What do unconscious bias tests really reveal about racism?

2020-08-26

YOU are biased. So am I. We all discriminate. It is both a source of concern and comfort that we don’t necessarily do so deliberately and that our prejudices aren’t always wilful.

If societies are to truly confront the pernicious effects of racism and prejudice, the importance of examining these biases and how they become etched into the brain is becoming increasingly clear. The death of George Floyd under the knee of a police officer in Minneapolis on 25 May shook the world to attention, but it was no isolated incident. Every day there are stories of people being treated with suspicion – or far worse – based on their skin colour while going about their daily lives.

This is in spite of the fact that, for the past 40 years, opinion polls show a steady decline in racist views in the USUK and other countries. That has led some researchers to suspect that, as explicit racism has been driven underground, unconscious bias is playing a critical role. This suspicion inspired the creation of the Implicit Association Test, a tool that aims to reveal unconscious biases with a few clicks of the mouse.

Unfortunately, the accuracy and reliability of this widely celebrated test isn’t what it once seemed. Pinning down the nature and extent of hidden bias is proving to be extraordinarily complicated. Eradicating it is far from straightforward, too – and it turns out that some efforts to do so may further entrench the very prejudices they are meant to uproot. But we are making progress, not least in understanding the processes in our brains that perpetuate bias – and what we can do to change them.

What exactly is unconscious or implicit bias? In psychological research, the label “implicit” refers to processes that aren’t direct, deliberate or intentional self-assessments. When we can’t retrieve a memory explicitly, we might still behave in a way that is shaped by our past experiences, for instance. The conscious mind governs deliberate actions, rational thoughts and active learning, while the unconscious carries on with processes that occur automatically or aren’t available to introspection. The unconscious is a busy place: the brain is capable of processing approximately 11 million bits of information every second, but our conscious mind can handle only 40 to 50 of them.

As all of this information comes in, our brains categorise it without our deliberate attention. When we process information on a more superficial level – when we are in a hurry, tired or distracted, for example – we are more likely to rely on existing templates. Occasionally, such cognitive shortcuts can be useful, such as when we need to decide something quickly. But they can also be problematic, especially if these shortcuts were formed based on mistakes, misinterpretations, stereotypes or other biased information. When we use them, we may then be relying on and reinforcing these very mistakes and biases. When that happens with people in positions of power and authority, it can have far-reaching consequences, from discriminatory hiring practices to /poorer healthcare treatment or prejudice in the legal system.

The idea that we could pin down and study implicit bias was first hinted at in 1995 when social psychologist Anthony Greenwald, then at Harvard University, and his colleagues invented the Implicit Association Test (IAT) to measure the strength of links between different concepts and words. For instance, participants would be shown black or white faces and asked to pair them with descriptors such as angry, clever, good and bad (see “How the bias test works“). This was adapted for the web in 1998 by Greenwald and fellow Harvard psychologist Mahzarin Banaji.

There have since been several adaptations of the test, measuring views on race, body type, gender and even names. The array of applications and easy online access have amplified the test’s appeal. It is hard to overstate just how influential it has been in both academic research and the public understanding of implicit bias. In his 2005 book Blink: The power of thinking without thinking, journalist Malcolm Gladwell summed up the prevailing view: “The IAT is more than just an abstract measure of attitudes. It’s a powerful predictor of how we act in certain kinds of spontaneous situations.”

Yet for all this, its results are inconsistent and hard to reproduce. Many studies have challenged the idea that the IAT reveals only unconscious processes. The reliability of results also appears to decline the more times you take it in a sitting.

What the IAT really measures is reaction time, based on the assumption that the speed with which we make associations reflects underlying mental processes. But everything from reflexes and physical ability to whether the user is distracted can influence this. Several studies have now shown that, for individuals, carrying an implicit attitude is only weakly linked to biased behaviour in the real world.

Part of the problem may be with how the test is used. Neuroscientist Calvin Lai at Washington University in St Louis, Missouri, studies implicit bias and is on the executive committee of Project Implicit, the non-profit research collaboration that studies implicit social cognition and examines the data gathered using the different versions of the IAT. He and others admit that the test is imperfect, but stress that it isn’t intended to be a one-off measure. “IAT results should be used as an educational experience for self-reflection but should not be treated as a tool for diagnosing one’s self or others,” he says. “A single administration of the race IAT tells you as much about your enduring racial attitudes as a single measure of your blood pressure tells you about your blood pressure over time: not very much.”

