You are here

Research & Current Work

Present Research

An Introduction to Implicit Bias: Knowledge, Justice, and the Social Mind (co-editing volume with Erin Beeghly, Routledge) (Proposal and authors here)

It is now well-established that unintentional or unreported biases shape all aspects of social life.  Imagine walking through a grocery store.  The smaller the floor tiles, research shows, the slower people tend to walk.  The slower people walk, the more they buy.  These unconscious biases are well-known to marketers and consumer psychologists.  Yet store shoppers do not notice the ways in which the floor’s tile-size affects how they walk or their spending decisions.

Examples such as this are only the tip of the iceberg.  Increasingly, psychologists cite unconscious mental processes to explain persistent social inequities and injustices in a broad range of contexts, including educational, corporate, medical, and informal contexts.  Implicit biases have been invoked to explain heightened police violence against black US citizens, as well as subtle forms of discrimination in the criminal justice system and underrepresentation of women and people of color in the workplace.  Now a popular buzz word, “implicit bias” was even discussed by likes of Hillary Rodham Clinton and Donald Trump during the 2016 U.S. Presidential debates.

Our handbook is the first philosophical introduction for beginners on topic of implicit bias.  It addresses fundamental questions about such biases, including:

  • What is implicit bias? 
  • How do implicit biases relate to other, more familiar mental states, such as beliefs, desires, and intentions? 
  • What is ‘implicit’ about implicit bias?
  • How do implicit biases relate to explicit biases and prejudices?
  • How do implicit biases compromise our knowledge of others and social reality?
  • How do implicit biases affect our social and political institutions, and what can we do to combat these influences?  
  • What are the empirical and normative sources of skepticism about the nature and social-political significance of implicit bias?

Social Psychology, Phenomenology, & the Indeterminate Content of Unreflective Racial Bias (for a volume on Race and Phenomenology, edited by Emily Lee, Rowman & Littlefield) (Materials from WashU Prejudice conference, November 2017: Handout; PPT)

Social psychologists often describe “implicit” racial biases as entirely unconscious, and as mere associations between groups and traits, which lack intentional content, e.g., we associate “black” and “athletic” in much the same way we associate “salt” and “pepper.” However, recent empirical evidence consistently suggests that individuals are aware of their implicit biases, albeit in partial, inarticulate, or even distorted ways. Moreover, evidence suggests that implicit biases are not “dumb” semantic associations, but instead reflect our skillful, norm-sensitive, and embodied engagement with social reality. This essay draws on phenomenological and hermeneutic methods and concepts—principally represented in work by Merleau-Ponty, Fanon, Gadamer, Dreyfus, and Alcoff—to better understand what social-psychological research has begun to reveal about the conscious access individuals have to their own racial attitudes, as well as the intentional contents of the attitudes themselves.

First, I argue that implicit racial biases form part of the “background” of social experience. That is, while they exert a pervasive influence on our perceptions, judgments, and actions, they are frequently felt but not noticed, or noticed but misinterpreted. Second, I argue that our unreflective racial attitudes are neither mere associations nor fully articulated, propositionally structured beliefs or emotions. Their intentional contents are fundamentally indeterminate. For example, when a white person experiences a “gut feeling” of discomfort during an interaction with a black person, there is a question about the meaning or nature of that discomfort. Is it a fear of black people? Is it anxiety about appearing racist? There is, I argue, no general, determinate answer to such questions. The contents of our unreflective racial attitudes are fundamentally vague and open-ended, although I explain how they nevertheless take on particular shapes and implications—that is, their content can become determinate—depending on context, social meaning, and structural power relations. (If, for example, a perceived authority figure, such as a politician, parent, or scientist, encourages you to believe that your uncomfortable gut feeling is a justified fear of other social groups, then that is what your gut feeling is likely to become.)

Intersectionality as a Regulative Ideal (collaboration with Katherine Gasdaglis)

What is the intersectional thesis a thesis about?  Some understand it as a claim about the metaphysics of social kinds, oppression, or experience; about the limits of antidiscrimination law or identity politics; or about the importance of fuzzy sets and multifactor analysis in social science.  We argue, however, that intersectionality, interpreted as a thesis in any particular theoretical domain, faces classical regress and reference-class problems.  Taken, e.g., as a metaphysical claim about social kinds, intersectionality entails that “blackness” is not a genuine kind, because gender and race are “mutually constitutive.”  Likewise, “black-womanhood” isn’t a kind, because gender, race, and class intersect, ad infinitum.  There is no in-principle way to carve social reality at its joints, if social kinds are intersections of indefinitely many factors.  Moreover, there’s no a priori reason to privilege one grouping (or reference class) over another when making generalizations.

