Understanding Race: The Case for Political Constructionism in Public Discourse

Abstract The aim of this article is to develop an understanding-based argument for an explicitly political specification of the concept of race. It is argued that a specification of race in terms of hierarchical social positions is best equipped to guide causal reasoning about racial inequality in the public sphere. Furthermore, the article provides evidence that biological and cultural specifications of race mislead public reasoning by encouraging confusions between correlates and causes of racial inequality. The article concludes with a more general case for incorporating empirical evidence about public reasoning into philosophical debates about competing specifications of the concept of race.


Introduction
Debates about racial ontologies raise difficult questions about public understanding. On the one hand, racial realism can contribute to racialism in the sense of false biological or cultural accounts of racial inequality (Montagu 1942;Graves 2005;Sussman 2014). On the other hand, eliminativist rejections of racial categories can undermine conceptual resources for addressing racial inequality and therefore contribute to a color blind "racism without races" (Bonilla-Silva 2006;Neville et al. 2013). The result seems to be a dilemma of public understanding in which different racial ontologies run the risk of contributing to either racialism or ignorance toward racial inequality.
This article argues for political constructionism as the account of race that is best equipped to respond to this dilemma. Political constructionism in the sense of this article 1 specifies "race" in terms of hierarchically structured social positions rather than biological or cultural features. The main goal of the article is to provide evidence that political constructionism supports adequate resources for public reasoning by identifying causes of racial inequality. In contrast, biological and cultural specifications of "race" are prone to producing public misunderstandings by foregrounding correlates rather than causes of racial inequality.
The next section introduces an explanatory argument according to which political constructionism provides adequate explanatory resources for engaging with racial inequality while competing specifications of race fail on explanatory grounds. While this argument has a lot of initial plausibility, it fails to acknowledge that alternatives to political constructionism can provide adequate explanatory resources if they specify the causal roles of hierarchically structured social positions through substitute concepts. Sections 3 and 4, therefore, develop an alternative argument for political constructionism that relies on considerations of understanding rather than explanation. It is argued that political constructionism supports adequate causal reasoning, while alternative specifications of race mislead public reasoning through confusions of correlates and causes of racial inequality. Section 5 considers racial eliminativism as an alternative to any realist appeal to race in the public sphere. While there is some preliminary evidence to suggest that eliminativism lacks core benefits of political constructionism, it is argued that the currently available empirical evidence is inconclusive. Section 6 concludes by arguing for a closer integration between philosophy of race and social science research on public understanding of race and racial inequality.

The causal-explanatory case for political constructionism
While current debates about racial ontologies share a concern with the nature of race, it has become widely recognized that they pursue a variety of related but nonetheless distinct projects. First, different speakers, from lay publics to biomedical scientists, use the notion of race in different ways and may ultimately refer to different phenomena. Questions about the nature of race may therefore have different answers in different contexts (Alcoff 2006;Glasgow 2010b;Ludwig 2019). Second, it has become common to distinguish between descriptive accounts that try to identify the current concept of race, and prescriptive accounts that try to identify adequate concepts along current frameworks of "conceptual engineering" (Cappelen 2018), "ameliorative projects" (Haslanger 2000), and "conceptual ethics" (Burgess and Plunkett 2013).
The project of this article is best understood as conceptual engineering of the concept of race for public discourse. "Race" is a deeply ambiguous and vague concept (Ludwig 2015) that could be specified through very different features from ancestry and morphology to cultural and social properties to false racialist or essentialist ideas. Furthermore, there remains considerable philosophical disagreement about how to analyze the current concept of race and its various biological, cultural, and political connotations (e.g., Glasgow 2010a;Hardimon 2017;Spencer 2014). Rather than joining this debate about core features of the current concept of race, this article develops the prescriptive thesis that political constructionism is best equipped to provide resources for public reasoning without creating racialist misunderstandings about biological or cultural causes of racial inequality. 2 In this context of conceptual engineering, political constructionism is best understood as arguing that race should be specified in terms of hierarchically structured social positions. Racialized societies create complex social systems that shape opportunities and disadvantages often before birth (e.g., through nutrition and health care) and develop through all major areas of social life. Political constructionists argue that to be of a certain race should be understood as being positioned in such a system that affects which neighborhood people grow up in, which school they attend, whether they go to college, what job they will be offered, what prejudices they occur, and so on. In this sense, races are perfectly real (pace eliminativism) but they are at the core of being positioned in a hierarchical social system rather than having certain biological or cultural properties (pace both biological realism and cultural constructionism).
