An argument for egalitarian confirmation bias and against political diversity in academia

It has recently been suggested that politically motivated cognition leads progressive individuals (e.g., liberals) to form beliefs that underestimate real differences between social groups (e.g., in academic performance) and to process information selectively to support these beliefs and an egalitarian outlook. I contend that this tendency, which I shall call ‘egalitarian confirmation bias’, is often ‘Mandevillian’ in nature. That is, while it is epistemically problematic in one’s own cognition, it often has effects that significantly improve other people’s truth tracking, especially that of stigmatized individuals in academia. Due to its Mandevillian character, egalitarian confirmation bias isn’t only epistemically but also ethically beneficial, as it helps decrease social injustice. Moreover, since egalitarian confirmation bias has Mandevillian effects especially in academia, and since progressives are particularly likely to display the bias, there is an epistemic reason for maintaining (rather than counteracting) the often-noted political majority of progressives in academia. That is, while many researchers hold that diversity in academia is epistemically beneficial because it helps reduce bias, I argue that precisely because political diversity would help reduce egalitarian confirmation bias, it would in fact in one important sense be epistemically costly.


Introduction
"Don't fight forces, use them." (Richard Buckminster Fuller) 1 Many studies have shown that people frequently engage in politically motivated cognition: we often tend to conform our assessments and beliefs about information to our political goals and political identity rather than to accuracy (Kunda 1990;Taber and Lodge 2006;Nisbet et al. 2015;Kahan 2016). Politically motivated cognition, which is taken to be less than fully epistemically rational because it involves a diminished sensitivity to facts (Huemer 2016;Ringel et al. 2019), has recently become a hot topic in philosophy (Gerken 2019;Peters 2019;Talisse 2019;Carter and McKenna 2020). The present paper offers a social epistemological investigation of a specific type of politically motivated cognition.
To hone in on it, notice first that politically motivated cognition is common among people from across the political spectrum. Both left-leaning individuals, i.e., subjects committed to values and goals such as social equality and social justice (henceforth progressives), and right-leaning individuals (e.g., conservatives) have been found to display it (Nisbet et al. 2015;Ditto et al. 2019;Clark et al. 2019;Gampa et al. 2019). In the following, the focus will be on politically motivated cognition that has been ascribed to progressives, in particular.
Specifically, several philosophers and psychologists have suggested that since progressives value social equality and social justice, this might incline them to form beliefs that underestimate, deny, or overlook real differences between individuals and groups when it comes to positive features, for instance, academic performance (as measured by standardized tests) (Kekes 2007;Jussim 2012, p. 398f;Clark et al. 2019;Peters 2019, p. 408;Anomaly and Winegard 2019). Empirical data are sometimes cited to support this claim. For instance, in one study, independently of intelligence levels, the more progressive a participant 2 (N 94) was, the higher their underestimation of the real variance between different ethnic groups (e.g., Jewish, Portuguese, Black, and British students) in academic performance (i.e., high-school grades) (Ashton and Esses 1999, p. 233). Relatedly, Anomaly and Winegard (2019, p. 9) report studies in which progressives rated evidence of group differences on, for instance, IQ and college entrance exams as much less credible when groups perceived as more powerful (i.e., Whites and men) were said to outperform groups perceived as less powerful (i.e., Blacks and women) than when the latter groups were said to outperform the former.
While I shall remain cautious about these data, 3 they cohere with many other studies on politically motivated cognition that found that one's political values can significantly shape one's beliefs and information processing (Taber and Lodge 2006;Nisbet et al. 2015;Kahan et al. 2017;Lewandowsky and Oberauer 2016;Washburn and Skitka 2018). It is thus worth taking seriously the possibility that when thinking about group differences, progressives too might be motivated to process information in a way that doesn't have accuracy as its goal but, for instance, social equality, leading to effects of the type just mentioned.
Suppose, then, that progressives do have a politically motivated tendency to (i) believe that differences between individuals and groups in socially valued traits (e.g., academic performance, intelligence, etc.) are less pronounced than they really are and to (ii) selectively process information so that it supports these beliefs and a politically progressive outlook (e.g., by depicting low status, stigmatized individuals more positively than high status, powerful individuals). 4 Since this tendency is by assumption driven by goals extrinsic to accuracy such as equality and social justice, it has been called "egalitarian denial", "motivated egalitarianism" (Jussim et al. 2016, p. 56), and "liberal bias" (Anomaly and Winegard 2019, p. 9). I shall call it egalitarian confirmation bias (henceforth ECB).
Notice that while the term 'bias' has a built-in negative connotation, when I use it here, this is only to indicate that the cognitive tendency at issue is driven by a goal other than accuracy. The term isn't meant to suggest that the tendency is unjustified (some biases might be epistemically and/or ethically justified).