We argue that these problems dissolve when intersectionality is modeled as a regulative ideal of social science, i.e., a unifying and guiding methodological principle, and not as an empirical theory or claim.  As an ideal, intersectionality requires social science to view its objects of study as if they were infinitely complex and particularistic.  As regulative, intersectionality suggests an expanding set of concrete heuristics for guiding theory and research.  On our view, the value of any particular scheme of social categories must be determined empirically and not a priori.  If the question “who’s left out” is taken empirically, the regress and reference class problems facing intersectionality dissolve.

Duties of Social Identity? Intersectional Objections to Sen's Identity Politics (collaboration with Shannon Doberneck) (Paper Draft - Comments welcome!)

Amartya Sen (2006) argues that sectarian discord and violence are often attributable, in part, to the pervasive tendency to see ourselves as members of a single social group standing in opposition to other groups (Sunni vs. Shia, East vs. West, etc.). Sen thus argues that we must recognize that we inevitably belong to many different groups (e.g., ethnic groups, socioeconomic classes) and that we are obligated to choose, in any given context, which among our multiple affiliations to prioritize. We argue that Sen’s account is both descriptively and normatively flawed, and overlooks significant lessons from feminist thinkers such as Elizabeth Spelman and Kimberlé Crenshaw. Sen’s account is descriptively flawed because it wrongly assumes that individuals’ identity affiliations are in-principle separable for the purposes of practical deliberation. It is normatively flawed, first, because requiring individuals to prioritize among “identities” that are not actually separable is unfair, and, second, because this requirement is particularly unfair for members of multiple disadvantaged groups, e.g., women of color living in poverty. Even if their identities were in-principle separable, requiring these individuals to prioritize among them may simply compound the injustices and disadvantages they already suffer. Such a forced choice obscures the particular injustices they face, i.e., injustices that reflect a complex combination of racism, sexism, and classism.

We Still Don’t Know “How Mental Systems Believe”: Gilbert’s Spinozan Theory Reconsidered

The contention in my dissertation that implicit biases constitute a distinctive kind of mental state has compelled me to think more broadly about the nature of mental kinds, such as belief, desire, emotion, and intention.  I am currently preparing a manuscript on the nature of belief.  I critically examine the influential “Spinozan” theory of belief formation, according to which the mind is dramatically more gullible than we tend to think.  This theory, introduced by psychologist Daniel Gilbert in the 1990s and now gaining popularity among philosophers (e.g., Bryce Huebner, Andy Egan, and Eric Mandelbaum), states that the mind automatically believes every proposition that it entertains, although it can subsequently reject that belief with mental effort.  I argue that, although the theory is not as radical as it might sound, it is nevertheless beset with internal confusions and, worse, is ill-supported by the empirical evidence, including the key studies Gilbert and colleagues performed to defend the theory.  I conclude by advising caution about the prospects of developing any purely cognitive, mechanistic account of belief.  The mechanisms of belief, I propose, are irredeemably entangled with emotion and desire.

Equal Rights for Zombies? Phenomenal Consciousness and Responsible Agency (Draft for Eastern APA 2018 here - Comments welcome!)

Intuitively, moral responsibility requires conscious awareness of what one is doing, and why one is doing it, but what kind of awareness is at issue? Levy (2014) argues that phenomenal consciousness—the qualitative feel of conscious sensations—is entirely unnecessary for moral responsibility. He claims that only access consciousness—the state in which information (e.g., from perception or memory) is available to an array of mental systems (e.g., such that an agent can deliberate and act upon that information)—is relevant to moral responsibility. I argue that numerous ethical, epistemic, and neuroscientific considerations entail that the capacity for phenomenal consciousness is necessary for moral responsibility. I focus in particular on considerations inspired by Strawson (1962), who puts a range of qualitative moral emotions—the reactive attitudes—front and center in the analysis of moral responsibility.​

The Inevitability of Aiming for Virtue (collaboration with Jennifer Kanyuk) (Here is our poster from the Bias in Context conference at Utah in October 2017)

We defend Fricker’s (2007) virtue-theoretic proposals for grappling with epistemic injustice against two leading criticisms. First, some have objected that her account of the cultivation of epistemic virtue is empirically implausible, suffering both from the general limitations that, situationists have argued, plague all ethical and epistemological virtue theory as well as from more particular pitfalls facing individuals who try to overcome their own “bias blind spots” (for discussion, see e.g., Alcoff, 2010; Alfano, 2013, 2015; Davidson & Kelly, 2015; Sherman, 2016a, 2016b; Washington, 2016). Second, and relatedly, many have argued that Fricker’s prescriptions for cultivating epistemic virtue are excessively individualistic, and fail to appreciate the underlying structural factors driving epistemic injustice and oppression (e.g., Anderson, 2012; Dotson, 2012; Langton, 2010; Maitra, 2010; Medina, 2012; Pohlhaus, 2012).

