When Yale University psychology professor Mahzarin Banaji posted a series of bias tests online, she simply wanted to diversify her participant pool beyond the confines of the Ivy League campus.
She never expected more than a million people to respond.
"I'm completely bowled over by the response," said Banaji, who created the project together with a colleague from the University of Washington in 1998. "It was an idea whose time had come and the technology was available."
The "implicit association tests" measure age, race and gender bias. Participants are surveyed about their explicit, or professed, attitudes toward different people before taking the five-minute test to uncover their hidden biases.
Although Banaji hoped to get a more diverse group of respondents by posting the test on the Web, an analysis of the questionnaires shows that most of the test-takers are young (under 50) and white.
But Banaji insists that the test is designed less as empirical research than as a way for individuals to discover their hidden prejudice.
"It's primarily an education site, not a research site," she said.
The tests use a Java applet that flashes faces and words on the monitor while respondents tap their keyboard to make associations between say, "young" and "nasty" or "old" and "pleasure." After completing a test, the respondents' answers are analyzed and compared with the other users' results.
A sample question is as follows:
Which statement best describes you? (Note: Understand Blacks and Whites as referring to the groups often identified as African vs. European in origin.)
- I strongly prefer Whites to Blacks.
- I moderately prefer Whites to Blacks.
- I like Whites and Blacks equally.
- I moderately prefer Blacks to Whites.
- I strongly prefer Blacks to Whites. The key to the test's accuracy is speed. Quick associations eliminate the ability to consciously manipulate the test. If respondents deliberate too long over answers, their responses are discounted.
Besides attracting heavy traffic from around the globe, the online test allows people to participate when and where they feel most comfortable.
"Psychological research often raises privacy questions," said Banaji, who's researched prejudice for 15 years. "We can ensure complete anonymity."
Normally, universities such as Yale pay students to take such tests; notices are plastered around campus and researchers hope that students answer questions with some degree of attentiveness before getting compensated. In contrast, Banaji's test is unpaid, attracting only people who are interested in the topic to begin with.
"These tests are as valid as any we might have in a laboratory," Banaji said.
So far the results reflect what we've already been told: Young people are biased against old, whites against blacks, and most people believe women are better adapted to liberal arts degrees than men.
But the site also warns people against jumping to conclusions based on their results. A female respondent who strongly associates science with men, for example, shouldn't rule out a degree in astrophysics.
The test is widely used. Many educators have incorporated it into their syllabi.
"It's one of the single most impactful interventions I do," said Princeton professor Susan Fiske, who uses the test in a psychology and racism course. "It makes students think harder about the material because there is a lot of automatic prejudice in the world."
The newly created Tolerance.org also offers the tests and had Banaji design three more to measure bias against skin tone, Asian-Americans and obesity.
"This fits into the whole concept of the site -- getting people to look into themselves," said director Jim Carrier. "The first step to overcoming the problem is recognizing bias in yourself."
Of course, there are those who disagree with their results and the Yale site offers the following: "If you are unprepared to encounter interpretations that you might find objectionable, please do not proceed further."
But that's another key to online testing: feedback. Users may suggest changes to the test or dispute the findings in a single mouse click.