ECB raises interesting philosophical questions. This is partly because there are at least two prima facie compelling rationales to counteract the bias: The Epistemic Rationale. Since (by assumption) ECB contributes to a kind of motivated ignorance of socially relevant facts (e.g., other people's differences in academic performance), the bias seems epistemically problematic per se: it undermines subjects' ability to track relevant truths. The Ethical Rationale. Since (by assumption) ECB contributes to a kind of motivated ignorance of people's differences in, e.g., academic performance, and subjects need to know 5 about these differences to be able to tackle undeserved or prejudice-based social inequalities related to them, the bias seems also ethically problematic per se: it undermines subjects' ability to counteract social injustice.
While these two rationales 6 suggest that we should counteract ECB, some philosophers have argued that, for instance, ignorance can in some cases, despite being epistemically detrimental, also have significant benefits (Pritchard 2016;Fricker 2016). Might the same apply to ECB? I shall develop an answer that will be relevant not only for the social epistemological analysis of ECB, but also for a broader and initially seemingly unrelated topic in the social sciences, meta-philosophy, and philosophy of science, namely the debate on diversity in academia (Intemann 2011;Duarte et al. 2015;Stanley 2016;Rolin 2017;Peters et al. 2020;Haidt 2020). In this debate, both philosophers and psychologists tend 4 Notice that (ii) might interact with (i) such that progressives' political viewpoint will not always lead to underestimation of real differences, but may also (due to cross-ideological hostility) incline them to "exaggeration of differences between their own and opposing political groups" even in contexts where these opposing groups (e.g., conservatives) are low-status and stigmatized (Jussim et al. 2016, p. 57). 5 I'm using the term 'know' loosely here; justified belief or warrant might be enough. 6 Versions of these rationales can be found in other recent philosophical and psychological debates (albeit with no reference to ECB) (e.g., Madva 2016, p. 193f, 200;Puddifoot 2017, p. 88f;Haidt 2020, p. 54). to accept that diversity in academia is highly epistemically beneficial because it helps counteract biases by increasing the range of potential social criticism of claims made in academia (Peters et al. 2020, p. 536f). For instance, Longino (2002, p. 134) writes: "A diversity of perspectives" is generally epistemically positive as it "assures that theories and hypotheses accepted in the community will not incorporate the idiosyncratic biases (heuristic or social) of an individual or subgroup". Similarly, Intemann (2017, p. 139) notes: "having diversity of values and interests is epistemically beneficial in terms of minimizing bias".
Here I shall motivate a qualification of this line of thought. I contend that because increasing political diversity in academia would reduce biases, specifically ECB, it in fact results in epistemic costs. This is because what seems like an epistemic vice among progressives, i.e., ECB, is in fact at least in one important hitherto unappreciated sense epistemically positive. More specifically, I shall argue that ECB is often (borrowing a term from Morton 2014) "Mandevillian" in nature: while it is epistemically problematic in one's own cognition, it frequently has effects that significantly improve other people's truth tracking, especially that of stigmatized individuals in education in academia. Due to its Mandevillian character, ECB isn't only epistemically but also ethically beneficial, as it helps tackle unjust social inequality. Furthermore, since ECB has Mandevillian effects especially in academia, and since progressives are particularly likely to display the bias, there is an epistemic reason for maintaining (rather than counteracting) the often-noted political majority of progressives in academia (Inbar and Lammers 2012;Honeycutt and Freberg 2017;Peters et al. 2020). That is, while many researchers hold that diversity in academia is epistemically beneficial because it helps reduce bias, I argue that precisely because political diversity would help reduce ECB, it would in fact in one important sense be epistemically costly.
Three clarifications. First, I have elsewhere made a case for a political diversification of academia (Peters 2019;Peters et al. 2020) and stand by it. The argument against political diversity that I shall develop here is compatible with this earlier case in favor of it. For instance, as will become clear below, the point here focuses specifically only on a diversification of the educational aspect of academia, not on the research aspect. Moreover, notice that highlighting epistemic benefits of ECB and a lack of political diversity in academia is compatible with the view that accuracy about group differences and political diversity in academia are overall much more epistemically and/or ethically beneficial. The argument proposed here is meant to provide one significant so far overlooked epistemic reason in favor of ECB and against politically diversifying academia, not an all-things-considered verdict on whether both are preferable to accuracy and political diversity, respectively.
Second, the argument against political diversity in academia below is not against diversifying academia by introducing more left-leaning progressives than it currently contains (e.g., increasing the number of feminists, Marxists, etc.). Rather, it is directed against political diversity that would reduce the overall number of progressive vs. non-progressive individuals in academia.
Finally, in investigating whether and if so how ECB can be epistemically beneficial, one needs to make assumptions about what 'epistemic benefits' are, and this will depend on one's preferred epistemological commitments (e.g., whether one is an epistemological deontologist, virtue epistemologist, or epistemic consequentialist). I shall assume a veritistic type of epistemic consequentialism introduced by Goldman (1999). It holds that a cognition C is epistemically beneficial for a subject S (or group G) if C maximizes the acquisition and/or retention of true beliefs relevant to S (or G). 7 In Sect. 2, I outline the overall argument of this paper. In Sects. 3-5, I support each of its premises inter alia by introducing experimental data on a psychological phenomenon known as a "self-fulfilling prophecy" (Biggs 2009) and relating these data to ECB. In Sect. 6, I consider and rebut three objections to the argument developed in the previous sections. Section 7 briefly summarizes and concludes the discussion.

