Implicit Bias

It just never ceases to amaze me how you miss the point.The test doesn't work.Once you associate X with A on the LEFT or RIGHT sides, it is difficult to switch. Even if the software tries to train you.Have you noticed that all the stereotypical associations are first?
And that it also happens to be the favored one?

riVeRraT writes:

quote:

---

You see in my head, I don't associate either with or without weapons at all. If anything the whites I saw in the pictures looked like white supremacist anyway, and the blacks reminded me of people starving in Africa.

---

How did you manage to come to that conclusion considering that the only thing you were shown of the people was a cropped picture of the face revealing only the area from eyes to the top of the lips?Am I detecting another bias?

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

riVeRraT responds to me:

quote:

---

Once you associate X with A on the LEFT or RIGHT sides, it is difficult to switch.

---

But that's the point: It doesn't matter. The difficulty in switching will affect all sides equally. In researching the test, they worried about that and have tried making every single adjustment you could think of: Which axis to test first, which pair on the axes to connect to first, even down to switching left versus right. None of it makes any difference. It doesn't matter which way the tests are administered, which one comes first, which one gets paired with what first.

quote:

---

Have you noticed that all the stereotypical associations are first?

---

But they're not. For example, I just ran the gender/science one and the first pairing was female/science with male/arts. Look, I'm sorry you feel upset over being told something you didn't want to hear, but this is a time for you to make a decision: You can either yell and scream and rant and rave or you can decide to use the information to learn more about yourself and what you do. The test isn't telling you that you're a racist. It's telling you that you have a race bias. It's now up to you to determine what you're going to do with that knowledge.Isn't that what your holy book tells you to do? Stop worrying about the mote in your neighbor's eye when there is a great plank in your own? That all people are sinners and it should come as no great shock to find that all are wanting?

quote:

---

And that it also happens to be the favored one?

---

How can you tell what the "favored one" is unless you already have a bias as to what that "favored" connection is supposed to be?Again, I'm sorry you are upset that you found out something about yourself you don't particularly like. The question now is what are you going to do about it? Whine or engage in some self-examination?

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

The tests, while intriguingly crafted, cannot be said to produce truly accurate results of personal bias against groups of people. They are after all based on singular criteria and limited lists for association which may contain their own reasons for bias.Take for example the arab/muslim test, this involved a set of name associations. It took a group of muslim sounding names and placed them against a list of different ethnic sounding names. At best this test would suggest which kinds of ethnic names one would prefer, not people.Unfortunately that is really "at best". It gets much worse. The list of names of all the groups are not complete and so cannot be said to represent all or a great majority of any ethnic group's names. It could be that a person simply disliked (more) the selection of names given for any particular group, but a different list using different names might have produced something else.It is also strange to have arab v all other cultures. That might tend to diffuse reaction one might have had against one of the other cultures, or been boosted by positive reactions to a couple. Perhaps one became bored with the greater amount of arab names as compared to the diversity (rarity) of the rest.And of course familiarity with names may also play a part.This is one example, but it works for picture tests as well. As interesting as this idea is, it is a test which cannot significantly remove other factors which might come into play, heck even color blindness and other cognitive differences (like pattern perception) can play a part.I suppose if one gets a majorly biased against (or for) result, that might indicate something, but even then it would be hard to say for sure without looking at other factors associated with groups in real life.

---

holmes
"What you need is sustained outrage...there's far too much unthinking respect given to authority." (M.Ivins)

holmes writes:

quote:

---

Unfortunately that is really "at best". It gets much worse. The list of names of all the groups are not complete and so cannot be said to represent all or a great majority of any ethnic group's names. It could be that a person simply disliked (more) the selection of names given for any particular group, but a different list using different names might have produced something else.

---

This ignores well-established results regarding bias. As an example, interviews with HR managers has them directly state that they are actively recruiting black people but when they are then sent identical resumes except for the name, the white-sounding names receive 50% more callbacks than black-sounding names. And when the experience level described on the resume is varied, we find that it doesn't matter if the name is black-sounding: Highly qualified resumes from blacks get the same number of callbacks as average resumes from blacks.Plus, not all the tests are with regard to names. As the race/weapon test goes, you are given pictures of people.

quote:

---

And of course familiarity with names may also play a part.

---

Of course. But if you don't have an implicit bias against the names, you're less likely to have an implicit bias against the group who commonly use those names.

quote:

---

This is one example, but it works for picture tests as well.

---

So you just contradicted yourself.

quote:

---

As interesting as this idea is, it is a test which cannot significantly remove other factors which might come into play, heck even color blindness and other cognitive differences (like pattern perception) can play a part.

