An Introduction to Implicit Bias

Knowledge, Justice, and the Social Mind

Erin Beeghly, Alex Madva, Erin Beeghly, Alex Madva

• Publisher: Routledge
• Year: 2020
• Print ISBN: 9781138092228
• eBook ISBN: 9781351607599
• Topic: Philosophy
• Subtopic: Philosophical Criticism

Written by a diverse range of scholars, this accessible introductory volume asks: What is implicit bias? How does implicit bias compromise our knowledge of others and social reality? How does implicit bias affect us, as individuals and participants in larger social and political institutions, and what can we do to combat biases? An interdisciplinary enterprise, the volume brings together the philosophical perspective of the humanities with the perspective of the social sciences to develop rich lines of inquiry. Its twelve chapters are written in a non-technical style, using relatable examples that help readers understand what implicit bias is, its significance, and the controversies surrounding it. Each chapter includes discussion questions and additional annotated reading suggestions, and a companion webpage contains teaching resources. The volume is an invaluable resource for students—and researchers—seeking to understand criticisms surrounding implicit bias, as well as how one might answer them by adopting a more nuanced understanding of bias and its role in maintaining social injustice.

1 The Psychology of Bias: From Data to Theory

Gabbrielle M. Johnson

What’s going on in the head of someone with an implicit bias? Attempts to answer this question have centered on two problems: first, how to explain why implicit biases diverge from explicit attitudes and second, how to explain why implicit biases change in response to experience and evidence in ways that are sometimes rational, sometimes irrational. Chapter 1 introduces data, methods, and theories to help us think about these questions. First, the chapter briefly outlines the features of good, explanatory psychological theories built on empirical data, and the pitfalls they must avoid. Next, it presents an overview of the empirical data relevant to two main questions: implicit-explicit divergence and rationality. Finally, it surveys the theories intended to provide psychological explanations for those empirical data, providing examples of each. The chapter ends with some summarizing reflections, and in particular it confronts the possibility that bias is in fact a mixed-bag of many different sorts of psychological phenomena, making one unified psychological explanation misplaced.

1 The Psychology of Bias: A First Pass

What’s going on in the head of someone with an implicit bias? Often psychologists answer this question by saying such a person has an unconscious mental association. On this view, when we say someone has an implicit bias against the elderly, for example, we’re saying they quickly, automatically, and unconsciously associate someone’s being elderly with, say, that person’s being frail, forgetful, or bad with computers. This view of implicit bias comes very naturally to us, as we’re used to our minds making associations quickly, automatically, and without our conscious awareness. For example, when I say salt, you automatically think pepper; or when I say hip, you think hop; when I say Tweedledee, you think … . You don’t need to deliberate about what comes next; you just know (for more discussion, see Siegel, Chapter 5, “Bias and Perception”). These characteristics of the associationist picture may help us explain one of the most vexing aspects of implicit bias: divergence. Divergence occurs when our unconscious mental states differ, or diverge, from our consciously-held mental states. Consider, for example, an individual who, when asked, says women are just as capable of succeeding in leadership roles as men are but continues to act in ways that seem at odds with that sentiment. They might, for example, rate a male applicant for a leadership role more favorably than a female applicant with equally impressive credentials. In this case, we say that this individual’s explicit (or consciously accessible) beliefs about gender diverge from their implicit attitudes. At the conscious level, the person believes men and women are equally capable; however, at the implicit level, the person has a bias against women.

The associationist view provides a natural explanation for divergence: this person’s conscious beliefs diverge from their unconscious beliefs because there are two distinct and independent mental constructs involved at each level of consciousness. At the conscious level, this person has deliberately considered evidence and is convinced that men and women are equal. However, at the unconscious level, this person’s automatic, reflex-like processes lead them to associate male with leader. Because distinct and independent constructs operate at each level, we get different results depending on which level the individual relies on at any given time. I call this approach, which distinguishes between different kinds of states or processes for explicit and implicit attitudes, a dual-construct model. Dual-construct models excel at explaining divergence, or the differences between explicit and implicit attitudes.

Dual-construct models—like the associationist picture above—have gained favor among psychological accounts of implicit biases. However, more recently, interesting studies exploring the malleability of implicit biases—that is, our ability to change implicit attitudes—suggest that the operation of mental processes at the two levels might not be so different after all. In particular, our implicit attitudes sometimes change when confronted with reasons to do so. I’ll call this newly emerging data rationality of bias, or rationality for short.