The main argument
The argument I shall develop consists of four sub-arguments each involving modus ponens. They take the following form ('P' premise; 'C' conclusion): I will now support Sub-arguments I-IV .

Sub-argument I
In this section, I will motivate the following: (P1): Overall, academia mostly contains progressive faculty. (P2): If, overall, academia mostly contains progressive faculty, then ECB is currently more common among faculty in academia than if academia contained more politically diverse faculty. (C1): ECB is currently more common among faculty in academia than if academia contained more politically diverse faculty.
As for (P1), in line with the Oxford Dictionary of English, I take the term 'academia' to refer to the "world of learning, teaching, and research at [colleges or] universities". 8 Similarily, I take the term 'faculty' to refer to "all the teachers in […] a college or university", 9 where these teachers (lecturers, professors, etc.) might also engage in research. Several surveys suggest that the social sciences, arts, and humanities currently contain a vast majority of progressives among faculty (Yancey 2011;Inbar and Lammers 2012;Honeycutt and Freberg 2017;Peters et al. 2020). And while there might be differences between fields, countries, and disciplines, in the empirical research on political diversity in academia, there is little doubt that, overall, academia does at the moment mostly contain progressive faculty (Gross and  Turning to (P2), to be sure, progressives are a politically diverse group of subjects, including individuals who are somewhat left leaning (e.g., liberals) as well as very leftleaning subjects (e.g., communists). Moreover, among progressives, value conflicts exist (think, e.g., of feminist critiques of liberalism). Still, despite these differences, progressives tend to share a commitment to egalitarianism, social justice, and fairness (Janoff-Bulman 2009; Arneson 2015). And, as noted above, this is likely to influence their social cognition in the way ECB does (Ashton and Esses 1999;Nisbet et al. 2015;Clark et al. 2019). This isn't to deny that other subjects than progressives might also display tendencies such as ECB. The point is just that ECB is likely to be much less common among non-progressives because non-progressives (e.g., conservatives) tend to be more critical of equality (Anderson 1999;Janoff-Bulman 2009). And this should make them less likely to be inclined to conform their information processing and belief formation to goals such as equality, social justice, etc. (Sterling et al. 2019;Gampa et al. 2019). Relatedly, studies suggest that non-progressives are indeed less susceptible to an underrepresentation of group differences 10 .
By extension, it is plausible that if academia contains now overall predominantly progressive faculty, then ECB should currently be more common among faculty in academia than if academia contained more politically diverse (read: less progressive) faculty. After all, viewpoint diversity in groups tends to facilitate social criticism that then in turn helps 'cancel out' individual group members' biases (Longino 2002, p. 131f;Intemann 2011, p. 115f). With an increase in political diversity among faculty, which here and below specifically only means an increase in non-progressive individuals, ECB should thus decrease in academia. This supports (P2) and together with (P1) leads to (C1): ECB is currently more common among faculty in academia than if academia contained more politically diverse faculty.

Sub-argument II
To recap, Sub-argument II was the following: Supporting (P3) and (P4) requires a brief discussion of psychological research on a phenomenon known in the social sciences as a 'self-fulfilling prophecy' (SFP). It occurs when initially false beliefs about subjects or social states of affairs can exert effects on cognition and behavior that contribute to gradually bringing about the truth of these beliefs (Merton 1948;Biggs 2009). In the next sub-section, building on earlier work (Peters 2020), I provide examples of and review evidence pertaining to SFPs. In Sect. 4.2, I will then use findings from studies on SFPs to support (P3) and (P4).

SFPs and self-fulfilling effects: examples and empirical data
To illustrate SFPs, three examples may help (see also Merton 1948;Velleman 2000, p. 358f;Mameli 2001, p. 610f;Peters 2020, p. 8f): (i) A solvent bank is publicly prophesized to be insolvent. This leads customers to withdraw their money from the bank, resulting in the bank becoming insolvent. (ii) Jack inaccurately believes that Jill is angry with him. This leads him to act in a hostile way toward her, which makes her angry with him. (iii) A teacher inaccurately believes that a student's academic performance is above average and so gives her challenging material, encourages her, and communicates high expectations. Because of that, the student increases her efforts, resulting in her gradually performing above average.
SFPs of the kind captured in (iii) will play a key role in my argument. Since that is so, what support is there for believing that effects of that type are real?
