Comment by TheGRS | Hacker News Reader

Comment by TheGRS | Hacker News Reader

Our company recently did a trial of looking for unconscious bias when we are making decisions, especially in terms of hiring practices. We all took a test to see if we had any unconscious bias toward gender and then discussed our results as a group. What really stuck with me is not that we have these sort of tendencies, but how shocked most people were at their own results. Most men in the room trended toward a bias toward men and women toward women. That shouldn't be that surprising really, but many were very disappointed in their results.

I've been thinking it would be good to do another round of that with the race test, especially considering how predominately white our workplace is. I don't believe everyone is actively trying to be racist or sexist (though some certainly are), but we might not realize that we're doing that on an unconscious level.

The tests are available here if you'd like to give it a try yourself, doesn't take very long. https://implicit.harvard.edu/implicit/selectatest.html

Implicit association tests have little predictive validity--that is, if we know that someone has more "unconscious bias" in an IAT, we should expect zero difference in their behavior.

See:

http://www.law.virginia.edu/pdf/faculty/reassessingpredictiv...

https://www.researchgate.net/publication/239732934_Predictin...

http://onlinelibrary.wiley.com/doi/10.1111/sjop.12288/abstra...

Great studies. I always felt something was off about the IAT and couldn't tell exactly why. Maybe it primes us for racism/sexism/whatever it's testing?

I think many people posess a self-image that race is not a salient characteristic in their thoughts and they describe this as being "not racist". There is a premise in discussions about race that being racist (judging someone because of their race) and acting racist (treating someone differently based on their race) are two different things. We act not-racist to prove that we are not-racist in our brains. The accusation of racism can be rebutted by the argument that you can't read my mind.

The implicit bias test shows that either this brain-focused definition of racism is wrong, or we need to accept that we are more racist that we would like to admit.

Inside the German version there is a test for biases against Western or Eastern Germany. I knew that I had a bias, but I was surprised that I found it that difficult when moving a bunch of Eastern German cities to one group to associate negative things to the group with the Western German cities.

The test might not be totally fair, though. Is it actually a bias if it is based on facts? Soviet architecture, higher unemployment, high rate of extremism, lower buying power, less people, people moving to the West, the list goes on and on. I think stereotypes and preconceived notions are bad most of the time, but I see no reason to get rid of the ones that are actually true for political correctness.

What is the german version of nextdoor? I've never heard of such a thing.

I think he's referring to the german version of the implicit association test

the test is fair if the goal of reducing racial bias is not to reduce crime but to achieve a more integrated society. in the long term, if such a policy contributes to a more integrated society, then more crime in the short term is deemed an unfortunate but necessary sacrifice. this is, in essence, the implication of such social pressure.

The test is totally fair in a sense that it does detect bias. What you're actually saying is that the bias itself is fair.

Now, normally, it would depend on whether the facts that you're citing are true and relevant or not. I don't know enough to judge that.

But secondly, even if the group bias is fair, it's still problematic when you apply it to a specific person. They do not necessarily have to be representative of their group average, and you owe at least giving them a shot to demonstrate that.

> The test is totally fair in a sense that it does detect bias. What you're actually saying is that the bias itself is fair.

Sure, phrased that a bit wrong.

> But secondly, even if the group bias is fair, it's still problematic when you apply it to a specific person.

The test was just about associating cities with words, I do not think you can say anything about persons or specific persons alone from that information.

That is also true. But tests like these do make a point that many of the things that happen in our mind are not the things that we are aware of. I mean, if those people who passed the test were asked in advance whether they'd be grouping based on any particular trait, or even specifically asked about the trait they grouped by, I doubt that many of them would believe that it would work out the way it did. It's not a far stretch to assume that snap judgments of other people in course of day-to-day life can be similarly affected, and at least self-check against the possibility.

Implicit assertions test is often used to "prove" biases, but honestly I have never seen a large (or any) number of publications that proves it works.

I like tests like these that can potentially point to internalized prejudices and biases. Knowing one's biases can cause pain and disappointment if we see that what we think of ourselves, and what we truly are, are two different things.

However, having taken one test (just now), I was left with the feeling that these tests may probably not be very accurate for everyone. When I tried it, I did it more from a memory point of view to associate the word and category. I didn't even think about the sequences and any implicit connections. Of course, my result came as "little or no bias", which I do agree with because I have thought about these for a long time and have been contemplating biases (and biased thoughts/decisions) that were not clearly obvious in the past. What I'm not sure of is the applicability of this methodology for all kinds of people.

Most people are carrying racist/sexist biases. Very few are aware of it. It's really challenging to address.

Reply on news.ycombinator.com