Implicit association tests show that we all have forms of unconscious prejudice on gender and race. But does that make us sexists and racists? Initial studies are finally yielding concrete results.
Philadelphia, April 2018. Two people are waiting for an acquaintance in Starbucks. The manager calls the police, justifying his action with the fact that consumption is mandatory in his establishment. A viral video showed police officers handcuffing innocent people – African Americans. It led to public protests, and eventually to an apology from the managing director. He also announced that the company’s 8,000 employees would undergo training to combat implicit racial prejudice.
A large volume of scientific literature in social and cognitive psychology attests to the existence of implicit biases, namely automatic associations made by our brain. Tests can measure them, the most well-known being the implicit association test developed in 1998 by the American psychologists Mahzarin Banaji and Anthony Greenwald. It measures in milliseconds the reaction time of a person to make certain associations. A majority of individuals thus take longer to associate “good” with “black” than with “white”, or “science” with “woman” than with “man”. This is claimed to prove the existence of prejudices, conscious or otherwise. The test is used in many fields and has gained popularity in studies on racial, gender and age discrimination.
Over the last two decades, this test has become established as a relevant scientific tool, despite some criticism. Numerous studies have documented the existence of various biases. But most often, they are conducted in laboratories and in simulations, leaving open an essential question: Do implicit biases really have consequences on behaviour in day-to-day life?
Not enough on their own
Two recent studies have shown interesting correlations between implicit biases and actual behaviour. The first was published in the summer of 2019 and described around 40 experts responsible for evaluating the applications for research management positions at the Centre national de la recherche scientifique (CNRS) in France. Its analysis looked at whether the evaluation process discriminated against female candidates. “Our approach is unprecedented”, explains Isabelle Régner, a professor at the Laboratory of Cognitive Psychology at the University of Aix-Marseille. “It’s studied the activities of real evaluation committees in all scientific disciplines, from maths to sociology”.
The research team included Pascal Huguet, the research director of the CNRS. It subjected the members of the evaluation committees to an implicit associations test, revealing that more than 70 percent of them have a bias associating science with men, in other words the same average as among the general population. The subjects also completed a questionnaire asking them whether they felt that women are less likely to be appointed to management positions because of intrinsic factors (e.g., their skills or motivation) or because of extrinsic barriers (e.g., hiring biases, barriers to advancement, etc.). This is an explicit bias test.
Implicit biases alone do not explain discriminationIsabelle Régner
The study was conducted in two phases. In the first year, the evaluation committees were informed by senior management of the research being conducted. The decisions made by their members were somewhat parity-based, showing no relation to their implicit or explicit biases. The study continued into a second year, which was not known to the committee members, and revealed that those showing both implicit and explicit biases appointed fewer women, unlike those showing implicit biases only. “Implicit biases alone do not explain discrimination”, says the author of the study. It is only in combination with explicit biases that they influenced committees’ decisions. “The study also shows the effect of knowing that one is being observed: these differences were not present in the first year, when the evaluation committees were aware of participating in a study on gender discrimination.
The indirect effect of prejudice
In 2017, a study conducted by France’s INSEAD revealed the subtle effects of implicit racial biases against employees shown by supermarket chain managers. It analysed the situation of cashiers on temporary six-month contracts (the sample consisted only of women). Every day, the supermarket’s human resources system assigned employees at random to a manager. An implicit association test meanwhile revealed biases in 80 percent of the managers.
The study explored the impact of the origin of employees, which was determined on the basis of their first names. The results showed that people from minorities (either of North African or sub-Saharan origin) showed higher rates of absence and error when working for a biased manager. “What is interesting is the cause of this drop in performance”, says the study’s leader Dylan Glover. “During our survey, no employees mentioned inappropriate or overtly racist behaviour on the part of biased managers. On the contrary, they noted a low level of interaction, and few requests from them for thankless tasks, such as cleaning”. The scientists concluded that biased managers simply interacted less with minority employees and that this resulted in a decrease in their performance. “Our study indicates that the effect of implicit biases on behaviour is not necessarily the one you would expect”, says Glover. “In our case, we find ourselves more in a self-fulfilling prophecy: managers are biased; this affects the productivity of minority employees, which in turn confirms managers’ prejudices”.
So should supermarkets set up training courses for managers to work on their unconscious prejudices, for example by presenting images that contradict gender or racist stereotypes? Glover thinks not, saying that further research would be needed to demonstrate the usefulness of such measures. And the study on the CNRS evaluation committees concludes that any training should focus more on the knowledge of the existence of implicit biases and on the functioning of discriminatory behaviour. It should also include awareness of external factors that disadvantage women. “Our research indicates that these kinds of measures could be effective”, says Régner. “But the subject still needs to be explored with further studies before systematic policies can be developed. This is what we are currently doing”.
A domain not yet ready for application
Both studies indicate a link between implicit bias and direct or indirect discriminatory effects, but they fail to determine that specific training on implicit biases can impact discriminatory behaviour. According to a meta-analysis of 492 studies on modifying implicit biases, the literature currently shows a mixed picture. “The scientific literature does not allow us to conclude that training courses aimed at changing implicit biases are either effective or ineffective”, explains the author Patrick Forscher, a postdoctoral fellow at the University of Grenoble Alpes. “Nor does it demonstrate that any successful changes actually have an effect on discriminatory behaviour. On the basis of current knowledge, we just don’t know anything about it”.
I consider this field of research as not yet ready for any practical applicationPatrick Forscher
Forscher adds that he was struck by one thing: The vast majority of the research analysed was based solely on laboratory tests, using experiments that generally did not last longer than 5-10 minutes. No long-term changes in implicit bias have been measured yet. “I consider this field of research as not yet ready for any practical application”, he says. Food for thought for the many companies and institutions that are paying out for employee training on implicit biases.