Consider a naturalistic experiment conducted by Madon et al. (1997). Madon et al. investigated SFPs in educational settings by tracking maths performance in US students transitioning from 6th to 7th grade. They first assessed the students (N 1539) via a standardized maths test taken by the beginning of the 6th grade. Afterwards, teachers (N 98) were asked to fill out questionnaires about (inter alia) the (i) performance, (ii) academic aptitude, and (iii) effort of each student in their math class. The questionnaire tracked teachers' beliefs about the students' current academic standing. Many teachers held false beliefs about their students in that they over-or under-estimated them. To identify teacher over-and underestimates, Madon et al. regressed the three teacher perception/belief variables (i.e., (i)-(iii)) on the student background variables (i.e., previous grades, test scores, motivation etc.). Teachers' over-and underestimates indicated the extent to which teachers over-or underestimated students compared to what was suggested and predicted by each student's own background variables. Madon et al. then used the teachers' estimates to measure SFPs by relating these estimates to each student's future achievement, tracked (a year later) via another standardized test taken by all 7th grade students. Interestingly, they found pronounced SFPs; indeed, "one of the strongest self-fulfilling prophecy effects ever obtained in naturalistic research […] occurred when teachers substantially overestimated [previously] low achievers" (Madon et al. 1997, p. 805).
These findings cohere with the results of studies by, for instance, Jussim et al. (1996) in which a range of powerful SFPs emerged especially with stigmatized students such as African-American students, students from lower socio-economic status backgrounds, and students with histories of low prior achievement. More specifically, in some of these cases, SFPs were comparable to changing a student's SAT score from 450 to 550 (Jussim 2012, p. 240). Or, as Biggs (2009, p. 11) calculated (on the basis of the mentioned data), especially in African-American students: "moving from the lowest to the highest expectation is predicted to raise the student's grade from C to B+".
The results of Madon et al.'s (1997), and Jussim et al.'s (1996) studies aren't flukes. For instance, Hinnant et al. (2009), too, found that positive SFPs of overestimations were strongest among children with low-income background (in math), and minority boys (in reading). SFPs have also been found in various other contexts (Peters 2020), albeit often with attenuated effect sizes (Jussim 2012).
Why do SFPs occur? A common explanation is that when individuals (perceivers) hold beliefs about other people (targets), these beliefs produce certain expectations in the perceivers about the target. The expectations at issue then influence the perceivers' interactions with the targets in ways that elicit from the targets behavior that is consistent with the expectations (Snyder and Klein 2007). Targets tend to respond in this conformist way partly because people have a general inclination to fulfil others' expectations out of an interest in coordinating and getting along with each other (Bacharach et al. 2007).
This process could be highly pernicious when the beliefs at the center of the SFPs are negative, capturing racist or sexist stereotypes, anxious expectations, fear, or hostility (Langton 2009, p. 310;Fine 2012). However, research comparing the strength of SFPs of negative vs. positive expectations suggests that in some contexts there is an asymmetry in the effects. For instance, Madon et al. (1997, p. 806) found that teacher overestimates "increase[d] achievement more than teacher underestimates tended to decrease achievement among students". 11 One explanation of this asymmetry is that when students with negative educational experiences find themselves faced with a supportive, encouraging teacher who also insists on high performance, they might feel as if they have caught a breath of fresh air and be inspired to work especially hard (ibid, p. 793). Relatedly, SFPs crucially depend on whether targets accept their perceivers' expectations. And since subjects usually strive to think well of themselves, it is perhaps to be expected that in order to protect their self-concept, people will respond asymmetrically to positive vs. negative perceiver expectations (Peters 2020).
However, even when it comes to positive expectations, SFPs are neither all-powerful nor inevitable (Felin and Foss 2009). They occur, for instance, only when targets are able to change in accordance with the trait ascriptions, and when the latter are believable rather than unrealistic (Snyder and Klein 2007). Moreover, the effect sizes of SFPs in experimental settings are often modest (Jussim 2012).
Still, incremental effects in some contexts might accumulate across contexts (particularly cooperative ones; Peters 2020). Indeed, while most studies that found only relatively weak SFP effect sizes have so far focused just on dyadic social interactions (i.e., between one target and perceiver) (Madon et al. 2018), recent research on SFPs in contexts in which targets keep encountering the same social conception from different perceivers provides evidence of such an SFP-effect accumulation (ibid). In any case, notice that the specific kind of pronounced SFPs in teacher-student settings mentioned above have been found repeatedly in different studies (e.g., Madon et al. 1997, Hinnant et al. 2009). And even, for instance, Jussim (2012Jussim ( , 2017, one of the most ardent critics of SFP research, acknowledges that in studies he and his colleagues conducted specifically to see whether SFPs could ever have significant effects, [p]owerful self-fulfilling prophecies occurred among: (1) African-American students, (2) students from lower SES [i.e., socio-economic status] backgrounds (regardless of ethnicity), (3) students with histories of low prior achievement who were from lower SES backgrounds […] [, and] (4) students with histories of low achievement who were the target of high expectations. High expectations uplifted such students more than they uplifted high achievers, and more than low expectations harmed achievement. […] [SFPs] seemed to ameliorate more than cause social inequalities (uplifting students with histories of low achievement). (Jussim 2017, p. 8) 12 The points in this passage are all that I shall here assume about the efficacy of SFPs. That is, I take it that in educational settings in interactions between teachers and students belonging to stigmatized groups, SFPs are "real, reliable, and occasionally quite powerful" with an asymmetrical impact, i.e., overestimations tend to help more than underestimations harm students (Jussim 2017, p. 8;Willard and Madon 2016).
