) belief(s) that leads a person to favour certain viewpoints, reach certain conclusions, or take certain facts to be true. In her discussion of the bias paradox, Antony gives a similar initial analysis of bias as “the mere possession of belief or interest prior to investigation” (1994, p.100). This general analysis of bias conflates bias and partiality, and Antony instead distinguishes between useful and problematic forms of partiality by using the terms ‘good biases’ and ‘bad biases’ (1994, p.101). Overall she gives us a view of bias as any form of partiality resulting from background assumptions or beliefs that can either hinder or aid the development of knowledge. In her view, ‘bad bias’ (or the ordinary conception of bias as distinct from partiality) is problematic because it leads us away from truth (1994, p.137). Her focus is therefore primarily on the epistemic aspect of bias, which she argues is distinct from ethical issues. In cases where moral concerns are particularly relevant, we can also consider problematic bias as unethical, as it can lead us to unjust conclusions. I suggest that this moral conception of bias can be understood as similar to Fricker’s notion of prejudice, which she defines as “judgements… which display some (typically, epistemically culpable) resistance to counter-evidence owing to some affective investment on the part of the subject” (2007, p.35). What makes Fricker’s conception of prejudice in line with Antony’s bias is the fact that it is an ‘affective investment on the part of the subject’; it is the nature of the partial interests or prior beliefs of the subject that create bias. What Fricker’s account highlights is the interconnectedness of epistemic and ethical issues when discussing prejudice or bias. She gives a similar definition of ethically problematic identity prejudice as stemming from “ethically bad affective investment” (2007, p.35). Fricker explains that moral bias in the form of prejudice is still epistemically problematic, because it affects how we interpret evidence or causes ‘resistance to counter-evidence’. As will become evident from my later discussion of the relationship between evidence and truth, Fricker’s reliance on the notion of ‘resistance to counter-evidence’ in highlighting epistemic problems neglects certain difficulties with the issue of interpreting and identifying the relevant evidence, however the point remains that moral bias presents an epistemic problem due to the often false conclusions it produces. Initially then, we can see that bias (or in Antony’s terms ‘bad bias’) presents both a moral and an epistemic problem. To illustrate the basic notion discussed so far of bias as epistemically and morally problematic, I will give four examples. The first three are examples of morally significant bias, and the final example is designed to illustrate a specifically epistemic notion of bias. I believe it is important to distinguish between these aspects in order to understand Antony’s focus on an epistemic account of bias and how this influences her discussion of the bias paradox.
1) Biased inquiry: a scientist with sexist background assumptions uses social and psychological experiments designed to reveal facts about gender differences. Because the scientist and the experiments presume the notions they are investigating, and prior beliefs about gender influence how the evidence is interpreted, the scientist and the results are biased. This example is drawn from Anderson’s discussion of gender classifications in psychological theory (1995a, p.48) and is similar to Antony’s example of racist inquiry (1994, p.140).
2) Biased portrayal: an historian uses facts to give an account of historical events that focuses on particular aspects or perspectives and ignores others, giving an overall picture that is biased. For example, consider a biased account of the discovery or colonisation of the Americas by European settlers which ignores the perspective of the native populations. The Secret Relationship, a book
about the role of Jews in the Atlantic slave trade discussed by Anderson (1995a, p.38), is another example of such a biased portrayal of events.
3) Biased jury: In a court of law, the purpose of a jury is to provide a judgement from a selection of the public who have no personal connection with the case and so are non-biased. However, juries are often still biased due to social prejudices. Fricker discusses the racist jury in To Kill a Mockingbird as an example (2007, p.23). Racist background assumptions constitute bias and mean the jury are unable to make a fair judgement.
4) Biased judgement: In drug trials, placebo-controlled, double-blind trials are carried out to ensure that biases of the participants and experimenters, or the well-known placebo effect and wishful thinking, do not interfere with assessing the effectiveness of the drug under trial. This is Anderson’s example (1995b, p.55) and shows how bias might distort judgment in a non-moral situation and still constitute an epistemic problem; the experimenters and participants are impaired in their ability to accurately interpret results of the treatment.
The bias acts in each case differently. In the first case, the experiments are biased because they use a certain framework or set of assumptions which causes the evidence to be interpreted in a certain way. In the case of the biased portrayal, the choice of which facts are taken as relevant is subject to bias, meaning that the account highlights certain facts as significant and does not consider the wider context. The Jury’s verdict is biased because of distorted perception or interpretation due to racist beliefs. Similarly in the final example, wishful thinking leads to biased interpretations of data that do not accurately assess the effectiveness of the drug. In each case, bias arises due to prior, influencing interests or beliefs, and in some of the cases this bias amounts to Fricker’s prejudice.
The idea of distorted perception or cognitive distortion is a significant aspect of how bias manifests itself. Fricker elaborates her account of identity prejudice, which can be understood as harmful bias stemming from an unreliable social stereotype, by explaining that “[identity prejudice] distorts the hearer’s perception of the speaker” (2007, p.36). Similarly, Anderson talks about ‘cognitive distortions’ (1995b, p.66) that constitute bias. Similar to Fricker’s focus on the unreliability of certain stereotypes in identifying harmful prejudice, Anderson discusses epistemic reliability, explaining that rational inquiry must cancel out “unreliable belief-formation mechanisms” (1995b, p.55). Both authors draw significance to the cognitive and perceptive processes of knowledge-formation, and how these can be susceptible to bias in the form of an unreliable process, which creates a distortion. The need for epistemic reliability in preventing distortion is also applicable to Antony’s conception of good bias, which she believes can be identified by how well it guides us to a truthful theory, or how reliable it is (1994, p.137).
Considering the above discussion, the overall problems with bias can be summarised by the following points. Bias can be understood as background assumptions or a framework that leads us to interpret evidence in a way that gives false and sometimes morally problematic results. It can influence the aims of our inquiries and determine the relevant facts in a way that produces accounts which ignore the wider context. Finally, bias can distort our perception and general ability to perceive information. Problematic bias in particular leads to falsehoods or unethical conclusions. Before considering these points in more detail and how they feature in Antony and Anderson’s accounts, I will firstly outline the notion of impartiality as discussed by these authors and how it could be used as an epistemic ideal to counter bias.
