Neutrality


Harvard IAT – How to Beat It and Why Bother – Part I

 

Bias

The field/profession of “Diversity” has been taking a beating for a while now.

Despite its continued and increasing presence in the business community, the pecuniary value and benefit of Diversity interventions remain in serious question. Not that it wasn’t a fine idea a half century ago as the vague concepts of “valuing differences” began to take root in the 1960s. But the oft maligned concept has remained poorly defined, and therefore poorly measured, if measured at all. Once quotas were yanked from the Compliance-based model of Diversity, it was as if the limb had been sawn off after the Diversity folks had climbed out on it. Compliance with equal opportunity objectives still stand in law, despite the nearly complete breakdown of post-Bakke progress in equal employment opportunity by about the year 1980. Since then, the Diversity field slid even deeper into definitional chaos and programmatic anarchy.

Diversity objectives, on the other hand, have found their way into neither law nor regulation, remaining largely social initiatives with dubious business case linkages. With the onset of “Diversity Fatigue” and a general sense that it has failed to either make a solid business case or achieve anything approaching equality or equity in the workforce after half century, without professional standards, without uniform definition and metrics, there is a line is forming to replace “Diversity” with any number of new concepts and ideas.

Enter neuropsychology, implicit association and “Unconscious Bias.”

In 1998, social psychologist Anthony Greenwald introduced the concept of “implicit association.” Greenwald & Banaji, co-authors of Blindspot: Hidden Biases of Good People (Delacorte Press, 2013) provided a linear history of the evolution of the terminologies, from “unconscious mental function” to “unconscious cognition” (1990s) to “implicit cognition” to “implicit bias” to “implicit association” to “hidden bias” and more. The essence is this: Memories that are not accessible to awareness can influence our actions, and therefore, related memory-based associations can also influence our attitudes and behavior. In short, what we don’t know we are thinking can affect our attitudes and the choices we make.

Today, the widely known Harvard Implicit Association Test (IAT) measures the strength of an array of individual associations related to a variety of human “differences.” At one time attributed purely to conscious prejudice, societal failings are increasingly being attributed to “unconscious bias.” These hidden preferences, lurking in the psychological recesses of average citizens, are now hypothesized to be a major culprit in lack of progress in the field of Diversity. Now, through the miracles of modern technology, people all over the world are getting in contact with the unconscious realm.

In the book, Dr. Banaji gave away the tricks of how the IAT actually works to detect bias. I was able to work it through, not because I read Banaji, but because I have some programming training, as well as survey research. The input had to convert to the outputs related to the IAT survey. Greenwald & Banaji understood the physiological responses to bias from a clinical standpoint, while I just backed into it by understanding that a computer can only measure Input-Output (I/O) as opposed to reading your mind to divine implicit association. It wasn’t hard to deduce what the I/O measurements had to be, and thereby the psychology as it relates to survey technique and I/O limitations.

Computers are stupid. They don’t know what you’re thinking or how you feel. Computers must be programmed to recognize inputs, and they can calculate measures only with programming. So, what inputs could a computer program collect, track and send to Harvard Project Implicit?

1. Which key(s) you hit.

2. The time between each keystroke.

3. The time between the first and last keystrokes.

4. The total number of keystrokes.

5. How many associations you make.

6. Which key you hit in association with each picture.

7. The amount of time you took to make each individual association.

8. The total amount of time you took to make all the associations, and,

9. Which associations you were not able to make at all, or wild keystrokes, which is going to be rare, because you only use one of two keys for all associations.

There may be other things, but you get the idea. After that, it’s just interpretive statistics about your keyboard performance that get factored into your “preference” score by the program.

So, how do they program the test so that they can infer preference? Here’s Dr. Banaji’s huge in-your-face two-by-four to the forehead clue on how the IAT works…and I use italics to zero in on it.

Dr. Banaji says: “So when I took the test … it was stunning for me to discover that my hands were literally frozen when I had to associate black with good. It’s like I couldn’t find the key on the keyboard, and doing the other version, the white-good, black-bad version was trivial…it sunk in that this test was telling me something so important that it would require a re-evaluation of my mind, not of the test.”

Music isn’t only the notes that are played, as the great French composer Claude Debussy observed, “Music is the space between the notes.” If the IAT cannot measure anything more than keystrokes and the space between them, then that’s the only thing the psychologists will have to work with to interpret the meaning of your performance. If they know that your negative implicit cognition inhibits speed/motion, and positive implicit cognition facilitates speed/motion, then your negative cognition will produce slower or wilder keystrokes, while your positives will facilitate more accurate and speedier keystrokes. They are reading your performance, not your mind.

From there, it’s just a matter of measuring two or three things, 1) the actual choices made, 2) the statistical variances in time between choices, and/or 3) variances from average/mean (norm) choices of yours and others. (That’s an oversimplification of the algorithm, but you get the picture.) The bottom line is that your own keystrokes and speed variances betray your bias. This is not kept a secret at all. IAT FAQ #11 openly reveals the need for speed.

Q 11. What does it mean that my IAT score is labeled ‘slight’, ‘moderate’, or ‘strong’?

Answer: Assume that you respond faster when flower pictures and pleasant words are paired on a single key than when insect pictures and pleasant words are paired on a single key. Your score would be described as showing automatic preference for flowers. (In general, a result shows an association between concepts that, when paired, get fast responding.) The labels ‘slight,’ ‘moderate,’ and ‘strong’ refer to the strength of the association (i.e. how strongly you associate flower pictures with pleasant words). No matter which IAT you took, if a speed difference between different pairings was so great as to be obvious to you, it would likely be labeled a ‘strong’ effect. The ‘moderate’ label also indicates a difference large enough so that you would probably notice it. A ‘slight’ effect is one that is noticeable in statistical analysis, but you may not have been aware of it.”

