There are many reasons why the implicit bias construct took root in everyday conversation, but one of them is that millions of people have experienced the most commonly used measure of implicit bias – the Implicit Association Test (IAT; Greenwald, McGhee, & Schwartz, 1998) at the Project Implicit websites. Project Implicit is a non-profit organization and international research collaboration between behavioral scientists interested in implicit social cognition. The organization’s primary public contribution is its education websites (https://implicit.harvard.edu) where more than twenty-eight million IATs have been completed. This chapter provides an overview of Project Implicit and the contributions and challenges of more than twenty years of internet-based data collection on implicit attitudes and stereotypes. The first section describes Project Implicit’s history and organizational structure; next, some of the key insights gleaned from the data collected at the Project Implicit websites are reviewed. These include assessment of the pervasiveness and correlates of implicit bias, comparisons across time and by geographic area, and reactions to learning about one’s own implicit bias. Finally, we reflect on some of the challenges of being uniquely situated between academic researchers and the general public, and describe how changing scientific knowledge has changed scientific communication about implicit bias.

We offer a conceptual framework by which to consider implicit bias. In contrast to a far too common presumption that implicit bias involves unconscious attitudes and stereotypes, i.e., ones for which individuals lack awareness, we emphasize a view of implicit bias as an effect of attitudes of which individuals are unaware. The perspective is grounded in decades of social psychological theory and research concerning the constructive nature of perception and the potential biasing influence of attitudes on perceptions and judgments. Attitudes that are automatically activated from memory can exert such a biasing influence, without individuals’ awareness that they have been affected. We articulate the advantages of such a perspective for both the science and the politics of implicit bias. We also discuss how individuals can overcome the influence of an automatically activated attitude, given appropriate motivation and opportunity to do so, and briefly review evidence concerning the joint influence of these factors on prejudicial judgments and behavior.

Implicit bias has always been understood as an individual attitude that is rooted in one’s social environment. However, in practice, the field has focused more heavily on the individual attitude, to the neglect of the social environment. In this chapter, we describe an alternative view of implicit bias – the Bias of Crowds model – that reinterprets implicit bias as a feature of social contexts more than persons. In doing so, we argue that, akin to the “wisdom of crowds” effect, implicit bias may emerge as the aggregate effect of individual fluctuations in concept accessibility that are transitory and context-dependent. We also explain how this novel interpretation of implicit bias resolves long-standing concerns regarding the temporal instability and weak predictive validity of implicit attitudes measures. Finally, we review direct empirical tests of the model and its predictions and consider future avenues for research, as well as theoretical and practical implications.

A recent debate on implicit measures of racial attitudes has focused on the relative roles of the person, the situation, and their interaction in determining the measurement outcomes. The chapter describes process models for assessing the roles of the situation and the person-situation interaction on the one hand and stable person-related components on the other hand in implicit measures. Latent state-trait models allow one to assess to what extent the measure is a reliable measure of the person and/or the situation and the person-situation interaction (Steyer, Geiser, & Fiege, 2012). Moreover, trait factor scores as well as situation-specific residual factor scores can be computed and related to third variables, thereby allowing one to assess to what extent the implicit measure is a valid measure of the person and/or the situation and the person-situation interaction. These methods are particularly helpful when combined with a process decomposition of implicit-measure data such as a diffusion-model analysis of the IAT (Klauer, Voss, Schmitz, & Teige-Mocigemba, 2007).

On the basis of dual-process theories, we propose a model that accounts for the (lack of) convergence between explicit and implicit dispositions, the effects of explicit and implicit dispositions on controlled and automatic behavior, and changes in explicit dispositions that are based on self-observed automatic behavior. The model is characterized by nine direct effects among the constructs and measures that are included the model. Importantly, in this model, each effect can be moderated by characteristics of persons, situations, and person x situation interactions. As a general theoretical framework, the model can be applied to a variety of individual difference constructs such as personality traits, values, norms, attitudes, prejudice, beliefs, stereotypes, judgment bias, and discrimination. We present evidence from our research and from studies by other researchers that speaks to the validity and usefulness of the model. We propose that implicit–explicit consistency should be considered a variable in itself and demonstrate its usefulness for understanding some findings that have been reported in the self-concept and self-esteem literatures. We make suggestions for future research with an emphasis on the discrimination of automatic and controlled behavior using facial expressions of emotion as an example.

During the past century, racial attitudes in America have been radically transformed. One hundred years ago, this was a country of explicit racism, where separation of the races and discrimination against African Americans in particular were normative, formalized in laws, in the widespread practices of businesses and in the treatment of individuals by individuals every day. The civil rights movement of the 1960s brought about a landmark shift, eliciting widespread condemnation of racism, and setting the stage for the country’s embracing of multiculturalism and implementing policies in many arenas of life to level the playing field and compensate for past discrimination. These changes in public practices were accompanied by a gradual transformation of public opinion in the United States: surveys documented a steady growth of endorsement of racial equality and a decline in explicitly stated racial prejudice. More and more Americans endorsed principles of racial equality and expressed support for various policies preventing discrimination.

In contrast to most scientific research that goes largely unrecognized by the general public, the concept of implicit bias broke through into the public sphere. This success comes with the challenge that academic nuances and clearly stated limitations often get lost in translation. Moreover, given ongoing scientific debates about what implicit bias is and how to measure it, perhaps the phenomenon got out into public consciousness before scientists have fully understood it.