In response, we argue that Fricker’s account is both empirically oriented and plausible—indeed, pursuing epistemic virtue in roughly the way she describes strikes us as inevitable for anyone who takes seriously the systemic epistemic injustices she and others have identified. According to Fricker, the pivot away from epistemic injustice depends in part on moments of self-critical awareness, states of cognitive dissonance in which an individual realizes that she may, e.g., be underestimating an interlocutor’s credibility due to stereotypes or prejudices. Fricker further advocates that individuals cultivate epistemic habits that aim to consistently neutralize the effects of such prejudices on their credibility estimates. We agree with Fricker on both counts: working to cultivate such habits is an integral component of what we ought to do in the face of pervasive epistemic injustice. We assemble a range of findings from social and cognitive psychology that bear these claims out (Devine, Forscher, Austin, & Cox, 2012; Fiske et al., 2015; Gawronski, Brochu, Sritharan, & Strack, 2012; Monteith, 1993; Monteith, Ashburn-Nardo, Voils, & Czopp, 2002; Monteith, Lybarger, & Woodcock, 2009; Stewart, Latu, Kawakami, & Myers, 2010; Unkelbach, 2007).

That said, we do not read Fricker as advocating that her specific proposals constitute the only means through which individuals and institutions should combat epistemic injustice. We therefore build on her account—drawing in particular on several papers published after her groundbreaking Epistemic Injustice (Fricker, 2007, 2010b, 2010a, 2012, 2013)—by beginning to sketch a fuller picture of the structure of cultivating epistemic virtue. Virtue cultivation must, we argue, occur on two broad but interrelated fronts: first, the direct retraining of more automatic and unreflective patterns of thinking, feeling, and reacting to epistemic social reality (cf. Alcoff, 2006; Madva, 2016a,b, 2017), and second, the cultivation of more reflective, metacognitive virtues (cf., Medina, 2012; Kanyuk ms), such as the ability to swiftly identify contexts in which our first-order epistemic intuitions are likely astray. Although we articulate a range of individual-level obligations, our account is not individualistic. With Fricker and many of her commentators, we argue that individual self-transformation is a necessary but not sufficient component of the struggle for epistemic justice. Accordingly, we sketch several ways in which individual virtue cultivation must be a socially and institutionally embedded process. Moreover, we argue that this process is ongoing. While most individuals cannot actually achieve such moral-epistemic ideals, many can (and therefore should try to) get much closer than they already are.

Meta-Analyses and Predicting Behavior: In Defense of Implicit Measures (collaboration with Michael Brownstein and Bertram Gawronski) (SPP 2017 PPT)

What is the status of research on implicit bias?  Criticism is ubiquitous.   Recent meta-analytic reviews suggest that the Implicit Association Test is a “poor” predictor of behavior (Oswald et al. 2013) and that changes in scores on implicit measures may not be associated with changes in behavior (Forscher, Lai, et al., 2017).  Prominent philosophers have questioned the validity of research on implicit social cognition altogether.  Edouard Machery (2017), for example, describes an ongoing “rescue mission” within the field, implying that the relevant research is in peril of being discredited.  Machery argues that leading methods for studying and theorizing about implicit bias need to be rethought from the ground up, writing that we should not “build theoretical castles on such quicksand.”  Headlines in the popular press have been even more pointed.  New York Magazine reports, “Psychology’s Favorite Tool for Measuring Racism Isn’t Up to the Job,” while the Chronicle of Higher Education asks, “Can We Really Measure Implicit Bias? Maybe Not.” 

We argue that, while there is ample room for improvement, research on implicit bias is not in crisis.  It neither has been nor will be discredited.  Research on the causes, structure, and behavioral effects of implicit social cognition will and should continue to play a role in the sciences of the mind as well as in efforts to understand, and ultimately combat, discrimination and inequality.  We first describe the central issues that have been described as crises, anomalies, or puzzles for the field (§2).  To demonstrate that these alleged anomalies are empirical questions on which progress is steadily being made, we place them in the broader historical context of theorizing the relationship between mental states and behavior (§§3-4).  We respond to potential criticism (§5), then, finally, point to directions for future research (§6).