The FAQ answer is somewhat oversimplified. IAT describes the strength of preference for (A) or preference for (B). The degree of preference is arrayed from mild to strong — though it’s still implicitly stated as a preference for either (A) or (B). But in fact, there’s another way of portraying it, and it’s the one the IAT folks must actually use. There are three possible primary scores, but not those described in that “slight, moderate or strong” range reported in the IAT FAQ. The real scores are at polar ends of a spectrum of scores — and in the center. Let’s use IAT race preference scores as the example first, comparing the response structure of a black to white race bias. The actual range of diagnosis is:

(A) Preference for White, “slight, moderate or strong” OR
(B) Preference for Black, “slight, moderate or strong” OR
(C) No Preference for either, neither preferred over the other.

But…here’s how it could — and probably should — be remapped.

 

Competing BIAS (A) . . . . . . . . . . . . . . . . . . . . . Competing BIAS (B)
Preference for White ====> NEUTRAL <==== Preference for Black
Strong ======> . . . . . . . . No Preference . . . . . . . . .<======Strong
10 –9 –8 –7– 6 –5 –4 –3 –2 –1 — 0  — 1 –2 –3 –4 –5 –6 –7 –8 –9 –10

The score (C), “No Preference,” is the one not discussed much in the FAQ. Instead, I prefer to call that one “Neutral.”

The “No Preference” is actually a “culture neutral” state, while the left and right tails are the strongest polar cultural preferences, or biases. What IAT and other similar tests are actually measuring is the statistical distance (variance or deviation) from a mean state of a non-preferential neutrality.

It is useful to understand the objective of placing “Neutral” at the center of the scale. The objective is not to exchange one bias for the other, but to cancel out a bias or preference. Most would be horrified, for example, if testimonials for the racial bias tests were published reporting, “Before I took the Hidden Bias Test, I was clearly favoring Whites in my day to day work. Thanks to the test and the new bias therapy, I now favor Blacks, instead.” We can thereby easily understand that the objective of “neutrality” would be a desirable outcome.

In this example, neutrality can be defined as a state of balance at which a bias will not impel the subject to act in a manner that favors one person over the other based on the preference.

In other words, these folks are actually in the business of measuring how far from neutrality you are along specific lines of a given bias. They don’t say that, either because they tend not to look at it that way, or because it isn’t a score that large enough numbers of the respondents of any race achieve – which is troubling, considering that the idea of Managing Diversity has been around for a half-century already.

Consider IAT FAQ #9, containing the only reference to “No Preference.”

IAT FAQ #9. Do Black participants show a preference for Black over White on the race attitude IAT?

Answer: …Although some Black participants show liking for White over Black, others show no preference, and yet others show a preference for Black over White….”

The “No Preference” is a score some have apparently achieved, based on FAQ #9, and so can you, if you want to. Now that you know the rough scoring mechanism, and if you want to improve your bias score (meaning, get closer to a culturally neutral state in your tested associations), then just as you might have guessed, the path is in managing both your thinking and your keyboarding.

Trying several things might work for you, as they did for me:

  1. Do a minute or two of diaphragmatic breathing before starting the test. Relax.
  2. Remember: Unconscious Bias is NOT the same thing as prejudice. (See IAT FAQ #16)
  3. Before starting the race bias test, prepare your mind by repeating, “I’m Black and I’m Proud!” especially if you’re white.
  4. Choose an initial response speed that will give your brain just enough time to get the right signal to your key finger as often as possible.
  5. Keep an even rhythm for your keystrokes. Use a metronome, or just a favorite from iTunes or YouTube. The song that worked best for me is Michael Jackson’s “Billy Jean.”
  6. Practice using the left & right cursor keys and the space bar along with the song, or any other two equidistant key combinations with the space bar.
  7. Take the IAT as slowly as you need to. The program may invalidate very slow responses. But it’s still free. Speed it up a bit with each iteration.
  8. Remember: the computer won’t “out” you to anyone. It’s just a computer. Relax and have fun.
  9. Don’t worry, you won’t break the computer or damage the test on the other end at Harvard.

The likelihood is that within three or four encores of the Harvard IAT, you and Michael Jackson can get your score consistently to “No Preference.” Keep taking it every few days until you routinely get to a CultureNeutral score. Go for it. It’s free!

Send a note on how you did to nuclusiv@gmail.com – we’d love to know.

But now that you know how to beat the test, the corollary question is: Why would anyone want to go to all that effort to defeat a test? Is it cheating, “gimmicking” the test, winning? Who would want to be “neutral?” We’ll talk about that and more in Part II of Harvard IAT – How to Beat It and Why Bother.

TO SHARE THIS ARTICLE TWEET: http://www.UnconsciousAmerica.com

________________________________

Copyright © 2014 Robert D. Jones – All Rights Reserved

To learn more about “neutrality” and the CultureNeutral® Framework, visit us at http://www.InclusiveWorks.com

 

Reference Links to additional reading:

1. http://www.npr.org/templates/story/story.php?storyId=122327104

2. http://blogs.hbr.org/2013/06/is-your-firm-really-an-equal-o

3. https://implicit.harvard.edu/implicit/

4. http://tinyurl.com/modernprejudice

5. https://en.wikipedia.org/wiki/Implicit-association_test

6. http://tinyurl.com/Banaji-HBRArticle

7. http://personallyinspired.wordpress.com/2006/07/20/5/

8. https://implicit.harvard.edu/implicit/demo/background/faqs.html

 

Related Posts

No Comments Yet.

Leave a Reply

Your email address will not be published. Required fields are marked *