Last updated October 9, 2016.
Overview
I study the unconscious mind. Conscious experience tells a convincing story about what we like or dislike, but often this story is disconnected from how we act. For example, a judge may explicitly value racial equality but give a Black offender a harsher sentence than a similar White offender for the same crime. This disconnect between what we think and how we act can be captured in measures of implicit associations, which assess aspects of peoples’ minds that are outside of conscious awareness or control. Using a variety of methods including traditional experiments, meta-analysis, large-scale research contests, and longitudinal studies, I study how to change implicit associations to better align them with our explicit values and actions. This work has included studying when implicit associations are malleable and amenable to change; when implicit associations show remarkable stability; and when changes in implicit associations translate into behavioral change.
Malleability of Implicit Associations
The gap between explicit and implicit biases can be traced to differences in how they form and change. A persuasive argument that dramatically changes explicit attitudes and beliefs may have little effect on implicit associations. Conversely, classical conditioning linking Black people with good things and White people with bad things can change implicit associations but have no effect on what we consciously believe. In my research program, I have sought to map out the causes for malleability in implicit associations. This ambition has led to investigations on many mechanisms for changing implicit biases, including perspective-taking, reflecting on multicultural values, and efforts at the self-regulation of bias (Calanchini, Sherman, Klauer, & Lai, 2014, Personality and Social Psychology Bulletin). This pursuit has also led me to explore the emotions underlying implicit prejudice toward gay people (Lai, Haidt, & Nosek, 2014, Cognition & Emotion). Prior research found that the emotion of disgust leads to the degradation of others, particularly gay people. Could experiencing the theoretically opposite emotion of moral elevation un-do disgust toward gay people? In contrast to disgust, moral elevation arises from witnessing acts of moral virtue such as gratitude or generosity. I went to see if experiencing moral elevation reduces implicit and explicit anti-gay prejudice, and I found that it does. These findings suggest that moral elevation can be deployed for reducing prejudice.
Hundreds of studies from the past twenty years have sought to change implicit associations, spanning many areas of interest such as ethnic discrimination, social anxiety, dieting, and marketing. And yet, little work has empirically synthesized this evidence to understand when implicit bias will change (or not). In a forthcoming meta-analysis with over 500 experiments that will be one of the largest ever conducted in psychology (Forscher[*], Lai*, et al., under revision, Psychological Bulletin), we compared fourteen approaches to implicit bias change. We found approaches that exposed people to counter-stereotypical information and gave strategies for overriding bias were the most reliable for reducing implicit biases. Conversely, the most reliable approaches for increasing pre-existing implicit biases were exposure to stereotype-confirming information and mental fatigue. Some approaches were also inconsistently effective in the aggregate, such as inducing emotions (e.g., moral elevation), threat, and affirmation.
My meta-analysis reveals what categories of interventions work at changing implicit biases, and individual experiments like the moral elevation studies reveals which interventions can do so. But to understand the forces that are most influential for changing implicit associations, we must learn which specific strategies are most effective. To accomplish this, I organized a research contest to test many interventions simultaneously and find the most effective approaches for reducing implicit racial biases (Lai et al., 2014, Journal of Experimental Psychology: General). I asked researchers to submit a brief intervention they thought would be best at reducing implicit pro-White/anti-Black attitudes and received a total of seventeen interventions (and one sham intervention) that were tested over four studies (total N = 17,021). Direct comparisons of interventions were possible by using the same sample, setting, dependent variables, analysis strategies, and control conditions.
Nine of the eighteen interventions were effective at reducing implicit racial prejudice immediately. As with the meta-analysis, potent interventions tended to give people counter-stereotypical information or strategies to override bias. The best interventions were often highly emotional, vivid, and self-relevant. In the most powerful intervention, participants imagined themselves being kidnapped by a malicious White criminal, only to be rescued by a young Black man. This intervention shows that a countervailing experience that counters what is often seen in popular media can have a profound effect on implicit biases.