One way of identifying what it is about bias that hinders our abilities to gain knowledge and make fair moral judgements is the fact that it makes us problematically partial. To overcome this partiality and bias, we could aim for impartiality, meaning that we rid ourselves of preconceptions and assumptions which can distort our perception and guide us towards false conclusions. To return to the examples, the biased jury would reach a fairer verdict and fulfil its role better were it more impartial. Similarly, if the experimenters were more impartial when examining the results, they would be more successful in accurately determining the effectiveness of the drug. Impartiality seems useful both epistemically, for imparting and receiving knowledge, and ethically, for reaching a fair conclusion.
However, the case is not this simple. As both Antony and Heikes discuss, the traditional notion of impartiality has been criticised by feminists such as Haraway, Longino, Nelson, and Harding, as well as epistemologists working within the analytic tradition such as Quine, Putnam, and Goodman
(Heikes, 2004; Antony, 1994). Overall, the notion of impartiality needs clarifying before it can either
be endorsed or rejected as an ideal.
Antony and Heikes explain that the idea of impartiality as an epistemic ideal has its origins in 17th and
18th century epistemology and Enlightenment philosophy. Heikes focuses in particular on a Cartesian
conception of knowledge as giving rise to the ideal of impartiality, originating from Descartes’ famous abstraction from experience and a priori truth of the Meditations, ‘I think, I exist’, an argument known as the cogito. The significance of this claim relies on Descartes’ recognition of human experience as the partial, limited perspective of the world that it is, and therefore the fact that experience is not a reliable guide to what he conceives as objective reality (Newman, 2014). It is only by abstracting from experience and striving for impartiality through rationality and a priori methods that Descartes reaches his foundational, necessary truth that, as Heikes explains, “provide[s] the ground for impartiality on contingent matters” (2004, p.322). The consequence of the success of the cogito in particular is that “Descartes establishes a disembodied, detached reason as the domain of absolute clarity, certainty, and impartiality” meaning that “within a Cartesian framework, knowledge requires a complete detachment from the lives of actual knowers” (2004,
p.323). The notion of impartiality that Heikes draws is one that relies on reason and logic, for it is this
that ensures our beliefs are reliable and not products of distorted perception.
Significant for this view of impartiality as an epistemic ideal is the idea that flaws in our knowledge stem from distorted perception and experience, and that rationality is reliable for an individual gaining knowledge because it is not subject to the same distortions. Cartesian doubt relies on rationality as a method for arriving at truth; it is the senses and our perception that lead us astray. One problem with relying on a rationalistic understanding of impartiality is that even if we accept and ground all knowledge in this notion of rationality as truth-reaching and universal (a view which we could also criticise), experience will always be significant for gaining knowledge; as Antony points out, even rationalists such as Descartes did not deny the importance of experience in gaining knowledge (1994, p.110-113). This means that we will always be subject to some perceptual biases, and that the ideal of impartial rationality cannot account for all knowledge and cannot eradicate bias. It is impossible to escape experiential bias. In the Meditations, Descartes focuses on how sensory information can be subject to distortions and illusions, making experiential knowledge unreliable. Other forms of distorted perception that operate outside of rationality and that can be difficult for an individual to detect are, for example, the placebo-effect mentioned earlier and discussed by Anderson (1995b, p.55), or Fricker’s idea of testimonial perception, or the perception of a speaker involved in making a credibility judgement about the reliability of a testimony (2007,
(under the rationalist understanding of impartiality) agent can still be subject to these types of biases, and rationality does not ensure that biases will be detected. If we are to save rational impartiality as an epistemic ideal, further work must be done to form an account that accounts for experiential knowledge and bias.
Rather than criticising this view of impartiality by highlighting the limits of an individual’s rationality in forming beliefs and overcoming bias, Antony goes further and rejects the coherency of the notion of rationalist impartiality. She does this by clarifying the debate between rationalists and empiricists, explaining that “for the empiricists, the trick was to show how the content of all knowledge could be reduced to pure reports of sensory experience; for the rationalist, it was showing the indubitability of the innate notions that guided and facilitated the development of knowledge” (1994, p.115). Rather than being grounded in a Cartesian, rationalist conception of knowledge, Antony explains that impartiality as an ideal is in fact specific to empiricist conceptions of knowledge. She points out that rationalists do account for knowledge from experience and that rationalist conceptions of knowledge can be interpreted as claiming that humans are “natively biased towards certain ways of conceiving the world” (1994, p.112), this being simply a constraint on how our knowledge is structured and not entailing the impossibility of experiential knowledge.
From this, Antony explains that our innate cognitive functions can be understood as a form of partiality (1994, p.114) and therefore impartiality cannot be a rationalist notion. It was the empiricists who, as John Locke illustrated with his metaphor of the mind as a ‘blank slate’ or tabula
rasa, rejected innate ideas and argued that all knowledge originates from experience. Antony
explains that the empiricist ideal of impartiality can then be understood as “that spot which is least particular... where the regularities in one’s own personal experience match the regularities in the world at large” (1994, p.113). This entails that we must transcend our own embodied experiences and interpretation of the external world in order to discover an ‘objective truth’. To achieve impartiality we must disregard background assumptions, beliefs, or desires, which are not considered to be drawn solely from empirical experience.
Interpreting impartiality as empiricist allows Antony to use what she considers rationalist insights to highlight problems with the notion. For example, Antony states the problem with empirical impartiality as “the essentially rationalist insight that perfect objectivity is not only impossible but undesirable, that certain kinds of ‘bias’ or ‘partiality’ are necessary to make our epistemic tasks tractable” (1994, p.116). The idea here is that an understanding of the problem with bias as the fact that it makes us partial, or prevents impartiality, becomes untenable when we realise that partiality in general is not epistemically problematic and is in fact, both impossible to avoid and necessary for knowledge. Taking impartiality as, in Antony’s words, “giving equal weight to every hypothesis consistent with the data, or... drawing no conclusions beyond what can be supported by the data”
(1994, p.114), or in other words a disregard for prior and non-empirical assumptions, aims, and
desires, we are left with the problem of accounting for the fact that theoretical context and background beliefs are necessary for gaining knowledge; all knowledge requires us to be in a particular position or starting point. We must be partial in order to structure our knowledge and interpret our data.
