I was offered a test by several undergraduate psychology students on campus that left me puzzled. It was an implicit-association test (IAT) that attempts to identify implicit bias in the form of racism and prejudice. (Other IAT tests look at sex, age, religion, weight, etc. See here.)

The test measures accuracy and efficiency of response when associating negative and positive things while categorizing white faces and black faces. As a white person, I am said to have an implicit bias if I more efficiently associate white faces with good terms such as “joy,” “pleasure,” and “love,” and black faces with bad terms, like “horrible,” “evil,” “hurt.”

The students told me that most of us have an implicit racial bias, and that society is responsible for this widespread phenomena. (If that sounds very postmodern-y, it is.)

I pushed back a bit. It seems to me the test confuses unfamiliarity with certain ethnic groups and implicit prejudice or bias. In other words, their experiment only shows the accuracy and efficiency at which people are able to associate positive and negative attributes given their familiarity, not that the more accurate and efficient associations were caused by some conscious or unconscious prejudice or that less efficient associations were due to bias against people of a different race.

Consider a parallel test: suppose you know enough about auto-mechanics that you can change the oil on your car. Now suppose you’re asked, “What part of the car does the oil drain out of?” You wouldn’t hesitate to answer: the oil pan. But now imagine being asked, “What’s better for professional drag racing: a good supercharger or a good turbocharger?” To answer that question, you’d have to think a bit. Suppose you like reading racing magazines, and you’re able to recall an article about how powerful superchargers are good for drag racing. Right answer. But notice: responding to the second question took several milliseconds — maybe even a second or two — longer than the first. Why? Because it asked about concepts you were less familiar with. 

Does the result from this test show that you have an innate bias against drag racing? Of course not! The only valid conclusion seems to be that you were simply less efficient in recalling the less familiar question, and more readily able to recall the more familiar question.

And so it is with our ability to ponder questions about people of different ethnicities. Some ethnicities are simply less familiar. For instance, I didn’t grow up in a majority Chinese, Black, or Arab community; my parents and family and friends throughout my life have been majority white. None of this, however, means that I have an unhealthy unconscious prejudice or am in need of some kind of re-training or correction.

A simple way to determine a person’s views of other ethnic groups is to look at their behavior towards them — how they treat them and what they say about them when speaking honestly. Familiarity and more efficient reactions when associating objects more present in our memories does not demonstrate an implicit bias. More evidence is needed.

According to this video, my denial and/or rejection of the conclusion of my so-called implicit bias may be due to me being embarrassed by it, or even due to my own self-deception:

I say not so fast. I want to know at least two things. First, how they control for additional factors such as familiarity with concepts and life experiences, and second how their test is able to distinguish between the less familiar and genuine bias by response accuracy and efficiency alone.

Lastly, the IAT test fails basic diagnostics standards. Tests that attempt to produce a diagnosis must meet a criteria of reliability, repeatability and reproducibility, due to the many possibilities of error. However, taking the IAT test multiple times often produces different results. This means that any diagnosis of implicit racism has a huge margin of error, and such a test cannot provide consistent results.

So not only is the test invalid, it’s also unreliable.