The research contest demonstrates that large-scale comparisons of many interventions can teach us much about the social forces that are most influential on implicit associations. It also shows how researchers can make substantive progress on pressing social-psychological questions. In future research, I will expand on findings from these studies to develop more effective interventions and apply this research contest method to other pressing social-psychological questions.
Stability of Implicit Associations over Time
It is commonly assumed that immediate changes in implicit bias will generalize to persistent changes over time. And yet, my meta-analysis finds that over 200 interventions have been published with the goal of reducing implicit prejudice and stereotypes, but less than 10% of those studies have tracked their effects past one study session (Lai, Hoffman, & Nosek, 2013, Social and Personality Psychology Compass; Forscher*, Lai*, et al., under revision, Psychological Bulletin). Of the existing long-term studies, the results are mixed. It seems that we know less about the durability of change in implicit associations than previously thought. So, I sought to see if effective interventions from the research contest were effective in the long-term (Lai et al., 2016, Journal of Experimental Psychology: General). And if so, I wanted to know if the ranking of interventions would shift. For example, interventions that gave strategies to override biases could lead to habits that successfully down-regulate bias over time, but interventions that targeted associations directly through conditioning could show rebound effects after re-exposure to cultural stereotypes in everyday life.
With collaborators across the United States and Canada, I tested the nine effective interventions from the original research contest again in two longitudinal experiments. In all, we gathered over 6,000 participants from eighteen universities. I found that all nine interventions were successful at reducing implicit prejudice immediately (as in prior studies), but that none continued to have an effect after twenty-four hours. While much progress has been made in temporarily reducing implicit biases, these findings suggest that implicit biases are more stable than previously thought. In future research, I will test whether more intensive interventions will be effective in reducing implicit bias in the long-term. I will also apply insights from research on effective learning and memory retention (e.g., the spacing effect, retrieval practice) to see if those ideas can be used to enhance the effects of interventions for changing implicit bias.
The Relationship Between Implicit Associations and Behavior
An unexpected set of results from the meta-analysis came from mediation models examining the relationship between implicit bias, explicit bias, and behavior. I found that experimentally induced changes in implicit biases did not statistically mediate corresponding changes in explicit biases or behavioral outcomes. These findings go against fundamental assumptions about the causal role of implicit bias in conscious cognition and behavior. In my future research, I will gather evidence to develop a revised theory for how implicit associations relate to conscious thought and action. By one theoretical account, implicit associations are like a “sociometer” (Leary & Baumeister, 2000) that tracks changes in the social environment rather than an engine for changing behavior. Implicit biases as assessed by implicit measures may reflect the leftover mental residue of experiences that an individual has had, rather than being an active causal force in-itself. By one methodological account, researchers have asked too much of implicit measures. The same Race Attitude Implicit Association Test (IAT) has been used to predict many outcomes, including physiological arousal, interpersonal distance, willingness to have an interracial relationship, support for affirmative action, and hiring discrimination (e.g., Lai et al., 2014, Journal of Experimental Psychology: General). Following the resolution of the explicit attitude-behavior crisis in the 1970s, I will study whether implicit measures that are more closely tied to the behaviors they are intended to predict are more successful (e.g., a Black politician vs. White politician IAT for predicting political support).
Future Directions
The social psychologist Kurt Lewin once said, “If you want truly to understand something, try to change it.” That ethic has animated my approach to conducting research and my interest in understanding implicit bias. I conducted large-scale comparative studies to find the best interventions for changing implicit associations. At the same time, I found that the best interventions’ effects were short-lived. With one of the largest meta-analyses in psychology, I found the relationship between implicit biases and behavior to be more tenuous than previously believed. These findings provoke interesting questions for my future research. How do implicit associations change? What are the forces that create long-term sustained change in implicit associations? What explains the link between implicit biases and how we consciously think or act? I don’t yet know the answers, but I hope to make progress on these questions and more in the years to come.