To illustrate, epistemically useful forms of partiality include adopting a framework that supports our aims and interests in conducting an inquiry (e.g. an engineer classifying materials and properties in a way that aids construction), or having background knowledge that allows for a particular analysis of a situation (e.g. a doctor can give a medical interpretation). In conducting any inquiry or producing an account, a background framework is needed and a choice must be made concerning which facts are
relevant to the aims of inquiry. Fricker also discusses the value of partial judgements in her understanding of stereotypes as aids which are “a proper part of the hearer’s rational resources in the making of credibility judgements” (2007, p.30). She defines a stereotype in the neutral sense as “widely held associations between a given social group and one or more attributes” (2007, p.30) which can either be reliable or not. With this Fricker highlights the usefulness of partiality in the form of pre-judgements and importantly, value judgements, and hence why we cannot reject all forms of partiality in favour of impartiality. Without the types of judgements or stereotypes used in gaining testimonial knowledge (e.g. ‘doctors are a good source of medical advice’), effective transfer of knowledge between individuals would be impossible. An empiricist notion of impartiality is not desirable when judging which statements to trust as knowledge.
Antony draws on arguments made by Quine in recognising the significance of partiality in knowledge-formation. She explains that pure empiricism cannot work:
If we had to rely on nothing but logic and the contingencies of sensory experience, we could never get anywhere in the process of forming an opinion, because we would have too many choices. There are an infinite number of distinct and incompatible hypotheses consistent with any body of data (1994, p.132)
This is because of what Quine called “underdetermination of theory by data” (1994, p.132), or the fact that, as humans, our empirical data will forever be limited. This means that we need to make judgements and use our evolved psychology, itself a form of partiality, in order to interpret and draw conclusions from limited data; impartiality as pure objectivity would prevent us from being able to do this.
Antony uses arguments from Quine to criticise the epistemic distinction between normative and descriptive claims (2000, p.108), providing further argument that an empirical conception of impartiality as value-free will not suffice as an epistemic ideal. Antony explains how epistemic values or the Quinean notion of ‘extra-empirical’ virtues are needed to guide scientific practice (Antony,
2016; p.176), and Anderson similarly discusses the need for values in interpreting evidence (Anderson, 1995a; 1995b; 2004). I will consider these arguments in greater detail in section three
when I consider the epistemic role of values in inquiry.
Finally, I suggest that Quine’s criticism of the analytic synthetic distinction(Rey, 2013) can be used to
criticise the very distinction between empiricist and rationalist forms of knowledge, or empirical knowledge and values and justifications of knowledge. As Antony explores, Quine’s and ‘web of belief’ and holistic approach (whereby facts can only be tested in conjunction with other facts as part of a theory) complicates the difficulty with locating bias, making it harder to locate bias specifically in experience and therefore in the forming of synthetic truths; how we rationalise our experiences and form our analytic truths will also be affected by bias. In light of this, simplistic conceptions of epistemic impartiality as purely empirical or purely rational will not suffice in explaining knowledge formation.
Clarification is therefore needed in order to understand exactly what impartiality as an epistemic ideal could amount to and how it could help us with understanding bias. Consideration of the usefulness of partiality shows that it cannot be the simple fact that bias makes us partial that makes bias undesirable; further explanation is needed for understanding what it is about particular instances of partiality that we refer to as bias that are problematic. Without appeal to impartiality however, we are faced with the difficulty of needing grounds on which to criticise problematic forms of bias. Ultimately we must either refine our notion of impartiality to include useful forms of
partiality, or provide an alternative account of what makes problematic bias distinct from useful partiality. This difficulty is summarised by the bias paradox, which I will now outline before considering how Quinean insights, and ideas such as rationality and truth, have been used to solve the paradox.
What is the bias paradox?
Given the difficulties with formulating a clear notion of impartiality that can act as an epistemic ideal, we are now left with the difficulty of distinguishing instances of partiality in general, such as Fricker’s stereotypes and credibility judgements, from problematic bias, such as prejudice and distorted judgements. This problem was first discussed by Antony, who introduced the term ‘the bias paradox’. The ‘paradox’ comes from wanting to criticise bias for making us partial, whilst also wanting to criticise impartiality as an epistemic ideal. Antony explains that by rejecting impartiality we are faced with the difficulty of identifying what it is about bias that is problematic if it is not its partiality. To avoid the paradox, Antony explains that we must find a way to sort ‘the good bias from the bad’, and to achieve this further suggests that we “evaluate the overall theories in which the biases figure”
(1994, p.137) in order to determine which biases lead to truth; her response therefore distinguishes
‘bad biases’ through their unreliability, or the fact that they produce falsehoods (1994, p.102) Heikes summarises the bias paradox with the following argument.
(P1) The ideal of impartiality should be rejected
(P2) If we reject the ideal of impartiality, there can be no justified procedure for normatively distinguishing among competing epistemic views
(P3) If there can be no justified procedure for normatively distinguishing among competing epistemic views, then all accounts are epistemically equal
(C) All accounts are epistemically equal
(2004, p.320)
Heikes identifies the crux of the argument as P2. Given my previous discussion of arguments in support of P1, and given P2, P3 and the conclusion follows straightforwardly. It is clear that the conclusion of this argument is problematic; we do not consider all views to be epistemically equal, and have preferences over which theories ought to count as knowledge. This paradox has been discussed particularly within feminist philosophy because feminists have criticised the ideal of impartiality for falsely projecting men’s viewpoint onto the world as impartial, whilst also attacking sexist views for being biased.