One more clarification before returning to the discussion of ECB: the self-fulfillment of trait ascriptions just considered is not an all-or-nothing phenomenon. It might unfold over time such that an initially false proposition becomes gradually less inaccurate. These gradual changes, which might fail to result in a complete confirmation of the belief that triggers them, are what I shall henceforth call self -fulfilling effects.

From self-fulfilling effects to Mandevillian benefits of ECB
The preceding considerations help support (P3) and (P4) of Sub-argument II. As for (P3) (i.e., 'ECB among faculty in academia often has self-fulfilling effects.'), recall that ECB is the tendency to (i) believe differences between individuals (when it comes to, e.g., academic performance) to be less pronounced than they are, and to (ii) selectively process information so that it supports these beliefs and serves a politically progressive outlook, e.g., by depicting stigmatized subjects more positively than high status, powerful individuals. Now, in education, which is a key component of academia, ECB is likely to result in precisely the kind of mindset that a sub-set of the teachers in, e.g., Madon et al.'s (1997) study had: many teachers in the study held beliefs about stigmatized students' academic performance that overestimated these students and underestimated real group differences. Since academia is currently likely to contain predominantly progressive faculty, i.e., subjects particularly likely to display ECB, there is reason to believe that ECB is present among many faculty members in academia and often results in the kind of overestimates of students just mentioned. This is because even if ECB contributes only to an equal (vs. an over-) estimation of stigmatized and high-status individuals, faculty with ECB would still overestimate currently low-achieving stigmatized students. Moreover, as noted, these overestimates tend to initiate real, reliable, and occasionally powerful self-fulfilling effects especially among such students. There is thus ground to hold that ECB among faculty in academia often has self-fulfilling effects.
What about (P4) ('If ECB among faculty in academia often has self-fulfilling effects, then ECB among faculty in academia often has Mandevillian ethical and epistemic benefits.')? It isn't difficult to see that the self-fulfilling effects of ECB-based overestimates discussed above are ethically beneficial. Since they tend to be more pronounced in low achievers than in high achievers, these effects facilitate an equalization of high achievements among students. And while faculty members who overestimate stigmatized students might at the same time underestimate privileged students, as mentioned, studies suggest that there is an asymmetry in self-fulfilling effects. These effects tend to be weaker when it comes to underestimates especially in the academic contexts relevant here (Madon et al. 1997, p. 806). There is hence ground to believe that the equalization of academic performance facilitated by faculty overestimates of stigmatized students doesn't necessarily come at the cost of 'levelling down' other students' performance. If we relate these points to ECB, then since ECB is particularily likely to be present among progressive faculty and often results in performance overestimates that produce self-fulfilling effects reducing performance inequality, ECB helps reduce social inequality. And since this social inequality in academia often derives from prejudice (e.g., racism) (Wisdom et al. 2019), the self-fulfilling effects of ECB help tackle unjust social inequality, making them ethically beneficial.
But are these benefits 'Mandevillian' in nature? In recent social epistemological work, a mental state, process, or tendency M is called 'Mandevillian' if M is epistemically problematic within the information processing of the individual with M but conducive to epistemic success of other people, including at the group level (Morton 2014 coined the term, alluding to Mandeville (1705)). For instance, Smart (2018) and Peters (2018) argue that while confirmation bias is epistemically detrimental for individuals with the bias (as it leads to one-sided information processing), in a group of individuals (e.g., scientists), the bias can ensure that the group as a whole conducts a more thorough analysis of problem space than if each member impartially weighed the available data and lacked the bias.
I propose that ECB too (a kind of confirmation bias not yet considered by Smart 2018 and Peters 2018) can be Mandevillian. For even though ECB affects those who have it negatively by contributing to (politically motivated) inaccuracy about social reality, due to its self-fulfilling effects especially in academia, the bias promotes the development of students' epistemic capacities, helping them to form true beliefs, avoid error, and acquire knowledge in, e.g., mathematics (as in the example discussed). The epistemic benefits for stigmatized students, in particular, are likely to be significant. For, as noted, in some of the self-fulfilling effects at issue have been found to be comparable to grade changes from C to B + and SAT score changes from 450 to 550 (Biggs 2009;Jussim 2012). Moreover, as mentioned, SFPs can accumulate across multiple social interactions if different perceivers hold the same kind of view about a target (Madon et al. 2018). Assuming that many faculty members in academia underestimate group differences and overestimate stigmatized students, one would expect the epistemic benefits of ECB for these students to accumulate too. Given its likely significant social epistemic benefits, ECB can thus plausibly be viewed as a Mandevillian cognition: in academia, one subject's politically motivated, epistemically problematic cognition can lead to improvements in the epistemic processing of others. Since the epistemic benefits of ECB are directly related to its ethical benefits (i.e., its contribution to reducing unjust social inequality), I shall refer to both types of benefits as 'Mandevillian'.