* Co-lead-authors.
Overview
I study the unconscious mind. Conscious experience tells a convincing story about what we like or dislike, but often this story is disconnected from how we act. For example, a judge may explicitly value racial equality but give a Black offender a harsher sentence than a similar White offender for the same crime. This disconnect between what we think and how we act can be captured in measures of implicit associations, which assess aspects of peoples’ minds that are outside of conscious awareness or control. Using a variety of methods including traditional experiments, meta-analysis, large-scale research contests, and longitudinal studies, I study how to change implicit associations to better align them with our explicit values and actions. This work has included studying when implicit associations are malleable and amenable to change; when implicit associations show remarkable stability; and when changes in implicit associations translate into behavioral change.
Malleability of Implicit Associations
The gap between explicit and implicit biases can be traced to differences in how they form and change. A persuasive argument that dramatically changes explicit attitudes and beliefs may have little effect on implicit associations. Conversely, classical conditioning linking Black people with good things and White people with bad things can change implicit associations but have no effect on what we consciously believe. In my research program, I have sought to map out the causes for malleability in implicit associations. This ambition has led to investigations on many mechanisms for changing implicit biases, including perspective-taking, reflecting on multicultural values, and efforts at the self-regulation of bias (Calanchini, Sherman, Klauer, & Lai, 2014, Personality and Social Psychology Bulletin). This pursuit has also led me to explore the emotions underlying implicit prejudice toward gay people (Lai, Haidt, & Nosek, 2014, Cognition & Emotion). Prior research found that the emotion of disgust leads to the degradation of others, particularly gay people. Could experiencing the theoretically opposite emotion of moral elevation un-do disgust toward gay people? In contrast to disgust, moral elevation arises from witnessing acts of moral virtue such as gratitude or generosity. I went to see if experiencing moral elevation reduces implicit and explicit anti-gay prejudice, and I found that it does. These findings suggest that moral elevation can be deployed for reducing prejudice.
Hundreds of studies from the past twenty years have sought to change implicit associations, spanning many areas of interest such as ethnic discrimination, social anxiety, dieting, and marketing. And yet, little work has empirically synthesized this evidence to understand when implicit bias will change (or not). In a forthcoming meta-analysis with over 500 experiments that will be one of the largest ever conducted in psychology (Forscher[*], Lai*, et al., under revision, Psychological Bulletin), we compared fourteen approaches to implicit bias change. We found approaches that exposed people to counter-stereotypical information and gave strategies for overriding bias were the most reliable for reducing implicit biases. Conversely, the most reliable approaches for increasing pre-existing implicit biases were exposure to stereotype-confirming information and mental fatigue. Some approaches were also inconsistently effective in the aggregate, such as inducing emotions (e.g., moral elevation), threat, and affirmation.
My meta-analysis reveals what categories of interventions work at changing implicit biases, and individual experiments like the moral elevation studies reveals which interventions can do so. But to understand the forces that are most influential for changing implicit associations, we must learn which specific strategies are most effective. To accomplish this, I organized a research contest to test many interventions simultaneously and find the most effective approaches for reducing implicit racial biases (Lai et al., 2014, Journal of Experimental Psychology: General). I asked researchers to submit a brief intervention they thought would be best at reducing implicit pro-White/anti-Black attitudes and received a total of seventeen interventions (and one sham intervention) that were tested over four studies (total N = 17,021). Direct comparisons of interventions were possible by using the same sample, setting, dependent variables, analysis strategies, and control conditions.
Nine of the eighteen interventions were effective at reducing implicit racial prejudice immediately. As with the meta-analysis, potent interventions tended to give people counter-stereotypical information or strategies to override bias. The best interventions were often highly emotional, vivid, and self-relevant. In the most powerful intervention, participants imagined themselves being kidnapped by a malicious White criminal, only to be rescued by a young Black man. This intervention shows that a countervailing experience that counters what is often seen in popular media can have a profound effect on implicit biases.