Heikes discusses responses to the paradox that “reconstruct a notion of impartiality so that it avoids the problems of a traditional Cartesian conception” (2004, p.323). Antony’s response is to use truth instead of impartiality as a means of assessing biased accounts; she primarily addresses the falsity of biased accounts, rather than how bias can be unethical. As she explains, the reason that certain biased accounts are considered unethical in the first place is that are in fact false (1994, p.136). Although they do not explicitly discuss the bias paradox, both Fricker and Anderson use some form of epistemic reliability and empirical justification in their accounts. As mentioned previously, Fricker refers to evidence in her discussion of prejudice as an example of the epistemic significance of prejudice (2007, p.35). Anderson in particular discusses the role of reason in our knowledge-forming processes, and makes use of the notion of rationality as “social enterprise” (1995b, p.53) as a means
of assessing bias. I will discuss the social approach to understanding epistemic justification in the fourth section.
I will now assess Antony’s own solution to the bias paradox before discussing some of its benefits and shortcomings, and moving on to consider alternative solutions. In the next section I will highlight the significance of context in understanding truth and bias, and use arguments from Anderson which I believe clarify issues neglected by Antony’s focus on absolute, empirical truth.
Responding to the bias paradox
Antony’s naturalised approachAntony takes a Quinean naturalised approach to the bias paradox, endorsing an empirical assessment of biases to see which are conducive to truth (1994, p.137). Antony explains a naturalised approach to epistemology as “the empirical study of the actual processes – not ‘rational reconstructions’ of those processes – by which human cognizers achieve knowledge” (1994, p.119). Naturalised epistemology therefore draws on natural science, entailing that we “study knowledge by studying the knower” (1994, p.131), and that we do this empirically. As discussed, Antony draws on Quine’s holistic approach to explain that in the case of scientific inquiry, background assumptions, “inferential disposition[s]” (2016, p.161), and epistemic values play a role in interpreting empirical evidence – something a naturalised approach can show us – and that bias or partiality can be understood as these background assumptions which guide theory choice.
By understanding bias in this way, Antony explains that we can “evaluate the overall theories in which the biases figure” to determine which biases are truth conducive (1994, p.137). The problem with bias therefore is not its partiality, but rather that some forms of bias or partiality can lead us to false theories. Impartiality is not an epistemic ideal, rather truth is, and sorting the “the good bias from the bad” is a matter of empirical investigation; ‘good bias’ being one that brings us closer to truth, and ‘bad bias’ one that hinders access to truth or produces ‘oppressive falsehoods’ (1994,
p.141). Antony’s response to the bias paradox is therefore to reject P2 and shift the focus from
assessing the bias itself to assessing the truth or falsity of a theory the bias figures in using empirical methods. In other words, empirical assessment as a guide to truth provides a ‘justified procedure for normatively distinguishing among competing epistemic views’.
Antony explains that her focus on truth also aids ethical concerns, since when it comes to oppressive or dominant groups in society, “the real problem with the ruling-class worldview is not that it is biased; it’s that it is false” (1994, p.135). Similarly, “The Dragnet theory [or the particular view of knowledge criticised as oppressive] is not false because it’s pernicious; it’s pernicious because it is false” (1994, p.136). Notice that this ordering of the judgement – a view is problematic because it is false, not false because it is problematic – implies a notion of truth as the ultimate standard of assessment, both epistemically and ethically, applying to both normative and descriptive claims. How can we assess the truth of a theory? Considering the realisation that science is not value-free, and that “theory infects observation” so that confirmation of a theory through evidence “is a multidirectional relation” (1994, p.137), recognising a false theory resulting from bad bias is not simply a matter of checking whether it matches the evidence:
[theories based on bad biases] can’t be distinguished on the basis of their formal relation to the ‘facts’ because (1) there are no ‘facts’ in the requisite sense; and (2) there are too many
good biases whose relation to the data will appear as tenuous as those of the bad ones
(1994, p.137)
The upshot of this is that assessment of theories is not a straightforward matter of checking claims against empirical data. Combining this insight with Antony’s endorsement of empirical assessment as a guide to truth, we are left with the need for both a clearer understanding of how empirical assessment of bias is possible and a method for identifying the truth and falsity of a theory, particularly in the face of bias.
Antony addresses this difficulty, explaining that to empirically assess the truth of a theory we do not simply check, for example, its hypotheses against empirical evidence, but we must also subject the assumptions behind its aims to “empirical scrutiny” (1994, p.139), which will entail considering how well the background theory is supported, and comparison with other theories. Because ‘theory infects observation’ we must be sure our theoretical framework is also justified. For example, Antony attacks the use of race as a factor in psychological studies, asking “how could one defend the proposition that race and intelligence are connected without confronting the embarrassing fact that there’s no biologically defensible definition of race?” (1994, p.140). Discovering correlation between intelligence and race and drawing conclusions from the results will not leave us with true theory because race fails as a justified criterion for psychological study; there is not enough empirical data to support it, and the naturalised approach means we must justify our choice of categories and aims of inquiry empirically. The choice of race as a factor therefore counts as a bad bias. In this specific example, Antony relies on biological insights into race to highlight the problem with the psychological study, and later refers to “the wealth of competing explanations available at the social and personal/intentional levels” (1994, p.140) as an alternative to biological explanations. This suggests that that drawing on a wide range of empirical evidence and well-supported theories from different fields will give us reason to have faith in the truth of a theory, as well as enable us to ‘weed out’ the bad bias of a particular theory or inquirer.