While this broad picture is shared among political constructionists, it still leaves space for substantial variation in the specification of racial social positions. Clearly, not all social hierarchies are racial in character (think of class or gender differences), and political constructionists have developed different proposals of how to specify racial social positions. For example, Haslanger (2012, 236) appeals to social hierarchies that rely on "observed or imagined […] bodily features presumed […] to be evidence of ancestral links to a certain geographical region." In contrast, Mills's (2003) defines race in terms of social positions that are shaped by "white supremacy as a sociopolitical system." Mills therefore ties race to one specific historically developed system of domination while Haslanger's more general definition can be satisfied by different sociopolitical systems. These definitional differences are by no means trivial as they lead to the identification of different groups as races. For example, Ludwig (2019; see also Hochman 2017) has argued that Haslanger's account will recognize many racial groups on a global scale (e.g., Tibetans in China, Chinese in Indonesia) that cannot be understood through a system of white supremacy and therefore do not qualify as races in Mills's account.
While these definitional differences matter, this article will not focus on making the case for one particular specification of political constructionism. Instead, the main goal of the article is to argue that prominent versions of political constructionism, including Haslanger (2012) and Mills (2003), share crucial benefits for conceptual engineering in the public sphere because they support adequate causal reasoning about racial inequality, while alternative accounts of race in terms of biological and cultural features mislead public understanding.
The basic argument for political constructionism can be formulated through two simple premises. First, political constructionism identifies causes of racial inequality. Second, race should be specified in terms of the causes of racial inequality. Therefore, "race" should be specified in terms of political constructionism.
To illustrate the first premise, consider racial inequality in incarceration in the United States as addressed in Alexander's The New Jim Crow (2012). Alexander provides a complex analysis of the causes of racialized incarceration rates that involve a large number of interacting factors such as (implicit and explicit) biases of various actors, legal frameworks, economic opportunities, and so on. These causal factors are neither biological (e.g., genetic differences) or cultural (e.g., alleged "cultures of violence") but aspects of hierarchically structured social positions. By specifying race in terms of these positions, the concept is therefore also specified in terms of the causes of racial incarceration differences. In contrast: If race is specified in terms of biological or cultural features, or nonexistent structures such as racial essences, then "race" will not pick out causes of racial inequalities. In fact, interpreting biological or cultural features as causes of racialized incarceration rates would lead straight to false racialist causal narratives.
While the first premise largely articulates the common ground of nonracialist perspectives on racial inequality, 3 it is much less clear why the second premise should be accepted. Surely, one could recognize causes of racial inequality without defining race in terms of these causes. For example, a proponent of biological specifications of race can accept that biological differences between human populations are of limited causal significance in the social domain yet still insist on a specification of race in terms of these differences. Furthermore, an eliminativist can accept everything that a political constructionist has to say about the social production of racial inequality and still insist that "race" should be understood as a failed concept without referent. In other words: there needs to be an additional argument for specifying race in terms of the causes of racial inequality.
One possible strategy is to point out that such a specification is necessary for adequate explanations of racial inequality. If we do not specify race in terms of hierarchically structured social positions, we lack crucial conceptual resources for explaining racial inequality. In contrast, political constructionism provides the most "adequate tool to help achieve social justice" (Haslanger 2012, 304) because it provides resources to explain (and therefore to address) racial inequality. Haslanger makes this line of argument explicit in her analogous account of "gender" in terms of social positioning when writing that her "definitions are proposed […] as providing a better explanation of how gender works" (133).
This causal-explanatory argument can be further motivated by the more general philosophical literature on social categories and social kinds. Bach (2016, 187), for example, has recently argued that we should aim for social categories that "accurately map objectively existing causal-explanatory structures" and therefore qualify as natural kinds. As Bach points out, this appeal to objective social categories should not be misunderstood as a depoliticized scientism that tries to keep political concerns out of our conceptual practices. On the contrary, the objectivity of social categories also matters for political reasons: "in cases where the selected property or properties do not accurately map objectively existing causal-explanatory structures, the social theorist will not have access to forms of understanding and empirical possibilities that bear on politically important features of the target social phenomena" (Bach 2016, 187). Social categories should be specified in terms of causal structures of explanatory relevance, and political constructionism specifies "race" in terms of causal structures that explain racial inequality.
