Summit 2022: A New Way to View Implicit Bias


It’s something you’ve probably experienced or witnessed in the workplace: the desire to hire talent from a more diverse candidate pool, or to ensure all employees are included in discussions. We know being aware of our biases isn’t enough, and that we need easy-to-recall strategies to mitigate them. But researchers have discovered something unexpected in our brains that might help explain why we still struggle to bring equity into the workplace – even when we want to.

To find out more about this groundbreaking work, we spoke with Dr. Will Kalkhoff who is the executive director of the Electrophysiological Neuroscience Laboratory at Kent State, and his co-director Dr. Josh Pollack, about the ways scientists first understood implicit bias, and what his lab’s recent research is showing about the contextual nature of expectations. Below are edited excerpts of our conversation.

What’s the current understanding of implicit bias?

To start, we need to understand that the implicit association test (IAT) was developed in the late 1990s and has a popular online quiz site – Project Implicit. It’s supposed to test for bias and prejudice. It’s caught on in different settings, and you can even take the quizzes yourself. So, everybody’s talking about implicit bias, and for good reason. We need to be aware of the fact that these biases exist, and we need to mitigate those biases in our workplaces.

Research over the years has revealed some problems with the IAT, though. Take the racial implicit association test, for example. You’ll see pictures of Black faces and pictures of white faces. Sometimes the faces are paired with the word “good” or “bad,” and sometimes they’re paired with random words like “flower.” The claim has been that people show bias by associating good words with their own race, but not with other races. Those tests show how you have a preference for this or that. But, at the end of the day, the test has been criticized because as it turns out, it doesn’t seem to have very good test-retest reliability (i.e., people’s scores on the test vary over time), and, more importantly, it doesn’t do a very good job of predicting behavior.

So we can test for bias, but these tests don’t help us understand what we might actually do in any given situation?

Right. To be more specific, let’s say I have a bias towards my own race. I could walk into a situation and expect the white people to speak the most and others to speak less. But that may be different if my manager is a woman, a person of color, or a woman of color. The status differential changes. Different status attributes combine in complex ways. This helps us understand why we shouldn’t expect bias to be temporally stable. Any one person’s results change over time. Different contexts matter, and people can show bias toward a racial group, gender, or sexual orientation in one setting, but not another. We need a theoretical framework and measurement approach that helps us make sense of these complexities and develop better interventions to mitigate bias.

What are you doing to better understand bias and how it plays out in real life?

We are finding it useful to reconceptualize bias in terms of “expectations,” or “expectation states,” more specifically. The sociologist Harold Garfinkel was one of the first to point out the fundamental importance of unconscious expectations in social life. Implicit bias shows up as a gut feeling rather than a conscious thought. Isn’t this fundamentally what we all mean when we talk about “bias”? Expectations are a powerful driving force of interaction, even though we aren’t usually consciously aware of them, and they change depending on the unique configuration of status-based differences among interactants in a social setting.

For instance, I might put it this way: in this setting, with these people, I expect that I should be treated this way. In this setting over here, with this different group of individuals, now my status has changed. I’m the lower-status member of the group, so now I need to listen more. Our expectations, our implicit biases, change.

Compare this to the original implicit bias research. If you think the IAT taps into something about individuals that is fairly stable, then our biases just wouldn’t be shifting all over the place when we examine behaviors in different contexts. We need to figure out a better way to test for implicit bias, and in a way that is sensitive to contextual changes. Again, we believe thinking about “implicit bias” in terms of expectation states and using expectation states theory to guide our research is a useful path to take

Let’s get to the science behind this. Specifically, how do you measure these biases, or “expectation states?”

We use electroencephalography, or EEG. We put caps on people’s heads with electrodes and they measure brain activity. We’re measuring expectations by using a method known as “event-related potential component analysis” This technique is used to look at the brain’s responses to different classes of stimuli.

We know from years of EEG research that when people encounter something that is unexpected, the brain represents and supports that intruding feeling of, “whoa, what’s this?” Well, when that happens, about 250 milliseconds after it happens, you see this event-related potential component — or this spike in brain activity — that is called the “feedback-related negativity,” or the FRN for short. So we see the FRN becomes more pronounced the more unexpected a stimulus or event is.

With our scenarios, we randomly assign people to different status positions and observe how their brains and actions change in response to different behaviors from other group members who have lower, higher, or equal status. What we’re finding is that we can measure expectations in different experiments and, in fact, looking at these status differentials and partner behaviors then predicts people’s own behavior in particular situations. In short, by rethinking implicit bias as expectation states, we are starting to see a clear link between bias and behavior, including discrimination-type behaviors.

What can leaders learn from this?

The good news is that we have many interventions that we can explore to disrupt this process of expectations. These can make groups more functional and effective.

One simple intervention, for instance, in a hiring context is simply to ask people who are hiring to justify their decisions. Just asking, “So, what are your reasons for hiring this person? Why do you think this person is actually going to be better than this person?” You’ve interrupted the process by which expectations translate unconsciously into behavior, likely in a way that the hiring manager just doesn’t realize. You help them confront their own bias in a manner that helps them take ownership of it and take corrective action.

Well-trained leaders who understand that these processes are in play can make sure members of lower-status groups are given time to contribute. Leaders can start a meeting by talking about the amazing things that some of these members have accomplished. So that immediately elevates their status in the eyes of the group. Or, as a leader, you can really bust up expectations by saying, “I want to hear from everybody, not just the managers. I want to hear from everybody in the room because this is a complicated problem. You’ve all got unique talents, abilities, backgrounds, and experiences. Chances are you know something we don’t, so don’t be shy.”

Even having a group conversation about how complex a task might be can be helpful. If somebody steps up and says, “Hey, I just want to remind everybody this is an incredibly complicated task. We know that people on our team have different skills and different abilities, and we need to capitalize on that.”

So what’s next in your research, and what can we expect to hear from your work soon?

As a scientist I can provide solid research that shows consistent links with bias, or these “expectation states,” and behavior in different contexts. We want to understand, theoretically and empirically, why our biases aren’t stable across time, and why they change as you move from this group to this group. Then, once we prove it, we can make better leadership and policy decisions that recognize the role of context as we provide training on implicit bias and expectations.

Author: Troy Hicks, Ph.D.

Want to know more?