I identify two remaining difficulties with Antony’s account. To summarise, firstly, given Antony’s support of Quine’s holistic approach or ‘web of belief’, criticism of distinctions between fact and value (1994, p.118), and the pervasive nature of bias, problems with relying on an ‘empirical scrutiny’ of bias remain. Specifically, I suggest we need further explanation of how empirical assessment can account for norms and values which feature as background assumptions and are not always strictly evidentially based. I also identify a circular aspect to using an empirical assessment of theories to determine bias, one which Antony appears to mediate with her notion of truth. Secondly, I argue that there are difficulties with Antony’s concept of truth. The use of empirical methods as a determiner of truth entails a specific conception of truth as being empirically attainable and absolute, something which Antony herself addresses (1994, p.102). I suggest that the use of absolute truth as a determiner of the nature of a bias is unsatisfying when we recognise the context-relative aspect of both bias and truth. Furthermore we must be able to recognise truth in order for Antony’s solution to be satisfying, and I suggest more clarity on both our notion of truth and what counts as empirical and epistemic justification is needed for this. I will now discuss how empirical assessment and truth in particular feature in Antony’s solution to the bias paradox in more detail.
Empirical assessment of bias and empirical truth
As discussed, Antony explains that we also need to subject aspects of our partiality, such as background assumptions and aims, to empirical scrutiny when assessing theories. This can ultimately lead to an empirical assessment of values; Antony states, “it becomes an open question whether divestiture of emotions, prior beliefs, and moral commitments, hinders, or aids, the development of
knowledge” (1994, p.130), a question which the naturalised approach urges us to answer empirically. It is therefore also “an empirical question whether or not ‘androcentrism’… leads to bad theories”, or theories that do not derive in the ‘proper way’ from the data due to bias (1994, p. 138). At the heart of judging epistemic and moral worth is empirical assessment (in line with my earlier point that Antony uses an absolute notion of truth which is equally applicable to both descriptive and normative claims). Antony defends her position, explaining that “the suggestion that there cannot be an empirical approach to a normative issue begs the central question against Quine, which is whether there is a sharp distinction between the descriptive and the normative” (2000, p.107); it is because of the break down of the epistemic distinction between background norms and empirical data that Antony can endorse an empirical assessment of background values and norms. It is not that we choose our epistemic norms, or ‘emotions, prior beliefs, and moral commitments’ which then determine our interpretation of evidence, but rather that we use a wide range of empirical evidence to support our choice of norms. It is after all empirical insight that Antony uses to reject the classical conception of impartiality as an epistemic norm.
Confusion arises when we consider Antony’s explanation of how bias in the form of background values affects our understanding of the evidence in a “multidirectional relation” (1994, p.137); Antony shows that all knowledge is partial (as previously discussed), meaning that how we interpret evidence will always depend on aspects of our partiality, or ‘good’ and ‘bad’ biases. This means that assessing a theory for bias will itself be a biased endeavour; an account can only be empirically identified as being a product of ‘good’ or ‘bad’ bias relative to further biases required to carry out an empirical assessment. Furthermore, partiality can include value judgements and perceptual distortions which are systematic, pervasive, and difficult to identify empirically. Implicit bias provides an example of bias that operates unconsciously and without our awareness. As well as the risk of our assessment being subject to such implicit, or distorting ‘bad bias’ (consider Fricker’s prejudice), there is a risk of circularity in Antony’s account when we realise that empirical assessment of our background values will themselves be necessarily influenced by some form of bias, making it difficult to identify ‘bad bias’ in particular.
Anderson also observes the difficulty of escaping bias in her discussion of naturalised feminist epistemology (Anderson, 1995b), stating that bias can “distort the context of justification itself”
(1995b, p.69), making an empirical assessment of our own bias difficult. She explains that “any
influence that biases the development of the field of alternatives will bias the evaluation of theories”
(1995b, p.69). In other words, it takes “the relative merits of alternative theories” (1995b, p.69) to
evaluate a theory. However, biases inherent in a theory often inhibit the development of alternatives, making it difficult to see past them. This means that the evaluation of our current theories will be subject the same biases. Anderson explains further that in the case of sexist bias in theories, “a theoretical approach may appear best justified not because it offers an adequate model of the world, but because androcentric ideologies have caused more thought and resources to be invested in it than in alternatives” (1995b, p.69). This is a conceptual point highlighting the difficulty in assessing pervasive bias; Anderson discusses examples of theories that suffer from androcentrism in particular, and explains how the androcentrism constitutes a widespread bias that prevents the development of alternative, non-androcentric theories, making it difficult to assess our current ones, even against a wide range of empirical data.
To illustrate, I will outline Anderson’s example of androcentrism in economics. Anderson explains that economic concepts, such as distinctions between “labor and leisure” and “public” and “private spheres” place significance on the labor that in a gendered society is typically carried out by men
data for macroeconomic theory” in a way that “excludes women’s gender-typical unpaid domestic labor from gross national product (GNP) calculations” (1995b, p.71). Because the “conceptual framework” is designed in such a way that defines economic productivity from a typically masculine perspective, Anderson explains that this may enable theorists to “declare as an objective fact that, say, women have little causal impact on the “economy” when all that is going on is that they have not taken any interest in women’s productive activities, and so have not categorized those activities as “economic.”” (1995b, p.75). In criticism of such a theory, Anderson states that androcentric bias makes the theory “empirically inadequate” (1995b, p.74), because it leads to “resulting policies [which] are usually sexist” (1995b, p.74). Antony would assess the situation with similar reference to empirical adequacy, claiming that androcentrism constitutes a bias in this case because it is empirically unjustified. However, what Anderson’s account makes clear is that ‘empirically adequate’ is not an absolute notion and can only be understood in the context of certain values and norms that specify what counts as adequate; specifically, such a theory can only be understood as inadequate under the assumption that to be empirically adequate economic theory must account for a full picture of society, including the lives of women and children. An alternative construal of the epistemic value of economics might make the theory perfectly empirically adequate as determined by the values, yet still exhibit androcentrism in its aims. In order to uncover such bias through empirical assessment, an alternative understanding of which evidence is relevant and our overall aims is needed, one which bias often inhibits.
How can we empirically assess the bias in theories in light of this circularity? This is significant if we wish to follow Antony in using an empirical assessment of theories to respond to the bias paradox. Without a clear understanding of how empirical assessment of bias and of epistemic norms and methodology works, we are yet again left facing the bias paradox.