The term ‘rationality’ is intended as a term of art here, which I’ll use to pick out particular features of mental states that I’ll explain in more detail later. This notion of rationality is intentionally more robust than you might initially think when hearing the term. For example, you might think that my having an association between salt and pepper is rational—it makes sense that I would think of one right after the other since they often appear together. However, this sort of superficial rationality won’t be enough to capture the unique features I want this technical notion to capture. I’ll return to this point and explain exactly what unique features I have in mind in Section 3.3.

The possibility that implicit biases might ever be rational, albeit rarely, is surprising for the dual-construct model, which predicts that rational and deliberative processes are unique to the explicit level and, thus, entirely absent at the implicit level. Even stronger, the fact that rational processes might be in operation at both the explicit and implicit levels suggests that the dual-construct model, which attributes to each level distinct and independent kinds of states and processes, might be mistaken. Instead, one might think the evidence for rationality suggests that implicit attitudes are just like, or at least similar to, ordinary explicit beliefs, save for one kind of belief is unconscious while the other is conscious. I’ll call these sorts of approaches belief-based models, because they equate the kinds of constructs leading to explicit and implicit attitudes. Belief-based models excel at explaining rationality, or similarities between explicit and implicit attitudes.

In this chapter, I discuss these two fact patterns—divergence and rationality—in detail. I begin by reviewing standard assumptions about psychological theories more generally, such as what they aim to do and how we evaluate them. Following this preliminary discussion, I review the empirical data indicating patterns of divergence and rationality, and I examine how the two main approaches—dual-construct models and belief-based models—are each sufficient to deal with one of the fact patterns, but struggle to explain both. I’ll then look at views that attempt to carve out a middle ground between dual-construct and belief-based models. These views argue that implicit biases constitute a unique kind of mental construct, which is not easily explained by either standard dual-construct or belief-based models.

2 What is a Psychological Explanation?

Roughly, psychology is a scientific discipline that aims to explain an intelligent creature’s behavior in terms of that creature’s mental states and processes. In other words, psychologists look to a creature’s state of mind in order to understand why they acted the way that they did.

One fundamental assumption among most psychologists today is that humans have mental states that represent the world as being a certain way and that those representations of the world affect how they think and act in it. For example, you might explain your roommate’s going to Chipotle for lunch using her belief that Chipotle makes the best guacamole. Of course, this belief might turn out to be false or your roommate might have gone to Chipotle for a different reason. But the kind of explanation you gave is what psychologists understand and expect.

This psychological methodology of building theories that explain by making reference to distinctively mental states—beliefs, desires, fears, intentions, etc.—is an example of what philosophers of science beginning with Thomas Kuhn (1962) call a paradigm. Within any paradigm, scientists take certain fundamental assumptions for granted as shared among members of a scientific community—in this case, the assumption that humans have mental states in the form of representations.

Alternative to this methodological paradigm was a different approach made popular by B.F. Skinner called behaviorism. It claimed that psychology should only study objective, observable physical stimuli and behavioral responses, and not concern itself with subjective, private mental states. In its most radical form, behaviorism claimed that all behaviors of intelligent creatures can ultimately be explained in this way, without ever needing to mention internal, distinctively mental states. Although no longer popular, behaviorism made several important contributions to the methodology of psychology.

One contribution is a general suspicion toward the ease of relying on mental state explanations (see Dennett 1981: 56 citing Skinner 1971: 195). The fear is that we can’t explain an unknown fact—why your roommate went to Chipotle for lunch—by using an equally mysterious object—her internal belief about Chipotle. Because her belief is a mental state, it is observable only by her and no one else. So we weren’t really explaining anything at all, merely replacing one mysterious fact with another.

This worry is sometimes called the ‘homunculus fallacy’. The word ‘homunculus’ (plural ‘homunculi’) is Latin for ‘little person’. A theory that commits this fallacy attempts to explain some intelligent behavior by way of positing some equally intelligent cause of that behavior. The idea is that this was tantamount to positing a little person inside the head of the first intelligent creature whose own behavior goes unexplained.