The Irreplaceable Epistemic Value of Diversity (collaboration with Katie Gasdaglis and Brian Kim)

There is little debate that diversity can be valuable.  A diversity of experiences can paint a fuller picture of human life.  A diversity of ideas can inspire discussion and debate.  A diversity of perspectives can ensure that we achieve some degree of objectivity or inter-subjectivity.  However, what types of diversity is valuable?  Why is it valuable?  And for whom and for what is it valuable?  We hope to fill in the story about the value of diversity by exploring the fringes of the conceptual landscape.  We are specifically interested in the epistemic value of diversity rather than its ethical or political value per se.

It is widely agreed upon that diversity is epistemically valuable because diverse groups of individuals are often more likely to possess the information necessary for good decision making.  However, in order to further our understanding of the epistemic value of diversity, we hope to identify cases in which it is not only more likely for diverse groups to possess the required information, but it is also impossible for homogeneous groups to possess the required information. Thus, we hope to argue that there are cases in which the epistemic value of diversity is irreplaceable.

Cultivating Epistemic Virtue: A User’s Guide (collaboration with Jennifer Kanyuk)

We employ many heuristics or “rules of thumb” to form beliefs about the world.  For example, according to the so-called “truth effect,” we are more likely to believe a statement simply because we have encountered it before.  Familiar statements tend to seem “truer” than unfamiliar ones.  Are beliefs formed on the basis of such heuristics ever epistemically justified? 

Theorists who have debated the reliability and rationality of such heuristics have taken for granted that these heuristics are fixed, effectively hard-wired into our minds.  Both sides of the debate thus overlook the possibility of improving.  They dismiss out of hand the possibility of cultivating epistemic virtue.  This is a mistake.  For example, Unkelbach (2007) found that, with a little practice, individuals can overcome the “truth effect,” and learn that, in certain contexts, familiar statements are less likely to be true than unfamiliar statements.  (Consider how, e.g., one learns over time to be wary of news reports made by specific media outlets.)

We are developing an empirically-informed account of the structure of cultivating epistemic virtue.  Drawing on emerging evidence that some putatively hard-wired heuristics are revisable, we explain why cultivating epistemic virtue requires, in part, practice.  Agents must actively retrain their epistemic dispositions to become more reliable.  Certain heuristics do seem relatively fixed, however.  We argue, therefore, that distinctive training techniques must be implemented to correct for such incorrigible dispositions.  In particular, one can cultivate an ability to identify when to override one’s “gut feelings,” and switch from an intuitive, heuristic-based mode of reasoning to a reflective mode of reasoning.  How best to cultivate these epistemic virtues is an ongoing empirical project, but we make some systematic claims regarding which types of heuristics will be more amenable to which interventions, and why.  In doing so, we hope to enrich philosophical accounts of epistemic virtue, and defend the notion against a battery of recent empirically-based attacks.

Future Research

The Costs of Profiling: Social, Ethical, and Cognitive Consequences

In the opinion issued in Floyd v. City of New York, Judge Shira Scheindlin wrote of the NYPD’s stop-and-frisk program, “This case is about the tension between liberty and public safety in the use of a proactive policing tool.”  Such statements portray the debate about racial and ethnic profiling as a clash between opposing moral perspectives on crime and terrorism.  Defenders appeal to dramatic reductions in crime; opponents appeal to violations of rights.  Scheindlin conceded that stop-and-frisk might reduce crime, but insisted that its effectiveness was irrelevant to its constitutionality.  The result is a standoff between consequentialist or utilitarian considerations in favor of profiling, and deontological or rights-based considerations against it.

Conceding the effectiveness of profiling constitutes a grave oversight in building the case against it.  I plan to correct this oversight by developing an evidence-based analysis of the consequences of profiling.  My book will cover a series of interrelated questions in moral and political philosophy, race studies, legal theory, and cognitive and social psychology.  In addition to the central question how such practices affect crime, I will examine the social, cognitive, and behavioral consequences on the targets and perpetrators of profiling, as well as on the wider community.  Given what we know, I will argue that profiling is an ineffective, sometimes even counterproductive, policing tool, which brings a host of social and psychological costs in tow. 

Chapter 1 surveys the prevailing arguments in favor of the ethicality and legality of profiling, and the evidence that suggests it helps to effectively capture and deter criminals.  I also examine the remarkable commonplace that an empirical analysis of profiling is unnecessary, because “it follows mathematically” that selectively targeting “a smaller but higher crime group” will reduce crime.[1]  Finally, I consider defenses of profiling on deontological grounds, especially appeals to the rights of would-be victims of crime.