With this in mind, (P3) ('ECB among faculty in academia often has self-fulfilling effects.') and (P4) ('If ECB among faculty in academia has self-fulfilling effects, then ECB among faculty in academia often has Mandevillian ethical and epistemic benefits.') are now motivated. This supports (C2): ECB among faculty in academia often has Mandevillian ethical and epistemic benefits.

Sub-arguments III and IV
The points just made have interesting implications. Sub-arguments III and IV help illustrate this. Sub-argument III is based on (C1) and (C2) and adds: (P5): If (i) ECB is currently more common among faculty in academia than if academia contained more politically diverse faculty, and if (ii) ECB among faculty in academia often has Mandevillian ethical and epistemic benefits, then a political diversification of faculty in academia would reduce the overall probability of Mandevillian ethical and epistemic benefits specific to ECB in academia. (P6): A reduction of the overall probability of ethical and epistemic benefits currently gained in academia is an ethical and epistemic cost in academia. (C3): A political diversification of faculty in academia has (inter alia) an epistemic cost in academia.
As for (P5) (i), many surveys suggest that, overall, faculty in academia is currently predominantly progressive, and so a political diversification of academia would mean introducing more people into faculty in academia with other, i.e., non-progressive political viewpoints. The overall number of progressives, and so the overall number of subjects in academia committed to political goals such as social equality and social justice, would then be reduced. That is, people with the specific kind of politically motivated cognition manifested in ECB would decrease. With a political diversification of academia, ECB would thus become less common at universities, colleges, etc.
What about (P5) (ii)? Notice that the premise doesn't say that ECB always has Mandevillian ethical and epistemic benefits but only often. Still, the key point is that the overall likelihood of these benefits in academia would decrease with an increase of non-progressive faculty: if more teaching subjects who are less likely to display ECB enter academia, then the overall probability of ECB-specific effects should go down. Moreover while a decrease in social inequality and an increase in the epistemic performance of stigmatized groups might also result if faculty members hold accurate beliefs about group differences, this doesn't change the fact that the particular Mandevillian ethical and epistemic benefits that are specific to ECB are then likely to be less common. For they are defined as benefits resulting from the bias, and so a reduction in the distribution of ECB in academia, which would happen via a reduction of progressive faculty, will co-vary with a reduction in the overall probability of these benefits.
Turning to (P6), then, in general, if a cognition C now produces both ethical and epistemic benefits in a social environment E with a degree of probability D then any change reducing D in E will make it less likely in E to attain the C-related epistemic benefits that are currently gained. Since losing out on benefits that one currently enjoys is a cost, 13 a reduction of D is itself a cost. The same applies to ECB in academia. Hence, there is a basis for (P6) and so, given (P5), (C1), and (C2), (C3) is now supported: a political diversification of faculty in academia has inter alia an epistemic cost in academia.
I'm focusing only on epistemic costs here because the idea that diversity could be epistemically costly by reducing ECB strikes me as particularly interesting. This is because of the widespread assumption of the opposite view that political diversity is epistemically beneficial in reducing bias (Longino 2002;Duarte et al. 2015;Peters et al. 2020). The point here does, however, also apply to ethical costs (hence the 'inter alia').
Moving Notice that (P7) merely says that the epistemic cost that a course of action might produce in academia is one epistemic reason counting against that action. This is compatible with there being overriding epistemic and/or other reasons for adopting it. So (P7) is modest and I take it to be intuitively plausible enough to assume here. With this last premise in place, (C4) is supported: There is an epistemic reason for not politically diversifying faculty in academia.

Objections
Several clarifications and qualifications with respect to the argument captured in (P1)-(C4) are in order. I will now introduce them by considering and rebutting three objections to the argument. Objection 1 The argument captured in (P1)-(C4) assumes that the Mandevillian benefits that ECB might produce depend on the bias. This is false. All that is needed for the selffulfilling effects that generate these benefits is that people in academia convey high expectations to stigmatized individuals (to prompt behavioral confirmation). Academics can and often do convey such expectations verbally even when they accurately represent group differences and are unaffected by ECB. So the argument doesn't offer a reason to assume that if there were more people without ECB in academia, this would reduce the Mandevillian benefits discussed.