The nature of bias

But aggregated IAT results do tell us something about the nature of unconscious bias within societies. Information from Project Implicit reveals that, of the 630,000 people around the world who have taken a version of the IAT that examines associations between gender and science-related abilities, more than two-thirds correlate males more strongly with science roles and females more strongly with humanities, for instance. Test results from more than 1.8 million people in the US showed that in geographic areas where white residents show higher implicit race bias measured by a version of the IAT, there is also greater use of force by the police against black people.

Unfortunately, the IAT is still widely perceived as a diagnostic tool. Most anti-bias courses in the US and UK begin with the test, then give the results as a score that is seldom followed up by a deeper explanation. Occasionally, training programmes give examples illustrating the impact of unconscious bias and tips for how to reduce this influence (see “Ways to tackle your prejudice”).

Yet even with this kind of guidance, bias training is no magic wand that will cure individuals of their prejudices. It doesn’t seem to have a lasting impact on attitudes around diversity within corporations, for example. And while it appears to help reduce discriminatory behaviour by individuals for up to two weeks after attending, there is no evidence it leads to long-term change. Some kinds of training may even reinforce stereotypes, particularly if the participants are distracted or rushed.

That isn’t to say that we are without options. Advances in brain scanning techniques have helped reveal the neural underpinnings of our biases and in particular how prejudices about other groups of people activate brain areas associated with threat and fear (see “The roots of racism”). In an influential 2005 study, Mary Wheeler and Susan Fiske at Princeton University asked white volunteers who were in an MRI scanner to perform tasks while looking at black or white faces. They found that when the task involved thinking of the person whose face they saw as part of an out-group, rather than as an individual, the participants showed increased activity in the amygdala, the part of the brain that governs our threat response. Other brain scanning studies show greater activity in the amygdala when people view others from different ethnic backgrounds to their own.

Skin colour isn’t the only way our brains can automatically categorise people. Our response to different accents may be similar. In 2014Patricia Bestelmeyer at Bangor University in the UK and her colleagues found that when people heard accents similar to their own, there was increased activity in brain areas associated with positive emotional response; the opposite was true for different accents. “There is an increasing perception of the importance or relevance of those accents that are similar to ours,” she says.

Yet the imaging revolution in bias research has also demonstrated that our brains can change with experience and environmental influences. In 2013, Eva Telzer, then at the University of Illinois, and her colleagues conducted a study of 49 children and adolescents born in Asia, Europe and the US. They showed that the difference in amygdala activity in response to faces from different races wasn’t innate, but developed over a period of time.

This landmark study quashes any suggestion that we are somehow born prejudiced. What’s more, Telzer and her team found that study participants with a more diverse set of peers had less of a heightened threat response in the brain when shown faces from other racial groups. That suggests simply having more contact with people from different groups can reduce the importance of race in how we respond to people and that we can change our biases.

This wasn’t always a given. In social psychology there was a long-standing assumption that traces of past experiences linger on whether we want them to or not. But we now know that unconscious bias isn’t as stable as previously believed. Our biases are shaped by how we are brought up, what we see around us and the media we are exposed to. Knowing we can change their influence also means we can no longer shrug them off as beyond our control.

One day we may even have a tool that helps us to reliably measure them. “There is ongoing research to develop longer or more sophisticated versions of the IAT or other implicit measures that are reliable enough for diagnosis,” says Lai. Unfortunately, none are yet ready for public use.

We needn’t wait for new tools to assess the harms of bias, though. “Your best bet for understanding inequities in your organisation is collecting data about inequities within your organisation, not taking the IAT,” says Lai.

Even as efforts are under way to better measure the influence of unconscious bias, a growing number of researchers argue that we actually need to simplify this debate – to drive home that bias is bias, and whether it is unconscious or overt, whether individual prejudices shape social institutions or are shaped by them, they can cause irreparable damage. Unconscious bias is easier to ignore, but it cannot excuse discriminatory behaviour. It is important to remember that even if we cannot precisely measure our biases just yet, we can still overcome them.

newscientist.com, 26 August 2020
; https://www.newscientist.com/article/mg24732973-400-what-do-unconscious-bias-tests-really-reveal-about-racism/?utm_source=NSNEW&utm_campaign=adff1af429-NSNEW_270820&utm_medium=email&utm_term=0_e65fab71ff-adff1af429-378549783 ”>https://www.newscientist.com