People may be unaware of their own important underlying attitudes and beliefs, particularly in the domains of stereotyping and prejudice. Nevertheless, these "implicit biases" significantly influence inter-group behavior in everyday social interactions. Indeed, the influence of these biases often is greater than that of explicitly stated inter-group attitudes and beliefs, affecting important outcomes for members of minority groups. As such, it is critical to understand the nature and operation of implicit bias, as well as the factors that may change them or reduce their influence. To achieve these goals, Dr. Jeffrey Sherman of the University of California - Davis uses mathematical modeling with a tool called the Quad model that he developed in his past research. Application of this tool has shown that implicit attitudes and beliefs sometimes may reflect the automatic activation of biased associations in memory or may reflect failures to regulate the influences of such associations on behavior. The Quad model can be used to estimate the independent roles of each of these factors in producing biased and unbiased behavior.
There are four specific goals in this research. First, the Quad model will be applied to shed light on the contextual and individual factors that influence the extent of implicit bias. Implicit bias has been shown to vary significantly across social contexts and across individuals, often reflecting personal goals and motives. This research will model contextual and individual differences in implicit bias in order to better understand the factors that increase or decrease the automatic activation of biased associations and the ability to regulate the expression of those associations. The second goal is to apply a "treatment" approach to reducing implicit bias. In these studies, interventions that are designed specifically to influence either underlying associations or the ability to regulate those associations will be applied to individuals and contexts associated with enhanced activation of associations or diminished ability to self-regulate. The goal is to show how interventions can be tailored to address specific deficits in processing (over-activation of associations vs. failure of regulation) associated with increases in implicit bias. The third goal is to use the Quad model to better understand the relationships among different measures of stereotyping and prejudice, which frequently fail to correspond to one another. The present research proposes that dissociations among these different measures may reflect differences in the extents to which the measures reflect the automatic activation of associations versus the failure to regulate the associations. Application of the model will help to specify when and why different measures will and will not produce corresponding results. Finally, the fourth goal is to use the Quad model to better predict behavior in inter-group settings. Measures of stereotyping and prejudice are sometimes poor predictors of people's actual behavior towards members of minority groups. The Quad model can improve the ability of these measures to predict behavior by independently assessing the roles of automatic associations, the ability to regulate those associations, and interactions between these components.
In sum, the purpose of this research is to improve the measurement of stereotyping and prejudice, increase understanding of the factors that increase or decrease such biases, and improve the ability of measures of bias to predict people's behavior. Implicit attitudes influence behavior in important domains of life including law enforcement, health, and employment. It is therefore critical to gain a better understanding of their nature and operation.
People may be unaware of important underlying attitudes and beliefs, particularly in the domains of stereotyping and prejudice. Nevertheless, these "implicit biases" significantly influence inter-group behavior in everyday interpersonal interactions, in health care settings, in legal settings, and in other important domains. Indeed, the influence of these biases often is greater than that of explicitly stated inter-group attitudes and beliefs, affecting important outcomes for members of minority groups. As such, it is critical to understand the nature and operation of implicit bias, as well as the factors that may change them or reduce their influence. To achieve these goals, the present research relied on mathematical modeling with a tool called the Quad model that has been developed in my past research. Application of this tool has shown that implicit attitudes and beliefs sometimes may reflect the automatic activation of biased associations in memory or may reflect failures to regulate the influences of such associations on behavior. The Quad model can be used to estimate the independent roles of each of these factors in producing biased and unbiased behavior. There were three specific scientific contributions from this research. First, the Quad model was applied to shed light on the contextual and individual factors that influence the extent of implicit bias. This research examined contextual and individual differences in implicit bias in order to better understand the factors that increase or decrease the automatic activation of biased associations and the ability to regulate the expression of those associations. For example, we showed that threats to self-esteem can increase intergroup bias, and that this effect is related to increased activation of negative associations with outgroups but not to decreased effort to inhibit negative feelings about groups (which is a common explanation for the esteem effect). Second, the Quad model was used to better understand the relationships among different measures of stereotyping and prejudice, which frequently fail to correspond to one another. Our research proposes that dissociations among these different measures may reflect differences in the extents to which the measures reflect the automatic activation of associations versus the failure to regulate the associations. Application of the model has, indeed, shown that when the underlying processes of the different measures are directly compared, correspondence among the measures increases. This is an important methodological advance, as it improves our ability to accurately measure attitudes and beliefs and our ability to successfully predict behavior with implicit measures of bias. Finally, the Quad model was applied to better predict behavior in inter-group settings. Measures of stereotyping and prejudice are sometimes poor predictors of people’s actual behavior towards members of minority groups. The Quad model can improve the ability of these measures to predict behavior by independently assessing the roles of automatic associations, the ability to regulate those associations, and interactions between these components. In fact, we showed in a number of studies that we could predict both social judgments and behaviors more effectively by using the Quad model to separate activation of bias and the ability to overcome bias. This research has the potential to have significant societal impacts. Implicit prejudices and stereotypes have been shown to influence important behavior, such as the shooting of unarmed suspects, eyewitness testimony, legal outcomes, and medical treatment disparities. Thus, more fully understanding the nature and workings of implicit attitudes is much more than a mere academic enterprise. This project contributed to the training and development of four Ph.D. candidates and dozens of undergraduate research assistants, some of whom when on to Ph.D. programs. The findings were presented in public settings, such as NPR, the California Science Center, and in conferences addressed toward the general public. The findings also were presented at many professional conferences, both domestic and international, and in the publication of many articles and chapters. This research also enhanced the infrastructure for psychological research by developing a new instrument for analyzing data from implicit measures. In the past four years, I have been contacted dozens of times by and have assisted researchers working on addictions, phobias, interpersonal attractions, relationships, self-esteem, emotion regulation, social ostracism, cognitive development, and memory. The development and use of the Quad model will help to advance research in any domain in which implicit measures are being used (which is quite substantial at this point). To facilitate application of our methods, we have developed a web site with an instructional tutorial and example applications. In sum, the intellectual contributions of this research were to improve the measurement of stereotyping and prejudice, increase understanding of the factors that increase or decrease such biases, and improve the ability of measures of bias to predict people’s behavior. The human capital contributions were to train many promising students. The infrastructure contributions were to develop and share an important scientific tool.