To sum up, the causal-explanatory argument for political constructionism can be understood as involving two steps. The first (causal) step consists in the assumption that political constructionism identifies the causes of racial inequality as illustrated by examples such as the criminal justice and education systems. The second (explanatory) step consists in the assumption that race should be specified in terms of these causes in order to provide adequate resources for explaining how race works in society.
While this causal-explanatory argument has a lot of initial plausibility, there are reasons to think that it misrepresents alternative accounts of race in the current philosophical literature. Of course, we can imagine biological and cultural realists who use their accounts to advocate false causal explanations of racial inequality in terms of biological or cultural difference. Furthermore, we can imagine eliminativists who deny the existence of races to evade engagement with the causes of racial inequality. However, many philosophers of race reject political constructionism while accepting the need for adequate social explanations of racial inequality.
For example, Blum's (2010) eliminativism is not a strategy for avoiding engagement with the causes of racial inequality but proposes an account of racialized groups to capture the same causal factors as a political constructionist account of race. In a similar way, Hardimon proposes the introduction of a new concept of "socialrace" that "is by its nature hierarchical. Some socialraces are dominant; others subordinate" (2014, 71). However, Hardimon also insists that socialrace "is not a race concept proper" arguing that the "logical core" of the ordinary concept is limited to nonhierarchical aspects of ancestry and morphology. The proposals of Blum and Hardimon illustrate a fundamental problem with the causal-explanatory argument: political constructionism does not provide any explanatory benefits over accounts that propose substitutes such as "racialized group" or "socialrace." Whatever the explanatory benefits of political constructionism are, they seem to be preserved if the relevant causal-explanatory structures are acknowledged through some substitute.
Explanations of racial inequality require categories with an adequate intension of populations in hierarchical social positions. In order to explain racial inequality in the criminal justice and education systems, for example, we need to acknowledge how populations are affected by interacting causal factors from (implicit and explicit) biases to economic opportunities. While political constructionism provides an adequate category through its specification of race, so do substitutes such as Blum's "racialized group" and Hardimon's "socialrace." More generally, there seems to be relatively little disagreement in the philosophical literature about the need for a category of hierarchically positioned social groups in the explanation of racial inequality (Mallon 2009). What is controversial, however, is the identification of this category with race (rather than racialized group, socialrace, etc.). The fundamental shortcoming of the causal-explanatory argument is that it provides a justification for the largely uncontroversial issue of category intension while failing to provide any justification for the issue of category identity. As a consequence, alternatives to political constructionism such as eliminativism + racialized group or biological realism + socialrace do not seem to involve explanatory disadvantages.
3. The causal-hermeneutical case for political constructionism The last section developed and criticized an explanation-based interpretation of the causal argument for political constructionism. While political constructionism provides explanatory resources for addressing the social production of racial inequality, I argued that proponents of alternative specifications of race can employ substitution strategies to avoid explanatory disadvantages. Rather than defending the causal-explanatory argument, this section proposes a case for political constructionism that shifts the focus from explanation to understanding.
Even if substitutes such as racialized group provide adequate explanatory resources, political constructionism may still come with crucial advantages in the context of a broader focus on understanding. Recall that the causal argument involves two steps. First, it makes assumptions about the causes of racial inequality in terms of hierarchically structured social positions. Second, it suggests that we should specify race in terms of the causes of racial inequality. The proposed causalhermeneutical argument justifies the second step through considerations of understanding rather than explanation. While political constructionism fosters adequate understanding of the causes of racial inequality, alternative accounts create a risk of misleading public reasoning.
The basic idea of this causal-hermeneutical argument is that understanding-and especially public understanding-requires more than just explanation (De Regt 2017). For example, the mere availability of detailed social-science explanations of the causes of racialized mass incarceration does unfortunately not ensure adequate public understanding of these causes. Even if political constructionism and substitute strategies are on a par in terms of explanatory resources, it may turn out that political constructionism provides more adequate cognitive resources for public understanding of racial inequality. More specifically, I will argue that political constructionism facilitates adequate public reasoning about racial inequality because it specifies race in terms of these causes. In contrast, alternative specifications of race in terms of biological or cultural properties can mislead public understanding about racial inequality by failing to identify relevant causes.
There are at least two ways in which nonpolitical specifications of race can be misleading. First, they can mask causes of racial inequality by encouraging depoliticized perspectives on human diversity. For example, Morning (2011, 147) found that university students in the United States commonly endorse cultural accounts of race in terms of markers such as "what you eat," "what you wear," or "what values your parents teach you," because doing so "evades engagement with the history of oppression that has been part and parcel of racial stratification." Second, both biological and cultural specifications of race can also lead to public misunderstanding by directly encouraging racialist narratives about biological or cultural causes of racial inequality.