This same basic idea is depicted humorously in the 2015 Pixar film Inside Out. In this film, the perspective switches between that of a young girl, Riley, and the five personifications of the basic emotions that live in her head: Joy, Sadness, Fear, Disgust, and Anger. These five little people (or homunculi) inside Riley’s head operate a control center that influences all of Riley’s actions. According to the film, the explanation for why Riley acted the way she did—for example, why, when her parents feed her broccoli, she frowns, gags, and swats the vegetable away—is that there is a little person in her head prompting those reactions. In this case, Disgust finds broccoli disgusting.

If you’re like me, you might ask: if we looked into the head of these little people, would there be more, even smaller people inside their heads? Of course, the film never shows us what’s inside any of their heads. You might then wonder if the film has really provided any explanation of Riley’s actions, or if instead it has merely pushed the explanation of her behavior back a level. We can apply a point made by Skinner (1971: 19) and say the whole purpose of introducing the little people seems to be to help us understand why Riley acts how she does. But without providing an explanation of why the little people in Riley’s head act the way they do, we’ve failed to explain anything.

Over time, behaviorism itself was criticized for purporting to provide explanations without actually doing so, and there was a return to theories that unabashedly allowed for reference to mental states (Fodor 1981: 6). On such views, the way to avoid the homunculus fallacy is to slowly replace complex mental phenomena with combinations of simpler, more intelligible mental phenomena (Fodor 1968: 629). The hope is that eventually we arrive at an analysis constituted entirely by simple, elementary states (for example, thoughts and concepts) and the processes that combine them (for example, logical rules). We’ll call any collection of states and processes that enters into such an analysis a mental construct. Crucially, the explanation of how these states operate can be given without any reference to intelligent behavior.

And thus we return to the modern-day paradigm that explains behavior by reference to mental states. This paradigm has come to dominate theories of cognitive science and psychology, and is tacitly present in the theories of bias to follow. However, we should not forget the lessons of behaviorism. You should continue to ask yourself as we move through the theories: has this explanation rendered important parts of the psychological picture more understandable, or has it merely posited a convenient, but equally mysterious mental process? In other words, has it provided a genuine explanation or has it merely pushed the entire explanation back a level to equally intelligent homunculus-like states?

3 Empirical Data of Social Bias

At the onset of our investigation, we’re faced with several questions. What are the data surrounding social bias? In what ways do methods of testing for social bias differ from one another? What patterns emerge from these data?

3.1 Direct and Indirect Measures

Before the early 1970s, tests for social bias took a direct route: if a psychologist wanted to know if someone had a bias against a particular social group, she would ask her subjects directly. Such tests are called direct measures. Let’s focus on the case of racial attitudes in the United States. One of the earliest examples of a direct measure was a test created by Katz and Braly (1933) that asked 100 Princeton students to read through a list of 84 adjectives and write down those that they think best characterized a particular race or ethnicity. Characteristic of the time, the results indicated pervasive negative racial biases. The majority of participants in the study paired African Americans with traits like superstitious and lazy, while pairing Germans with traits like scientifically-minded and industrious.

Over time, the social landscape of the United States changed dramatically. The Civil Rights Movement of the 1950s and 60s strived to establish racial equality across the country, and ushered in a new public standard that discriminatory opinions about African Americans were socially unacceptable. During this time, direct measures began to show a decline in negative racial bias. However, although overt expressions of racist ideology were curbed, the pervasive and destructive effects of racism were still painfully evident. It seemed that people still harbored racist opinions, opinions that influenced their beliefs about and actions toward people of color; it’s just that either those individuals stopped wanting to admit those opinions to others or, more curiously, those opinions were not obvious even to them.

This prompted the emergence...

Table of contents

• Cover
• Endorsements
• Half Title
• Title Page
• Copyright Page
• Table of Contents
• List of illustrations
• List of contributors
• Acknowledgements
• Introducing Implicit Bias: Why This Book Matters
• 1. The Psychology of Bias: From Data to Theory
• 2. The Embodied Biased Mind
• 3. Skepticism About Bias
• 4. Bias and Knowledge: Two Metaphors
• 5. Bias and Perception
• 6. Epistemic Injustice and Implicit Bias
• 7. Stereotype Threat, Identity, and the Disruption of Habit
• 8. Moral Responsibility for Implicit Biases: Examining Our Options
• 9. Epistemic Responsibility and Implicit Bias
• 10. The Specter of Normative Conflict: Does Fairness Require Inaccuracy?
• 11. Explaining Injustice: Structural Analysis, Bias, and Individuals
• 12. Individual and Structural Interventions
• Glossary
• Index