Antony acknowledges this circular aspect to the Quinean naturalised approach, explaining that we will always need what Quine called ‘extra-empirical’ virtues to enable our empirical assessment:
Human theorists, [Quine] argued, choose hypotheses on the basis of a smallish set of “virtues;” we prefer simpler hypotheses to more complicated ones... and conservative hypotheses to ones that entail radical revision of our previous beliefs... we also prefer hypotheses that are empirically adequate and self-consistent (2016, p.176)
What distinguishes Quine’s naturalised approach is that rather than these virtues, or what I call epistemic norms, being posited as a justification of our theories, the virtues themselves are instead discovered empirically and therefore justified by the fact that they have led us to “empirical success”
(2016, p.177); Antony explains we must take a Humean stance1 towards justifying or vindicating such
virtues or epistemic norms, since they themselves cannot be empirically justified; it is rather the fact that they lead us to successful theories that justifies our reliance on them (1994, p. 118; 2016,
p.176-177). This highlights the importance of drawing on a range of supporting theories, as discussed
earlier with Antony’s example of the racist inquirer; by drawing on a variety of well-supported theories that also provide support for the theory in question, we can be justified in accepting the theory. Antony understands these extra-empirical virtues, which can be justified by their ‘success’, as forms of good bias. This again shows how she places success and truth at the centre of her account. Anderson also acknowledges a circularity in the Quinean approach, stating “nor does the prospect of circularity threaten the scientific validity of one’s reasoning, as long as the circle of reasoning is big enough” (1995b, p.78). This point is similar to Antony’s, since Anderson is suggesting that with
1 Hume famously recognised the problem of induction which necessarily could not be empirically justified. Instead, Hume used a pragmatic approach (Vickers, 2014)
support from enough theories and co-supporting claims, the ‘web of belief’ is made coherent. She concludes, “theories that incorporate value judgements can be scientifically sound as long as they are empirically adequate” (1995b, p.78). However, Anderson gives an understanding of what ‘empirically adequate’ amounts to in a way that clarifies confusions neglected by Antony; instead of understanding this notion as linked to truth, Anderson is clearer in discussing the role of non-empirical values in understanding the very notion of non-empirical assessment and what counts as success. I suggest that the idea of ‘empirically adequate’ as determined by evidence, referenced by Anderson here as a justification for value judgements, is to be understood as a minimum condition for justification; this means that in order to be epistemically considerable, empirical adequacy or ‘accounting for the facts’ is the minimum condition such a theory must meet. Anderson talks specifically about the role played by non-empirical, contextual normative values in her description of a “dual-track model of theoretical justification” (Anderson, 1995a), stating that evidential support for a theory must be sufficient, but that “contextual values do provide grounds for preferring theories that leave open the possibility of representing certain claims as true” (1995a, p.54-55).
Anderson recognises that ‘successful’ can be understood in varying ways and does not in itself provide a way of evaluating our epistemic practise and of escaping circularity; to illustrate the varying understandings of ‘success’ she gives an example of doctors designing drugs to control “premenstrual syndrome”, instead of considering the possibility that stress related to menstruation could be relieved by changes in societal attitudes to women’s menstrual cycles, or “claims on others to change their behaviour” (1995b, p.76). Even if the drugs are empirically “effective in controlling the phenomena of women’s hormonal cycles”, such technology can only be understood as ‘successful’ under a view which “does not consider whose power is enhanced by the theory and whose interests are served by it” (1995b, p.76). Whether or not considerations of power are relevant to such an investigation is a question of aims, not empirical adequacy in explaining data. Hence, “rival interpretations of the other epistemic values also depend on contested nonepistemic values”
(1995b, p.79), and moral and political considerations can also be relevant in judging the empirical
success and adequacy of a theory. Unlike Antony suggests, empirical adequacy and success cannot be considered in absolute terms and therefore used to deal with the circularity in assessing theories for bias.
Given Antony’s criticism of epistemic distinctions between normative and descriptive claims, she must also allow for normative claims including moral and political values to function epistemically as part of a theory and be subject to scrutiny. However, I argue that Antony’s explanation of scrutiny is reliant on absolute understandings of empirical adequacy, truth, and success. If moral or political values are of epistemic significance, they must themselves be subject to empirical scrutiny; according to Antony we presume our notion of empirical truth, and it is up to investigation to determine the instrumental role played by the moral and political values in discovering this truth.
This leads me to my second difficulty, namely the notion of truth. Antony initially states she endorses a “realist conception of truth, viz., one that allows a conceptual gap between epistemology and metaphysics, between the world as we see it and the world as it is” (1994, p.102). This is important, because realist truth is necessary for understanding Antony’s emphasis on empirical assessment as a guide to truth. Understanding truth as objective and determined by ‘the world as it is’, and not subjective or relative, supports the idea of empirical assessment as a guide to truth. Antony relates truth to notions of justification, specifically empirical justification, and epistemic success. She therefore uses truth as a “master value” (2000, p.114) to assess both the hypotheses of theories and their background assumptions, values or norms. Doing this requires us to be able to – using empirical methods in line with a naturalised approach - recognise truth in a theory, and what makes for
successful epistemic practise; Antony explains, “naturalised epistemology is concerned with explaining the success of human epistemic practise – but only to the extent that it actually is successful” (2000, p.110). Recognising successful practise requires that we take some knowledge for granted, since “there is no presuppositionless position from which to assess epistemic practice... The only thing to do, then, is to begin with whatever it is we think we know, and try to figure out how we came to know it” (1994, p.131). By determining ‘how we came to know’ we will uncover underlying empirical assumptions, thereby providing what Antony calls a ‘vindication’ of these assumptions. In the context of understanding bias however, and considering the role played by values, taking knowledge for granted becomes a very delicate process.
It appears that there is a feedback loop between bias and truth; we assess our biases using truth, yet bias in part shapes our notion of truth. This is where Antony’s concept of truth as empirical and realist comes to the fore, since she relies on empirical methods as providing a guide to truth and so although our empirical methods are always themselves subject to bias, the end goal of truth will always be the same. Antony recognises the role played by the ‘extra-empirical’ principles that guide inquiry, and believes in an empirical justification of values and norms, however I suggest that she does not consider in enough detail what the consequences are of this for her epistemic norm of truth.