While the next section will present more specific evidence of such misunderstandings, there is also a more general reason for concern: while biological and cultural specifications of "race" do not identify causes of racial inequality, they still tend to identify correlates (see table 1). Sometimes, Table 1. Only political constructionism identifies causes of racial inequality. Biological and cultural specifications of race do not identify causes but, rather, correlates of racial inequality.
"Race" identifies correlates of racial inequality "Race" identifies causes of racial inequality Biological specification yes no Cultural specification yes no Political specification yes yes correlates indicate causes. For example, difference in education level are not only a correlate but also a cause of differences in income level. Political constructionism identifies not only correlates but also causes of racial inequality in this sense. In contrast, biological and cultural specifications identify correlates from skin color and ancestry to cultural identities that lead to racialist misunderstandings when interpreted as causes of racial inequality. Of course, proponents of biological and cultural specifications of race can respond to this worry about racialist misunderstandings by pointing out that correlation does not imply causation. Surely, one can specify race in terms of biological and/or cultural properties without misunderstanding these properties as causes of racial inequality. However, the causal-hermeneutical argument is not about logical connections but rather about the actual effects of different specifications of race on public understanding. And it has become widely acknowledged in the psychological literature that noncausal correlational data is often sufficient for public endorsement of causal interpretations (Adams et al. 2017;Bleske-Rechek, Morrison, and Heidtke 2015;Mueller and Coon 2013).
To illustrate this risk with at least one specific study, consider Bleske-Rechek, Morrison, and Heidtke's (2015) study of understanding the distinction between correlation with causation among participants who read texts about potential links between (a) video games and aggression, (b) selfesteem and academic performance, and (c) pornography and relationship satisfaction. In each of the case studies, Bleske-Rechek, Morrison, and Heidtke presented participants a short text (the "experimental vignette") that provided evidence for a causal link or another text (the "nonexperimental vignette") that only provided evidence of a correlation. In the video game case, for example, the experimental vignette described how 500 students were randomly assigned to two groups. One group was asked to play video games every day for several hours while the other group was not allowed to play any video games for the same time period. The nonexperimental vignette described how 500 students were asked how many hours they spend playing video games. Both vignettes also included descriptions of teachers being asked how often these students engaged in aggressive behavior.
While the experimental vignette describes a study that has potential to test a causal contribution of video games to aggression, the second text provides merely correlational evidence that could be explained through many different causal relations. For example, socially isolated students may play more video games and be more aggressive even if the video gaming itself does not contribute to the aggressive behavior. More generally, factors such as a family situation or social network of students could lead to (positive or negative) correlations between video gaming and aggressive behavior without the former being causally significant for the latter.
Although only the experimental vignette provided evidence of a causal relation, Bleske-Rechek, Morrison, and Heidtke found that "participants drew causal inferences from noncausal data as often as they did from causal data" (2015, 65). After reading one of the two texts, participants largely endorsed the interpretation that confirmed their preconceived assumption about the link between video games and aggression rather than the evidence in the text about a possible causal relation.
Psychological research on the confusion of correlation and causation supports the worry that nonpolitical specifications of race can constitute "cognitive traps" by presenting correlational information that leads to causal misunderstandings. Of course, this does not mean that everyone will always misunderstand biological or cultural specifications of race as specifications of causes of racial inequality. As Mueller and Coon (2013, 288) point out, the "ability to distinguish between correlational and causal claims" can be acquired. However, they also emphasize that acquiring this ability "requires regular and systematic practice, feedback, and reflection" (288). Given that large parts of the public do not have access to this "systematic practice, feedback, and reflection" in reasoning about race, the causal-hermeneutical argument seems supported by worries about public misunderstanding of the causes of racial inequality.

Empirical evidence in support of the causal-hermeneutical argument
The previous section introduced a causal-hermeneutical case for political constructionism by arguing that public reasoning about racial inequality is facilitated by a specification of race that tracks causes rather than correlates. While political constructionism supports adequate understanding of causal relations, alternative specifications of race can constitute "cognitive traps" that mislead public understanding by encouraging confusions of correlates and causes of racial inequality. The aim of this section is to argue that empirical research in fields such as communication and education studies provides support for these worries but has not received sufficient attention in the philosophical literature.

