Novelists often offer deep insights into the human psyche that take psychologists years to test. In his 1864 Notes from Underground, for example, Russian novelist Fyodor Dostoyevsky observed: “Every man has reminiscences which he would not tell to everyone, but only to his friends. He has other matters in his mind which he would not reveal even to his friends, but only to himself, and that in secret. But there are other things which a man is afraid to tell even to himself, and every decent man has a number of such things stored away in his mind.”

Intuitively, the observation rings true, but is it true experimentally? Twenty years ago social psychologists Anthony Greenwald, Mahzarin Banaji and Brian Nosek developed an instrument called the Implicit Association Test (IAT) that, they claimed, can read the innermost thoughts that you are afraid to tell even yourself. And those thoughts appear to be dark and prejudiced: we favor white over black, young over old, thin over fat, straight over gay, able over disabled, and more.

I took the test myself, as can you (Google “Project Implicit”). The race task first asks you to separate black and white faces into one of two categories: White people and Black people. Simple. Next you are asked to sort a list of words (joy, terrible, love, agony, peace, horrible, wonderful, nasty, and so on) into either Good or Bad buckets. Easy. Then the words and the black and white faces appear on the screen one at a time for you to sort into either Black people/Good or White people/Bad. The word “joy,” for example, would go into the first category, whereas a white face would go into the second category. This sorting becomes noticeably slower. Finally, you are tasked with sorting the words and faces into the categories White people/Good or Black people/Bad. Distressingly, I was much quicker to associate words like joy, love and pleasure with White people/Good than I was with Black people/Good.

The test's assessment of me was not heartening: “Your data suggest a strong automatic preference for White people over Black people. Your result is described as 'automatic preference for Black people over White people' if you were faster responding when Black people and Good are assigned to the same response key than when White people and Good were classified with the same key. Your score is described as an 'automatic preference for White people over Black people' if the opposite occurred.”

Does this mean I'm a closeted racist? And because most people, including African-Americans, score similarly to me on the IAT, does this mean we are all racists? The Project Implicit Web site suggests it does: “Implicit biases can predict behavior. If we want to treat people in a way that reflects our values, then it is critical to be mindful of hidden biases that may influence our actions.”

I'm skeptical. First, unconscious states of mind are notoriously difficult to discern and require subtle experimental protocols to elicit. Second, associations between words and categories may simply be measuring familiar cultural or linguistic affiliations—associating “blue” and “sky” faster than “blue” and “doughnuts” does not mean I unconsciously harbor a pastry prejudice. Third, negative words have more emotional salience than positive words, so the IAT may be tapping into the negativity bias instead of prejudice. Fourth, IAT researchers have been unable to produce any interventions that can reduce the alleged prejudicial associations. A preprint of a 2016 meta-analysis by psychologist Patrick Forscher and his colleagues, made available on the Open Science Framework, examined 426 studies of 72,063 subjects and “found little evidence that changes in implicit bias mediate changes in explicit bias or behavior.” Fifth, the IAT does not predict prejudicial behavior. A 2013 meta-analysis by psychologist Frederick Oswald and his associates in the Journal of Personality and Social Psychology concluded that “the IAT provides little insight into who will discriminate against whom.”

For centuries the arc of the moral universe has been bending toward justice as a result of changing people's explicit behaviors and beliefs, not on the basis of ferreting out implicit prejudicial witches through the spectral evidence of unconscious associations. Although bias and prejudice still exist, they are not remotely as bad as a mere half a century ago, much less half a millennium ago. We ought to acknowledge such progress and put our energies into figuring out what we have been doing right—and do more of it.