In defence of an empirical assessment of bias, Antony states “there are no limiting cases of purely normative claims that make no descriptive demands on the world at all” (2000, p. 108). Anderson similarly recognises the centrality of empirical support for theories, however she also recognises the equally important role of “empirical claims couched in unreduced social, intentional, and evaluative vocabularies”, and that “it is important to note that modest empiricism is not committed to eliminating such claims from scientific theories” (1995b, p.52). Hence, she states that “evidential and normative considerations cooperate; neither usurps the role of the other” (1995b, p.54). By giving greater acknowledgment to the role of normative arguments – a role which Antony reduces to empirical adequacy, hence it is simply “an empirical question” whether or not androcentrism is a ‘bad bias’ (1994, p.138) – along with a contextual understanding of truth and empirical adequacy (which I will discuss next), I argue that Anderson can give a more complex analysis of the situation. Our methods of justification must go beyond what is empirically, and empirical adequacy and justification are normative notions. Furthermore, what counts as empirically adequate is context-dependant, meaning that absolute notions of truth and justification will not suffice.
Ultimately, Anderson explains that:
naturalized feminist epistemology... does not provide methodological arguments against the pursuit of sexist theories. It does claim, however, that it is irrational for theorists to pursue sexist research programs if they do not endorse sexist values. Moral and political arguments about the rationality of particular values may therefore have a bearing on the rationality of pursuing particular research programs (1995b, p.79)
Recognition of the role played by values in understanding what is empirical and in justifying theories can help us overcome the circularity mentioned earlier; in assessing our theories for bias, by directly assessing our background epistemic, moral and political values and arguments we can avoid the trap of further empirical assessment being subject to the same bias, or the same understanding of the empirical given our background non-empirical assumptions. I therefore agree with Anderson in that “the epistemic evaluation of theories therefore cannot be sharply separated from the interests their applications serve” (1995b, p.77). Moral and political values cannot be extricated from epistemic
ones when the topic in question is of moral or political significance, meaning that purely empirical assessment may be insufficient for assessing values and bias in these contexts.
The difficulties I have hinted at in this discussion can be summarised as follows. Antony uses truth as an end goal which justifies the use of any norms or values which lead us there. This requires us to be able to recognise truth in the face of bias and with reliance on empirical methods. The difficulty with this picture is that we cannot justify our norms and biases through pointing to our own success in using them because this success is in part dependent on our acceptance of these norms, something which will always go beyond the scope of what empirical justification can provide; hence Quine’s ‘extra-empirical’ virtues or norms. We can attempt to resolve this difficulty through greater consideration of how an empirical assessment of our background values or biases can be used justify them, however I suggest there remain consequences for our conception of truth and epistemic justification which need spelling out. Anderson’s discussion highlights the role played by values and how these can be influenced by moral and political as well as epistemic values, affecting our notions of epistemic success and empirical adequacy. I suggest that recognition of the significance of these values can provides a way of understanding ‘empirical adequacy’ and truth as normative and context-dependant, where the context includes values, a view supported by Anderson.
I now turn to Anderson’s view for further discussion of what makes a theory true and how we can understand truth as normative and context-sensitive yet still answerable to some notion of objectivity; ultimately Anderson seeks to achieve this through an understanding of epistemic inquiry as social and an appeal to rationality as an epistemic ideal. I will conclude from Anderson’s discussion that truth is context-relative and that the concept of bias can only be understood relative to this context.
Anderson on truth as contextual
Like Antony, Anderson also takes herself to be endorsing a naturalised approach to knowledge, meaning that she places empirical investigation at the heart of inquiry and supports breaking the epistemic distinction between the normative and the descriptive; she states that we must carry out “empirical inquiry” which “opens us up to the possibility of finding out that our value judgments were mistaken. The fact that we can do this shows that factual and value judgements do not occupy separate spheres” (2004, p.22). She does not explicitly address the bias paradox itself, however she does discuss the concept of bias and its role in inquiry. There are some differences between Antony’s and Anderson’s approaches, one of these being that Anderson’s discussion of truth is contextualist, and she does not use empirical truth as a sole means for assessing bias. Anderson is more explicit about how a “bidirectional influence of facts and values” (2004, p.11) works, and although empirical justification plays a central role in her account, she allows for moral and political arguments to have epistemic significance. Her understanding of bias is also contextualist, and she uses concepts such as ‘dogmatic’ and ‘fruitful’ rather than true or false and, as previously discussed, allows for moral judgements and values to play a role in determining the context.
Anderson discusses truth in a way that addresses issues neglected by Antony, explaining that we can understand certain theories as being problematically biased yet true in virtue of being empirically supported by evidence. Anderson defends her point by distinguishing between true facts and ‘whole truth’, explaining that “the significance of most truths can be adequately grasped only in the context of the whole truth” (1995a, p.37-38). She explains that although theories can be based on facts and count as empirically true, what counts as ‘the whole truth’ is not a strictly empirical matter and is instead determined by context and the aims of inquiry (1995a, p.40). Truth is therefore context sensitive. Anderson uses an historical case as an example, which I will now discuss and use to
illustrate the differences between Antony and Anderson’s approaches. Overall I believe Anderson gives a more in-depth understanding of truth.
Anderson discusses the Atlantic slave trade during the 17th century and The Secret Relationship
between Blacks and Jews, a book published in 1991 by the Nation of Islam (Anderson, 1995a: p.38),
as a specific example of the distinction between whole truth and fact. What is significant about this book is that it uses factual evidence to support what is overall a biased and “[un]acceptable account of the role of Jews in the Atlantic slave system” (1995a, p.38). What makes the account unacceptable and biased is not simply its being false; if all of the facts used in the book as evidence are true, how can we complain that the account is false? Instead, Anderson points out that the epistemic problem with the account is rather that it ignores additional facts and context that are needed to understand the whole picture. The account is not simply false, but it is not the whole truth. Because of this, Anderson explains the account is biased – an unbiased theory is “one that does justice to the whole truth” (1995a, p.39).