4.a Biological properties
The worry about cognitive traps that mislead public reasoning seems most straightforward in the context of biological specifications of race. To be sure, it is not sufficient to show that endorsement of biological specifications of race is positively correlated with racist beliefs and attitudes (e.g., Williams and Eberhardt 2008). Correlational data can be explained in different ways. For example, it could turn out that people tend to justify their racism by adopting biological accounts of race. As Williams and Eberhardt acknowledge, a correlation "leaves unaddressed the causal nature of this relationship. Does a biological conception of race produce greater acceptance of racial disparities or does greater acceptance of racial disparities lead people to adopt a biological conception of race?" (1037).
However, there is indeed evidence that biological specifications of race causally contribute to racist beliefs and attitudes. The most common experimental strategy is to manipulate participant beliefs about the nature of race through mock newspaper articles or textbooks and to measure the effects on further beliefs and attitudes toward race. Unfortunately, the results of such studies are deeply worrying as they indicate that the introduction of biological accounts of race (a) increases belief in essential racial difference, (b) increases genetically based racism, and (c) decreases emotional responses to racial inequality.
(a) Essential Racial Differences: Phelan, Link, and Feldman (2013) studied how different accounts of race affect beliefs in essential racial differences. Mock newspaper articles introduced race either (1) directly through claims about the biological reality of race, (2) directly through claims about the social construction of race, or (3) indirectly through claims about the relation between race and genetic diseases. Belief in essential racial differences was measured through endorsement (on a four-point scale) of statements such as "Although black and white people may be alike in many ways, there is something about black people that is essentially different from white people." Phelan, Link, and Feldman found that both direct and indirect introduction of biological accounts significantly increased endorsement of essential racial differences compared to both the introduction of race as a social construct as well as a control condition. (b) Genetically Based Racism: While Phelan, Link, and Feldman's results indicate that biological accounts of race contribute to essentialist beliefs, Donovan's recent studies (2014,2016) provide more direct evidence about genetically based racism. Donovan presented high school students with racialized and nonracialized textbook passages about genetic diseases. While the racialized version of the textbook tied genetic diseases to racial categories (e.g., Africans have a higher risk of sickle-cell anemia, Caucasians have a higher risk of cystic fibrosis), the nonracialized version explained disease patterns in terms of other factors (e.g., the sickle-cell allele provided an adaptive advantage in regions with malaria). Donovan found and replicated that the racialized passages led to significantly higher scores in the genetically based racism instrument (GBRI). Given that the GBRI is based on agreement with statements such as "Racial differences in academic ability are caused by genetics," Donovan's studies indicate that biological conceptualizations of race can directly contribute to false racialist beliefs about the causes of racial inequality. (c) Emotional Response toward Racial Inequality: Williams and Eberhardt (2008) provide evidence about effects beyond propositional beliefs through a study of how biological accounts of race affect emotional responses to racial inequality. They manipulated participants' accounts of race through two mock articles that described research results from the journal Gene. The first version, titled "Scientists Pinpoint Genetic Underpinnings of Race," insisted that we "generally inherit race along with everything else" while the second version, titled "Scientists Reveal That Race Has No Genetic Basis," argued that "the practice of classifying people into racial groups based on certain patterns of physical appearance is entirely cultural in origin" (1037). After reading one of these articles, participants were presented with an article from The New York Times titled "Plight Deepens for Black Men, Studies Warn." The article describes marginalization of young black men in American society and stresses that "in the country's inner cities, the studies show, finishing high school is the exception, legal work is scarcer than ever and prison is almost routine, with incarceration rates climbing for blacks even as urban crime rates have declined." After reading this article, participants were asked to evaluate their emotional response on a five-point scale by addressing whether the content left them "moved," "concerned," "upset," and "nervous." The study showed significant decline in emotional response in the context of biological accounts of race.
Studies such as (a) through (c) suggest a troubling picture about the effects of biological specifications of race in the public sphere. Endorsement of biological accounts of race seems to contribute to (a) essentialist beliefs about race, (b) an increase in genetically based racism, and (c) a decrease in emotional response to racial inequality.