Response I agree that subjects who accurately represent group variances in academic performance could still verbally express high expectations about stigmatized students and this might indeed lead to the same kind of expectation-related self-fulfilling effects that I connected to ECB. However, these effects are likely to be less frequent and less pronounced. This is because subjects who don't genuinely believe in group invariance but nonetheless express claims intended to convey equal (or higher) expectations to stigmatized students are likely to be less able to convincingly signal conviction about group invariance. The reason is that there will be a discrepancy in the subjects' beliefset and their overt behavior and it is well known that the suppression of discrepancies of that kind depend on cognitive resources (i.e., attention, time, etc.), which people often lack (von Hippel and Trivers 2011). Crucially, in the absence of credible signals of conviction, SFP processes are less likely to become operative (Peters 2020, p. 12). For instance, it seems clear that when a child notices that her parents and teachers (i.e., people she cares about) firmly believe that she is good at, say, mathematics, this will motivate her more to invest effort in mathematics so as to conform to the expectations than if she notices that her parents and teachers don't genuinely believe that she is good at math. Lending empirical support to this assumption, Swann andEly (1984, p. 1287) found that when a perceiver formed expectations about a target, SFP effects tended to "occur only when perceivers were certain of their expectancies". Since verbally expressing high expectations related to group invariance while accurately representing group differences is likely to be less convincing to targets than expressing such expectations while also believing in that group invariance (e.g., as the result of ECB), there is reason to assume that the Mandevillian benefits discussed are less likely in the former than in the latter case.
Objection 2 Even if ECB has the Mandevillian benefits discussed, accurate representations of group differences and political diversity are likely to be much more epistemically beneficial in academia than ECB and a lack of political diversity. This is for two reasons. First, when faculty members swiftly detect group inequalities in academic performance, they can implement specific policies to reduce them. This will help increase epistemic performance in stigmatized subjects more efficiently than if these faculty members displayed ECB and self-fulfilling effects occurred. Second, in general, viewpoint homogeneity in social groups is well known to contribute to unreliable belief formation, group polarization, research lacunas, etc., and viewpoint diversity, political or otherwise, helps counteract these significant epistemic problems (Longino 2002, p. 134;Solomon 2006;Duarte et al. 2015;Rolin 2017). Compared to the importance of tackling these epistemic problems, the Mandevillian benefits tied to ECB and to a lack of political diversity are outweighed by the epistemic benefits of accuracy and political diversity.
Response It does seem plausible to assume that accuracy about group differences is more efficient (than ECB effects) in improving subjects' epistemic performance in academia. There is, however, reason to be skeptical about this assumption. Notice first that research suggests children and adults tend to conform and, importantly, expect others to conform to what are in fact only descriptive generalizations about groups (Cialdini et al. 1990;Bicchieri 2017). For instance, in one study, children were first presented with groups of fictive creatures ('Hibbles' and 'Glerks') that were described to them in terms of morally neutral regularities (e.g., Hibbles eat berries of type A whereas Glerks eat berries of type B). Afterwards, the children were shown individuals of each group that either conformed or didn't conform to these regularities (e.g., a Hibble who ate berries of type B) and asked to assess their actions. Surprisingly, the children tended to judge non-conformity as 'not okay' despite only having been presented with purely descriptive generalizations about Hibbles/Glerks (Roberts et al. 2017).
Similarly, studies found that knowledge that men are more likely to be leaders than women made even "explicit egalitarians" and "feminists" less likely to treat a woman sitting at the head of a table as the group leader than a man in the same position (Porter and Geis 1981, p. 52;Mayo and Henley 1981, p. 6). As Madva (2016, p. 194) puts it, "merely knowing what is statistically likely about a group leads individuals to act in some respects as if those statistical generalizations were normative, as if members of that group ought to be treated in a certain way (e.g., as if marks of leadership confer authority to men but not women)". Now, there is little reason to assume that progressive faculty accurately encoding group differences are immune to the tendency to (e.g., inadvertently) transition from description to prescription. If they aren't, however, then accuracy about group differences might lead them to the corresponding normative expectations that group members ought to act accordingly. This evidently doesn't help but harm stigmatized students epistemically. Caution is thus warranted about the assumption that accuracy about group differences is more efficient in improving subjects' epistemic performance in academia than ECBinfluenced teacher overestimations.
Finally, notice that argument (P1)-(C4) is in fact compatible with the view that both accuracy about group differences and political diversity in academia are overall more epistemically and/or ethically beneficial than ECB and a lack of political diversity. As emphasized above, the overall argument here is only meant to provide one so far overlooked reason in favor of ECB and against political diversity in academia. Other considerations might override it. Relatedly, when it comes to the significance of the Mandevillian benefits tied to ECB, these benefits concern only one specific but important aspect of academia, namely education. In contrast, the arguments typically advanced for counteracting a lack of diversity in academia focus primarily on research in academia (e.g., enhanced social criticism, avoidance of research lacunas, etc., Longino 2002;Intemann 2011;Duarte et al. 2015). I agree and have argued myself (Peters 2019;Peters et al. 2020, p. 536) that there are good reasons to tackle a lack of political diversity when it comes to research environments in academia. When it comes to education, however, the argument captured in (P1)-(C4) suggests that a lack of diversity might produce significant ethical and epistemic benefits that point in the opposite direction. Whether these benefits are overall less important than those tied to more political diversity in academia will partly depend on how we weigh the two components of academia and view their potential interdependence. 14 If we focus primarily on education then, given argument (P1)-(C4) and the points just made on the risk of description-to-prescription transitions, it isn't obvious that the benefits of ECB and a lack of political diversity are outweighed by their costs.