How do we decide which facts are relevant to accessing the ‘whole truth’? Anderson points out that we must use judgement in selecting which facts are relevant; it is not that every fact in relation to a topic is significant in understanding the whole truth. To use Anderson’s example, not all facts relating to the Atlantic slave trade are significant to understanding it – we would under no context find answers to such questions as ‘how many waves did each slave ship surmount?’ and ‘how many times did each slave blink?’ illuminating for portraying the events of the slave system (1995a, p.39). The ‘whole truth’ relevant to a theory is therefore a matter of context, and the aims of inquiry guide which facts are relevant. Mere empirical truth is not sufficient justification; we must also consider significance, the aims of inquiry, and what counts as ‘whole truth’.
So, what is it about The Secret Relationship that ignores the whole truth? Given that “what counts as a significant truth, and the whole truth, can only be judged in relation to.. interests” (1995a, p.40), and given that the interests of the book are to explore the role of Jews in the Atlantic slave trade, what gives us ground to reject the theory presented in the book as inadequate? Anderson explains that in order to adequately assess an account we must look closer at its aims. Although the book might at first seem to be motivated by answering questions such as ‘what was the role of the Jews in the Atlantic slave system?’ or considering how this role compares with the roles of other ethnic groups, Anderson explains that these questions by themselves would not be enough to specify which roles and which comparisons are of interest. She explains, “the question that The Secret Relationship implicitly purports to answer is rather ‘Do Jews deserve special moral opprobrium or blame for their roles in the Atlantic slave system or bear special moral responsibility for that system’s operations?’”
(1995a, p.40). Anderson explains that in relation to this question, the account ignores the wider
context of historical facts about who was responsible for the slave trade that are relevant to understanding why the given account is unacceptable. This is what makes the account problematically biased as opposed to merely partial. Anderson summarises the situation by stating that the book “offers a biased account with respect to this question, because it ignores facts morally relevant to answering it” (1995a, p.40), facts such as the comparative role played by non-Jewish, Anglo-American and European land owners who benefitted from and perpetuated the slave trade. Furthermore, Anderson believes that the ‘background value judgements’ that determine which facts are significant can themselves be unjustified. The Secret Relationship is not only biased because it ignores the whole truth relevant to its aims, it is also unjustified in its aims; Anderson states that there is no more justification for “singling out Jews as a comparison class in such studies, rather than say, the class of people who have drooping eyelids” (1995a, p.40) and that it is “it is a moral mistake
to pass judgements of collective guilt or merit on whole ethnic groups” (1995a, p.40), making the choice of ‘Jewishness’ as a relevant category for investigation ethically unjustified. This choice of criterion relies on morally problematic and anti-Semitic value judgements, or views that take ethnicity or race as significant to understanding an individual’s actions, desires, and morality.
Clearly, both ethical and empirical concerns are at stake here. Like Antony, Anderson emphasises a need for justification of the aims and values of an inquiry, however, unlike Antony, Anderson does not rely on the ‘master value’ of truth as ultimate means of assessment. Concerning the above example, Antony would point to empirical reasons for the choice of ethnicity being unjustified; given her observation that “there is no empirical grounding for the selection of race as a theoretical parameter in the study of intelligence” (1994, p.140), similarly, Antony would reject The Secret
Relationship by pointing out that there is no empirical grounding for selecting Jewish people as a
group for inquiry (indeed, why not instead people with drooping eyelids?). The author’s motivation to do so can only be anti-Semitic, rather than relying on adequate empirical support. She must conclude that the account is biased because its background assumptions lead us to falsehoods. According to Antony, our aims being empirically justified ensures they will also be ethical; she need not characterise the use of Jewish ethnicity in this study as a ‘moral mistake’. In relation to the use of ethnicity, namely ‘Jewish’, as a category relevant to investigations into blameworthiness, Anderson would point out that it is not simply the empirical inadequacy of this classification that makes the choice unjustified, but that the moral component of this mistake is significant. A focus on empirical adequacy is only considered a possibility because it “reflects the supervenience of moral judgments on factual judgments” (1995a, p.41).
If moral judgements rely on empirical data for their support, as both authors allow, what then are the benefits of shifting the focus away from a purely empirical assessment, and highlighting what Anderson calls specifically the “bidirectional influence of facts and values” (2004, p.11)? What is the epistemic benefit of an assessment of moral and political values inherent in a theory? One reason for understanding the epistemic significance of the aims of The Secret Relationship being morally unjustified and not just empirically inadequate is that the notion of ‘empirical adequacy’ is itself subject to value judgements, which are in this case explicitly moral. Anderson states, “what makes a given factual criterion relevant to judging a theory’s... significance is its bearing upon the answer to the contextually value-laden question that motivates the inquiry, and whether it has such a bearing is itself determined by contextual values” (1995a, p.41). To illustrate, what makes empirical facts concerning the non-essential nature of ethnicity and race relevant in this case, and the inappropriateness of a category such as ‘Jewish’ in considering concepts such as blame, is the fact that the inquiry is motivated by moral questions concerning Jewish people and blame for ‘their role’ in the Trans-Atlantic slave trade. Underlying this question is essentially a moral value judgement – ‘Jewish people are blameworthy’ – and so any evidence used to counter this claim will rely on an understanding of the moral workings and significance of concepts such as ‘blame’ and therefore what would need to hold to make ‘Jewish’ a relevant category in this case.
Overall, to understand the issue as a purely empirical one will leave us without the resources to understand which particular facts are relevant in responding to the question (namely, ones that counter the moral that an ethnic group can be ‘blameworthy’), and why in this instance the use of ‘Jewish’ as a category is inappropriate. Empirical data must be central to any account of justification, however , as Antony points out, “it takes an extreme form of empiricism to believe that brute correlations between one arbitrarily selected characteristic another constitutes science” (1994,