4.b Cultural properties
It has become widely acknowledged in the social-science literature that challenges of biological racialism have contributed to a shift toward a cultural racialism that understands racial inequality in terms of alleged cultural causes. For example, Balibar (1991, 21-22; see also Bonilla-Silva 2006; Morning 2011) has influentially diagnosed a "neoracism" whose dominant theme is not biological heredity but the insurmountability of cultural differences." As Paul Gilroy (1990, 266) pointed out more than two decades ago: "The culturalism of the new racism has gone hand in hand with a definition of race as a matter of difference rather than a question of hierarchy." First, this culturalism has been discussed as a strategy for expanding racialism toward populations that are often not perceived as biologically racial such as Muslims (e.g., Meer 2013). Second, culturalist frameworks allow maintenance of racialist narratives in societies that discourage traditional forms of biological racialism. Given the arguments from the previous sections, these worries about a shift toward culturalist variations of racialism can be supported with a more specific hypothesis about public reasoning about racial inequality. Shifting specifications of race from biology to culture leaves the basic problem of "cognitive traps" intact because it does not shift attention from mere correlates (e.g., skin color, continental ancestry) to actual causes (e.g., access to socioeconomic resources) of racial inequality. Instead, cultural specifications of race shift the focus to other correlates that can be easily misunderstood and misused through false causal narratives that emphasize constructs such as an alleged "culture of violence" or "culture of poverty" that are assigned selectively to racial groups.
While much of the recent experimental literature on racial reasoning focuses on biological concepts (e.g., Donovan 2016;Phelan, Link, and Feldman 2010;Williams and Eberhardt 2008), the concern about cognitive traps can be supported with various qualitative studies in the social sciences. For example, consider Alexander's (2012) discussion of misunderstandings of the criminal justice system in The New Jim Crow. While Alexander develops a detailed analysis of the socioeconomic causes of mass incarceration, she also confronts a common culturalist narrative that relies on "the notion that black culture has devolved in recent years, as reflected in youth standing on the street corners with pants sagging below their rears and rappers boasting about beating their 'hos' and going to jail" (179). As Alexander argues, such culturalist framings systematically distort adequate causal analysis of mass incarceration as they focus on cultural symptoms of marginalization and therefore shift public attention away from its socioeconomic causes. Again, the confusion of correlates and causes is what makes this move so seductive in public discourse: cultural symptoms of socioeconomic marginalization are indeed correlated with incarceration rates but distort understanding of causal pathways when guiding public reasoning about mass incarceration.
Alexander's discussion of culturalist narratives about crime provides only one of many examples of what Bonilla-Silva has influentially called the "cultural racism frame." For example, Bonilla-Silva's Racism without Racists (2006) analyzes this frame on the basis of interviews from two surveys (the Survey of Social Attitudes of College Students and the Detroit Area Study) that include further examples of how cultural frameworks mislead about the causes of racial inequality. Some of Bonilla-Silva's respondents explicitly embraced cultural racialism in developing causal narratives about racial inequality. Asked about charges of "laziness," for example, a student responded: "I totally agree with that. I don't think, you know, they're all like that, but, I mean, it's just that if it wasn't that way, why would there be so many blacks living in the projects? You know, why would there be so many poor blacks? If they worked hard, they could make it just as high as anyone else could. You know, I just think that's just, you know, they're raised that way and they see what their parents are like so they assume that's the way it should be" (40). Other respondents understood the causes of racial inequality through a more careful causal narrative of a self-perpetuating "culture of poverty" which defines "family structure. Maybe it's not [being] able to support the child and, you know, in school and really encourage. It might be that it's a single-parent family and it's necessary [for them] to get out and get a job, you know, a full-time job and work a part-time job and still try to go to school. Maybe it's not encouraged as much for, like long term, it's mainly survival" (41).