The research contest demonstrates that large-scale comparisons of many interventions can teach us much about the social forces that are most influential on implicit associations. It also shows how researchers can make substantive progress on pressing social-psychological questions. In future research, I will expand on findings from these studies to develop more effective interventions and apply this research contest method to other pressing social-psychological questions.
Stability of Implicit Associations over Time
It is commonly assumed that immediate changes in implicit bias will generalize to persistent changes over time. And yet, my meta-analysis finds that over 200 interventions have been published with the goal of reducing implicit prejudice and stereotypes, but less than 10% of those studies have tracked their effects past one study session (Lai, Hoffman, & Nosek, 2013, Social and Personality Psychology Compass; Forscher*, Lai*, et al., under revision, Psychological Bulletin). Of the existing long-term studies, the results are mixed. It seems that we know less about the durability of change in implicit associations than previously thought. So, I sought to see if effective interventions from the research contest were effective in the long-term (Lai et al., 2016, Journal of Experimental Psychology: General). And if so, I wanted to know if the ranking of interventions would shift. For example, interventions that gave strategies to override biases could lead to habits that successfully down-regulate bias over time, but interventions that targeted associations directly through conditioning could show rebound effects after re-exposure to cultural stereotypes in everyday life.
With collaborators across the United States and Canada, I tested the nine effective interventions from the original research contest again in two longitudinal experiments. In all, we gathered over 6,000 participants from eighteen universities. I found that all nine interventions were successful at reducing implicit prejudice immediately (as in prior studies), but that none continued to have an effect after twenty-four hours. While much progress has been made in temporarily reducing implicit biases, these findings suggest that implicit biases are more stable than previously thought. In future research, I will test whether more intensive interventions will be effective in reducing implicit bias in the long-term. I will also apply insights from research on effective learning and memory retention (e.g., the spacing effect, retrieval practice) to see if those ideas can be used to enhance the effects of interventions for changing implicit bias.
The Relationship Between Implicit Associations and Behavior
An unexpected set of results from the meta-analysis came from mediation models examining the relationship between implicit bias, explicit bias, and behavior. I found that experimentally induced changes in implicit biases did not statistically mediate corresponding changes in explicit biases or behavioral outcomes. These findings go against fundamental assumptions about the causal role of implicit bias in conscious cognition and behavior. In my future research, I will gather evidence to develop a revised theory for how implicit associations relate to conscious thought and action. By one theoretical account, implicit associations are like a “sociometer” (Leary & Baumeister, 2000) that tracks changes in the social environment rather than an engine for changing behavior. Implicit biases as assessed by implicit measures may reflect the leftover mental residue of experiences that an individual has had, rather than being an active causal force in-itself. By one methodological account, researchers have asked too much of implicit measures. The same Race Attitude Implicit Association Test (IAT) has been used to predict many outcomes, including physiological arousal, interpersonal distance, willingness to have an interracial relationship, support for affirmative action, and hiring discrimination (e.g., Lai et al., 2014, Journal of Experimental Psychology: General). Following the resolution of the explicit attitude-behavior crisis in the 1970s, I will study whether implicit measures that are more closely tied to the behaviors they are intended to predict are more successful (e.g., a Black politician vs. White politician IAT for predicting political support).
Future Directions
The social psychologist Kurt Lewin once said, “If you want truly to understand something, try to change it.” That ethic has animated my approach to conducting research and my interest in understanding implicit bias. I conducted large-scale comparative studies to find the best interventions for changing implicit associations. At the same time, I found that the best interventions’ effects were short-lived. With one of the largest meta-analyses in psychology, I found the relationship between implicit biases and behavior to be more tenuous than previously believed. These findings provoke interesting questions for my future research. How do implicit associations change? What are the forces that create long-term sustained change in implicit associations? What explains the link between implicit biases and how we consciously think or act? I don’t yet know the answers, but I hope to make progress on these questions and more in the years to come.
* Co-lead-authors.