Chapter 2 examines arguments against profiling.  In addition to traditional concerns about rights violations, I examine the consequentialist arguments raised against profiling, e.g., that profiling can be profoundly “demeaning,” and fosters distrust between police and the community.[2]  These are reasonable, but regrettably underdeveloped, positions.  Because damaged trust and demeaning experiences are qualitative harms, they are difficult to identify and measure, and evidence for them has been largely anecdotal and correlational (rather than causal).

Chapter 3 explores the sophisticated tools social scientists are developing to predict and measure the causes and effects of profiling.  Jack Glaser has developed mathematical simulations that predict that profiling is ineffective in some cases and counterproductive in others.[3]  In particular, profiling is apt to create or exaggerate criminal justice disparities, such that rates of imprisonment exceed any genuine disproportions in rates of offense.  Researchers have also developed measures of our automatic tendencies to perceive people in stereotypical ways.  At UC-Berkeley, I have been integrally involved in one such project with Glaser, designing a computer task that assesses whether individuals judge that black men “look more suspicious” than white men who perform the same actions.  We hope the task can become a useful tool for police officers, and increase accuracy and objectivity in decisions to stop civilians.

Chapter 4 examines the effects of profiling on the individuals and communities subject to it.  Emerging empirical research suggests that suffering disproportionate police scrutiny can be profoundly wounding.  Rather than being a deterrent, the stress and anger caused by profiling may increase self-destructive and risk-taking behavior.[4]  To analyze profiling’s wide-ranging harms, I adopt an intersectional approach, according to which race, class, and gender interact to explain the injuries of profiling and the potential for antagonism in police-civilian encounters.

Chapter 5 considers the effects of profiling on those who engage in it, as members of institutions that sponsor or tacitly permit it.  The institutional sanction of such practices can desensitize and sap individuals’ empathy for those they intend to protect.  It can, moreover, encourage them to perceive structural inequality and poverty as permissible and perhaps even deserved.  I also draw on research that illuminates how assurances from authority figures can undermine subordinates’ sense of personal accountability for objectionable behavior.

Discrimination harms not just those who perpetrate and undergo it, but even those who merely observe it.[5]  Chapter 6 argues that profiling has important consequences for those who simply live with and around it.  Citizens may unwittingly internalize stereotypes about a racial group simply by observing others who suffer disproportionate police scrutiny.  In fact, in one study on college students, profiling had a “reverse deterrent” effect, whereby members of the non-profiled, majority group became more likely to engage in illicit behavior, leading to an overall increase in misconduct.[6]

Chapter 7 turns to individual and institutional reforms.  I contest the widespread conviction that, on some level, profiling is inevitable, because racial stereotypes are “so ingrained” that individuals cannot help but use them.  Drawing on prior work, I explain what institutions and individuals can do to become “de-biased.”  I describe concrete strategies that law enforcement agents (and ordinary people) can implement to limit the influence of race on perceptions of criminality, and to guide their attention toward more reliably predictive features.  I also lay out concrete proposals for how civilians can navigate potentially antagonistic encounters with police, and mitigate the harms of being profiled.  I conclude with a variety of institutional interventions that promise to reduce bias and promote public safety while simultaneously respecting civil liberty.  I will also develop a companion website that lays out these strategies in publicly accessible terms.

Philosophically, I am sympathetic to the concern that profiling violates individual’s rights and dignity, but the refusal to countenance evidence regarding its effectiveness strikes defenders as naïve.  (Consider, e.g., the defiant and inflammatory response of New York’s mayor to Scheindlin’s ruling.[7])  The moral burden falls to opponents to show that the efficacy of profiling is no self-evident “mathematical fact,” but is, rather, empirically unfounded.  Racial profiling is ineffective for the narrow aim of reducing crime, and brings in its wide wake a litany of destructive social and psychological consequences.

Here is a proposal for a seminar I'd like to give on these issues.

Here is an informative and comprehensive site on "stop and frisk" policing practices.


[1] Arkes & Tetlock (2004, p.272), Psychological Inquiry.

[2] Scheindlin (August 12th 2013, p.3).

[3] Glaser (2006), Journal of Policy Analysis and Management; Glaser (forthcoming), Oxford University Press.

[4] Jamieson et al. (2013), Psychological science; Chen & Bargh (1997), Journal of Experimental Social Psychology.

[5] Cortina et al. (2011), Journal of Management.

[6] Hackney and Glaser (forthcoming), Law and Human Behavior.

[7] Gray et al. (August 16th, 2013), “Bloomberg blasts stop-and-frisk judge as ‘some woman’ who knows ‘zero’ about cops.” New York Daily News.