Both Alexander and Bonilla-Silva argue that culturalist framings contribute to what Alexander (2012, 178) calls "a profound misunderstanding regarding how racial oppression actually works." Framings in the sense of Bonilla-Silva can be used to justify already existing prejudices but are also guides for shaping reasoning about the social world. Culturalist framings of racial differences therefore reinforce the worry that the problem "cognitive traps" extend from biological to cultural specification of race. In both biological and culturalist cases, race is primarily understood through correlates rather than causes of racial inequality. In both cases, emphasis on correlates can create misunderstandings that obscure the political character of the production of racial inequality and rather reinforce biologically or culturally motivated racialism. In the context of conceptual engineering for the public sphere, political constructionism seems to be clearly preferable in providing tools for adequate reasoning about causal relationships. 4 5. Toward a public discourse without "race"? The evidence from the previous sections suggests that a specification of race in terms of social hierarchies supports public reasoning while a focus on biological and cultural differences fosters public misunderstanding. However, this evidence alone is not sufficient to guide conceptual engineering of "race." Even if specifications of race in terms of hierarchical social positions are preferable over biological and cultural specifications, one could also reject any realist appeal to race in public discourse. Given the prevalence of racialist misunderstandings in the sense of the last two sections, one may specify race through clearly falsified (e.g., essentialist and/or racialist) assumptions that show that races do not exist at all. For example, Hochman (2017, 66) develops this line of argument by arguing that the "continued use of race as an analytic concept by social constructionists is likely to reinforce racial naturalism in the public imagination, even though it is the intention of most constructionists to distance 'race' from biology." The assumption that an elimination of "race" is beneficial in public discourse is also reflected in real-life cases of conceptual engineering in many European countries after the Holocaust. While realist racial discourse remains common in the United States, many European countries-including Germany, the Netherlands, and France-have built strong eliminativist norms around the notions of Rasse, ras, and race humaine that makes a realist appeal to races largely impossible. For example, Plümecke (2014, 12) describes Rasse in Germany as "the unword of the last sixty years, semantically entangled with the purity and destruction-oriented eugenic policy of the Nazis. Not even racist pamphlets […] are allowed to contain this word" (translation by author).
The considerations from the previous sections can provide some support for such eliminativist strategies in public discourse. I have argued that biological and cultural specifications of race mislead public reasoning through confusions of correlates and causes. While a purely political specification of race would avoid these problems, it is difficult to keep biological, cultural, and political features neatly separated in public discourse. From the perspective of conceptual engineering, one may therefore conclude that the most consistent strategy for public discourse is to avoid any realist appeal to race whatsoever.
However, a straightforward elimination of "race" would also have disadvantages in public discourse. The last sections did not only provide negative evidence against biological and cultural accounts of race but also positive evidence in favor of political constructionism. Political constructionism can support adequate public understanding of racial inequality by putting its causes at the center of reasoning about race. An elimination of "race" runs the risk of undermining this benefit by providing no guidance for thinking about the causes of racial inequality. And indeed, eliminativists like Blum (2002, 164) are very much aware of the risk that "racial thinking would continue largely unabated, simply driven underground by a norm discouraging its expression." More generally, eliminativism needs to address the risk of color blindness in the sense of strategic ignorance (e.g., Gines 2014;Ikuenobe 2013;Monahan 2006) toward racial inequality.
The arguments from the previous sections allow us to formulate a more specific hypothesis about the link between color blindness and an elimination of "race" from public discourse. The positive case for political constructionism has been based on the observation that it highlights the causes of racial inequality in the very conceptualization of race and therefore provides an important tool for public reasoning. Eliminating "race" from public discourse runs the risk of giving up this benefit and therefore giving up conceptual resources for addressing racial inequality in public discourse.
Is it possible to provide more conclusive empirical evidence for such a hypothesized link between color blindness and eliminativism? There are at least two potential sources of evidence to support this hypothesis. First, the elimination of racial discourses from European languages-including Dutch (ras), German (Rasse), and French (race humaine)-after the Holocaust can be seen as largescale experiments in eliminativist conceptual engineering in the public sphere. And while eliminativism about "race" is still widely considered a premise of antiracist politics in Europe (e.g., Amesberger and Halbmayr 2005;Cremer 2010;Jacob 2018), there has been increased concern about color blindness and the lack of discursive resources to talk about racial inequality in Europe.
In the Dutch context, for example, Gloria Wekker's recent book White Ignorance (2016) starts with the observation "that interracial situations, conversations, and phenomena that would be totally unacceptable in a U.S. context would pass without any frowns or critical comments in the Netherlands" (ix). Wekker's detailed analysis of color blindness in the Dutch context provides a striking example of how the elimination of conceptual resources surrounding race does not only contribute to a false sense of "having overcome racism" but encourages reactions "to deny the seriousness of a racist event, to belittle it, to hold it up to impossible definitional standards, to analyze it to pieces, so that it evaporates into thin air" (38).
In the French context, eliminativism has been deeply ingrained in wider debates about egalitarianism. For example, the collection of data on race and ethnicity in France has not only been opposed but even judged unconstitutional. Simon (2008, 8) has influentially described this situation as a "choice of ignorance" in which "France is officially a society without 'race' [but] racism and racial discriminations are as widespread as anywhere else." Simon's analysis does not only suggest that the elimination of "race" from public discourse has been inefficient in addressing racial discrimination but also that it may actively "conceal the extent of discrimination" (8).