In the study of racial prejudice in America, symbolic racism (and its close cousin, racial resentment) has been especially successful at predicting evaluations of race-related policies, evaluations of African-American politicians, voting behavior, and much more. This paper tests a proposal made by the theory of symbolic racism about the origin of racial prejudice: that symbolic racism is a blend of anti-Black affect and the perception that Black people violate traditional American values. Analyzed using a new approach that more fully meets the conceptualization of value-violation beliefs than in past research, data from college students and from a representative national sample of Americans disconfirmed the blend hypothesis. Instead, the data are consistent with a mediational chain: beliefs that Black people violate traditional values mediate the effect of anti-Black affect on responses to symbolic racism items, which, in turn, shape people’s attitudes toward racial policies. Thus, the previously suggested “blending” of proposed ingredients appears to be mediational rather than interactive or synergistic. These findings cast new light on the origins of symbolic racism.

The assessment of racial attitudes remains central to social science research, yet researchers differ widely in how they are measured. There is an ongoing debate over whether it is possible to assess racial attitudes, directly leading some researchers to develop measures of new racism such as modern racism and others to abandon the explicit assessment of racial negativity altogether in favor of implicit measures. Nonetheless, explicit measures of racial negativity remain pervasive in social and political psychological research. But unlike implicit attitudes, there is no consensus on the best way in which to measure them. In this chapter, we document current diversity in the measurement of explicit racial attitudes and demonstrate that component scale items can be divided empirically into three distinct concepts. Not all three concepts clearly reflect racial animosity, however. We map these three concepts onto racial resentment, a widely used measure of new racism, to demonstrate its questionable status as a measure of racial negativity. We conclude by suggesting the adoption of overt racism measures in psychological race-related research and urge for greater uniformity in the assessment of explicit racial attitudes.

Scholars have long recognized that successful prediction of behavior on the basis of explicit attitudes depends on the correspondence between the attitude measure and the focal behavior. Fishbein and Ajzen (2010) argued that behaviors vary in terms of their action, target, context, and time, and that the prediction of specific behaviors is greatly enhanced when explicit attitude measures reflect these features of the to-be-predicted behavior. We argue that the same principle applies in the case of predicting behavior from implicit attitudes, and we review relevant evidence relating to each of Fishbein and Ajzen’s parameters. Special attention is paid to the target parameter, given increasing awareness of the intersectional nature of bias. A global race bias may not extend equally to all members of a particular racial identity, and cross-cutting factors such as gender, age, or sexuality may qualify the extent to which global measures of race bias predict discriminatory behavior toward particular individuals.

Eight methodological issues relevant to improving the quality of research on implicit attitudes are considered. These include (1) formulating and implementing strong psychometric models for implicit attitude measures, (2) using modern theories of explicit attitudes as a base from which to test key propositions about implicit attitudes, (3) using sound psychometric practices to assess explicit attitudes, (4) addressing the problem of endogeneity, (5) addressing ecological fallacies when pursuing aggregate analyses of implicit attitudes, (6) evaluating the magnitude of effect sizes, (7) using structural equation modeling in implicit attitude research, and (8) using proper moderation analysis and incremental explained variance analysis in meta-analyses. The central elements of each problem are described and recommendations for addressing them are provided.

How do implicit and explicit racial attitudes compare in their ability to predict political attitudes and behaviors? Data from existing studies suggest that implicit measures may be less relevant than explicit ones for predicting vote choice. This chapter replicates that result using data from 2008 and 2012 and considers whether the dominance of explicit measures in this domain can be attributed to the fact that voting is a highly considered action, wherein individuals may have taken steps to mitigate their own biases. To assess this, we use nationally representative panel survey data to examine whether the relative dominance of explicit measures over the Affect Misattribution Procedure was similarly true across the campaign season and for alternative outcomes that may have encouraged less cognitive control than voting. Results indicate that explicit measures were more predictive for the vast majority of political outcomes. This raises questions about the added value of considering implicit measures in addition to explicit ones when measuring political attitudes and behaviors.

Implicit measures were introduced to explain phenomena that are characterized by a gap between self-reported attitudes and behavior. Recent meta-analyses revealed, however, that implicit measures have only limited predictive validity that goes beyond self-reports. We identify possible reasons for this failure: (a) A lack of validity that is due to the influence of extraneous processes, (b) a focus on evaluation instead of motivation, (c) a focus on associations instead of propositional beliefs, and (d) a focus on global instead of context-dependent attitudes. Recent developments in the field of implicit measures addressed these problems: (a) Statistical process models increase the internal validity of implicit measures, (b) implicit measures of wanting have the potential to predict behavior better than implicit measures of liking, (c) new paradigms provide measures of automatically activated attitudes for propositions that have an unambiguous interpretation, and (d) assessment of context-dependent beliefs is better suited to predict specific behaviors. Incorporating these developments into research on implicit bias will help to realize the initial expectations of describing, explaining, and predicting behavior in many situations.

While schema theory motivated the original measures of automatic cognitive associations between constructs in memory, researchers soon modified these to explore a different domain: implicit attitudes about social groups that elude standard self-reports. As the so-called implicit attitude revolution gained steam, the original measurement goal got much less attention, especially in political science. We believe the schema concept – automatic cognitive associations between features of an attitude object – continues to hold great value for political psychology. We offer a retrofit of the popular implicit association test (IAT), one more efficient than many lexical tasks, to tap these associations in surveys. The new technique captures the degree to which citizens link ideas about ostensibly group-neutral policies to specific social categories. We use this measurement strategy to explore the psychological mechanisms underlying group centrism in politics, an effort that has been largely abandoned due to measurement difficulties. Results from four studies offer practical suggestions about the application of implicit measures for capturing the automatic ways people link groups to important political objects. We conclude by discussing the broader promise of implicit measurement of group schemas, not just implicit affect, for political psychology.