Implicit bias is, according to the Stanford Encyclopedia of Philosophy, a phrase “referring to relatively unconscious and relatively automatic features of prejudiced judgment and social behavior.” Or in layman’s terms, implicit bias means that you could be prejudiced against certain groups of people without even being aware of it.
By testing automatic reactions and errors in grouping and classification, some cognitive psychologists have arrived at some tentative conclusions about the unconscious nature of prejudice. Research in this area dates back to the 1970s when researchers began to test subjects by using priming and automatic association and found that negative stereotypes were sometimes associated with racial groups.
The implicit association test, or IAT, is commonly used today and uses a computer program to test reaction times and association errors in word choice. For example, a subject may first be asked to use one button to select either “good” words (“happiness,” etc.) or ethnic groups (“white”) and another button to select either bad words or ethnic minorities. Then this technique is reversed. By measuring instant responses and error rates, researchers are able to determine whether an unconscious prejudice exists. For example, if a subject identifies the word “crime” with “black” when he should have selected “white,” this is taken as evidence of an unconscious bias.
Since 1998, Harvard researchers have been working on Project Implicit, which offers an online version of the IAT. The study has measured implicit bias in 17,000 unique subjects over the internet. Psychologists like Anthony Greenwald, Brian Nosek, and Mahzarin Banaji, leaders in the ongoing Harvard study, have brought their research to the wider public, both by expanding access to the IAT and by communicating their research in the media. Greenwald and M.R. Banaji published the book Blindspot: Hidden Biases of Good People about implicit bias in 2013.
In the last few years, the problem of implicit bias has become something of a political issue. Even Hillary Clinton, in the presidential debates against Donald Trump in 2016 answered a question on implicit bias, saying, “Implicit bias is a problem for everyone, not just police.”
Recently, institutions like police departments, universities, corporations, government offices, courts, and others, have adopted training programs to counter implicit bias. Implicit bias has the potential to significantly affect decision making in areas as diverse as hiring decisions, medical diagnoses, and voting, not to mention police-civilian interaction and court decisions—all areas where implicit bias has been shown to influence outcomes.
In light of recent unrest over the shooting of black citizens by white police officers, police departments have been under pressure to improve training programs to address race issues. Most recently the Northwest Indiana Times reported on an implicit bias training program adopted by the Griffith, Indiana police department.
In June of last year, the Department of Justice mandated implicit bias training for agents employed by Justice Department agencies like the FBI, the DEA, the ATF and the US Marshals Service.
Local courts, too, have taken implicit bias into consideration. Earlier this month, the Missouri Commission on Racial and Ethnic Fairness recommended implicit bias training as a way for state courts to hinder prejudice in the legal system.
Despite popular acceptance of the conclusions of the implicit association test, debate over the methodology employed continues within the academic cognitive psychology community. In the last several years, Phillip Tetlock, a noted cognitive psychologist and professor at UC Berkeley, has authored a number of papers in a back and forth with the psychologists behind the Harvard study, Anthony Greenwald and Mahzarin. Most of the criticism of the research focuses on methodology problems and possible misinterpretations of the results.
A particular point of controversy revolves around the issue of “test-retest reliability.” This refers to whether or not an individual subject’s results remain consistent through multiple iterations of the test. This is measured by a variable used by statisticians known as r. While an r of .8 is usually considered statistically significant, that same measurement for the IAT was less than .4.