In addition to these real-life cases of conceptual reform in post-Holocaust Europe, there has also been some attempts to address color blindness through experimental settings. For example, consider Apfelbaum et al.'s (2010) study of color-blind and diversity-oriented approaches in elementary-school education. Apfelbaum et al. presented eight to eleven year old students narratives that emphasized either similarity (e.g., pointing out that "we're all the same") or value diversity (e.g., pointing out that "race matters" and that we need to "recognize how we are different from our neighbors and appreciate those differences"). Following this prime, they presented students with cases of ambiguous and explicit racial discrimination. The ambiguous case involved a white student not inviting a black classmate to his birthday party because he was not able to buy any of the presents on his wish list, while the explicit case involved a white soccer player assaulting an opposing player because he was black. The effects of priming through value-diversity and color-blind narratives was considerable: given the value-diversity priming, 43 percent and 77 percent of the students identified discrimination in the ambiguous and the explicit case. Given the color-blind priming, the identification rates reduced to 10 percent and 50 percent. Experimental studies such as Apfelbaum et al.'s (2010) therefore support wider worries that have been articulated in the literature on color blindness. Rejecting the category of race may ultimately do more harm than good in the public sphere by fostering ignorance about the causes or the very existence of racial inequality.
To sum up, there is indeed some evidence that a straightforward elimination of "race" from public discourse has negative effects compared to political constructionism because it does not provide resources for thinking about the causes of racial inequality. That being said, it is much less clear that this evidence is sufficient to challenge philosophical versions of eliminativism that aim for a substitute rather than a straightforward elimination of "race." As Glasgow (2010a) has argued in detail, metaphysical antirealism can be combined with a large variety of prescriptive positions regarding conceptual engineering. His own reconstructionism largely overlaps with the arguments of this article as it embraces a concept of race* that refers to social kinds. Glasgow's suggestion therefore illustrates how a descriptive antirealism is compatible with political constructionism as a prescriptive proposal for conceptual engineering. However, antirealism can also be combined with a substitutionism that aims to "give up on race" (Blum 2002, 164) in the public sphere but also replaces it with a different concept such as "racialized group" (see also Blum 2010;Hochman 2017Hochman , 2018cf. Msimang 2019).
Given that neither Apfelbaum et al.'s experiments nor eliminativist discourses in Europe include a (well-established) concept of racialized group, it is indeed not clear how much this evidence challenges versions of eliminativism that propose to substitute "race" with "racialized group." Furthermore, it is not difficult to imagine that a substitute concept like "racialized group" could fulfill some of the positive functions of "race" such as creating a frame for political solidarity among marginalized groups (Blum 2002, 170) or even guide causal reasoning about racial inequality.
If this substitution of "race" with "racialized group" is supposed to be a proposal for actual conceptual reform in the public sphere, however, it will have to provide positive evidence of feasibility and effectiveness. While it is easy to imagine a concept of racialized group that avoids both racialist misunderstandings and color blindness, it is much less clear that real-life attempts at elimination and substitution would have such positive effects. In fact, European languages have come up with all kinds of substitute concepts such as the Dutch allochtoon or the German Migrationshintergrund ("migration background"). As the literature on race and racism in Europe suggests (e.g., Barskanmaz 2011;Bessone 2013;Wekker 2016), these substitutes have not been effective conceptual resources for navigating between color blindness and racialist misunderstandings. If the move from "race" to "racialized group" is supposed to be an applied proposal for conceptual engineering in the public sphere, it is therefore in need of empirical support that indicates the effectiveness of substitution strategies in guiding public reasoning about racial(ized) inequality.
6. Conclusion: Rethinking public understanding of race The aim of this article has been to develop an argument for political construction as a strategy for conceptual engineering in the public sphere. I have argued that political constructionism supports adequate reasoning about the causes of racial inequality while biological and cultural accounts of race can mislead public reasoning about racial inequality by encouraging confusions between correlation and causation. While I have argued that this causal-hermeneutical argument for political constructionism needs to be taken seriously, the discussion of eliminativism also highlights the complexity of unresolved empirical questions about the effects of different conceptual strategies on public reasoning. Addressing these questions requires that philosophers develop more sustained interactions with social-science research on public reasoning about race and racial inequality.
David Ludwig is an assistant professor at the Knowledge, Technology and Innovation Group at Wageningen University, Netherlands. His research is concerned with interactions between science and society and focuses on social epistemology and social ontology in global contexts.