Objection 3 The idea that progressives' tendency to underestimate, deny, or overlook real social differences between individuals and groups when it comes to, e.g., academic performance (as measured by standardized tests) is a bias is misguided for two reasons.
First, there is evidence that standardized academic tests are themselves biased against members of stigmatized groups (Santelices and Wilson 2010). Since that is so, when progressives are asked to estimate group differences in academic performance, their response is perhaps simply based on the thought that the tests at issue create such differences rather than reveal them and that different groups are, in fact, likely to be objectively the same in academic performance. Second, progressives' tendency to 'underestimate' real group differences in academic performance arguably just captures their view that individuals of different groups have equal academic potential, i.e., abilities, and that view is accurate.
Response While I shall not deny the possibility of social biases distorting standardized academic test data, a recent critical literature survey representing the current scientific evidence concerning the issue concludes that it is the "broad consensus among psychometricians" that "standardized tests are not systematically biased against members of minority groups" (Holmes 2016, p. 202). Moreover, it is commonly accepted that especially when it comes to quantitative skills and math performance, standardized tests do capture real academic differences in an unbiased manner (Reeves and Halikias 2017). And even if we focus only on this kind of academic performance, argument (P1)-(C4) will still apply. Additionally, if we take the data on politically motivated cognition at face value (Nisbet et al. 2015;Kahan et al. 2017;Lewandowsky and Oberauer 2016;Washburn and Skitka 2018), then it isn't unreasonable to assume that when thinking about group differences, progressives might in some cases be motivated to process information in ways that doesn't have accuracy as its goal but, e.g., social equality. If so, then it should be unsurprising to find progressives that do hold views of group invariance pertaining not only to abilities but also current performance of groups. Finally, the teacher overestimates in, e.g., Madon et al.'s (1997) study in fact did include teacher beliefs about students' current math performance, and several more recent experiments suggest that "[t]eacher overestimation of student achievement" is "common" (Urhahne and Zhu 2015, p. 305). 15 Since findings on politically motivated cognition make it plausible to assume that progressives might often hold inaccurate views about current features of stigmatized subjects, and a here especially relevant experiment (i.e., Madon et al. 1997) did in fact test SFPs in the context of inaccurate beliefs about students' current academic standing, objection 3 can be set aside.

Conclusion
Research on politically motivated cognition provides grounds to take the possibility seriously that progressives display a particular kind of confirmation bias, namely ECB. Supposing that they do, I argued that this bias is likely to lead progressives to act in ways that produce self-fulfilling effects resulting in significant Mandevillian ethical and epistemic benefits especially in academia. I maintained that these benefits would be reduced if academia contained more non-progressives. This is because nonprogressives are, given their different political values, less likely to exhibit ECB. The opposite direction of their politically motivated cognition is likely to 'immunize' them to ECB. While this can be viewed as a restricted epistemic advantage of nonprogressives, it also makes them less likely to have the type of cognitive disposition that produces the mentioned Mandevillian benefits in academia. Increasing the political diversity in a currently predominantly progressive academia thus comes with the epistemic cost of decreasing the probability of these benefits.
The argument developed here offers novel contributions to different areas of philosophical theorizing. It, for instance, casts doubt on the Ethical Rationale against ECB mentioned at the beginning of this paper, i.e., the idea that ECB should be avoided because we need to accurately represent group differences in order to be able to counteract unjust social inequality related to them. The preceding considerations suggest that progressives in academia in fact often reduce such inequality indirectly while inaccurately representing such differences. Argument (P1)-(C4) also challenges the Epistemic Rationale against ECB, i.e., the idea that ECB is epistemically problematic per se since it undermines subjects' ability to track relevant truths. This is because the argument suggests that while ECB is epistemically problematic in one's own cognition, it often has effects that significantly improve other people's truth tracking, particularly that of stigmatized individuals in academia. This point adds to the social epistemological research on Mandevillian cognition (Morton 2014;Smart 2018;Peters 2018), and to work on the epistemic profile of biases (Gendler 2011;Madva 2016;Puddifoot 2017). It also offers a new and perhaps counterintuitive contribution to the theorizing on diversity in philosophy of science. For it is widely accepted among philosophers in that area that diversity in academia is epistemically beneficial as it helps reduce bias (Longino 2002, p. 131f;Solomon 2006, p. 24f;Intemann 2011, p. 115f). In contrast, the preceding discussion suggests that precisely because political diversity would help reduce a particular kind of bias in academia, namely ECB, increasing political diversity in academia would in fact in one important sense be epistemically costly.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.