---

But that's precisely what the test works against. It's specifically designed to get around those aspects. Pattern perception would affect all sides equally.As I've noted, they've twisted the tests every which way they could and the results are always the same.

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

My response did not ignore these other bias studies. I am aware of them, and they are interesting as well but not directly relatable to this specific test.HR managers are selecting from resumes, not simply pushing buttons. While HR managers may very well have exhibited racial bias, based only on what they know from a name on a resume, that does not suggest you can spot racial bias based on how people associate names with words in a physical button pushing test. They are two different scenarios.I might add that one of the factors I raised regarding a potential factor in the association test, could be a factor in the employment studies. Although the realworld effect would be detrimental treatment of minorities, rather than bias against race it could very well be bias against names that are less familiar. Once again, I'd appreciate it if you read through my post before answering, or at the very least go back and erase points that are made moot by later elements within my post. This specific point of yours is addressed and you even reply to it later. I'm not sure which you are suggesting is the horse and the cart. I agree that if one is biased against such people one is likely to have a bias against common names from that group. But I don't agree with the opposite.The bias I was suggesting was an inability to connect words with unfamiliar names in the same was as with familiar names, which may be all this test measures. That is different than an active dislike of such names, or the people that might use them. I don't understand the contradiction. I said all the tests may actually involve confounding factors. I gave the example of a word test and then stated that the example works for pictures as well. By that I meant similar types of confounding factors. You will please relate how the test works against the specific factors I mentioned. As far as pattern recognition, perhaps you are thinking of pattern of questions or associations? I meant cognitive processing of visual images. How it would "affect all sides equally" I am not sure how that could be done or assured.As an example a study was done on facial recognition based on race. It was found that familiarity with a race allowed one to pick up on (identify) the features which most distinguished members of that race from each other. Such features are not the same for every race, and identifying the features was not racially inherent. Thus a black person could find blacks to "look alike" if they grew up around primarily white populations.Thus something which seems familiar, or involves patterns one can recognize, may be easier to process and so result in a different reaction, though have nothing to do with bias against a group.

---

holmes
"What you need is sustained outrage...there's far too much unthinking respect given to authority." (M.Ivins)

But that's just not the case. I found it confusing to associate whites with weapons, after I had spent the last 1.5 minutes trying to match up blacks with weapons.It was more game to me, than an actual psychoanalyses of myself.
When I was taking the test, I could care less about weapons and color, really. IF anything I associate weapons with whites more than blacks at the time of the test, because I was thinking of Nazi Germany when I looked at the white faces in the test.I am sure that the makers of the test did not account for that.So I stick to my original statement and say you can't make a test like that, and account for all the variables, when you just don't know them all. That is my problem with all of science. When people pretend to know all the variables.This of course does not make all of science wrong, but it doesn't make it right either.Sure everything in the physical world can be measured, unless our world stretches beyond what can be seen and measured. And if we know all the variables. But we just don't know.Still we can have fun trying. It's when something as silly as this test is even considered to be used in a court room, is when I have a problem with it.What's next? Tickle tests? No it's not rhain.If you stock something on a shelf for awhile, and then someone else tells you not to put it there anymore. You may in fact try to put it there the first few times, until you get used to the idea.Another case in point. January. It's a new year. You've been told this, yet you might still write the wrong date on a check.Sorry, this test taught me nothing about my race biased.I have never had a problem with any race, as long as I lived. We get along fine. Huh???
Can you please explain how I am pointing out a mote in someone’s eye?
I have done nothing of the sort.They made their own rules, so they can live by them (the scientists who made the test). They should examine why most of the tests put the stereo typical answers first. Argue the topic, not the person. How many times must you be reminded of this?*edit*
Since nothing in science is proven, I guess I am right.This message has been edited by riVeRraT, 02-12-2006 10:38 AM

holmes responds to me:

quote:

---

quote:

---

This ignores well-established results regarding bias.

---

​

My response did not ignore these other bias studies. I am aware of them, and they are interesting as well but not directly relatable to this specific test.

---

Except that's precisely what the test is for. It compares what people say to what they do. People claim not to have biases, but we find that the implicit bias test is a very strong predictor of what people do. People who took the implicit bias test with regard to race were then given the option to choose a white or a black partner with whom to work. Those who scored the strongest overwhelmingly chose the white person. Those who had low gay-bias scores were more likely to interact with a stranger who had a book with a title indicating a gay theme. Those who had a strong gay-bias score wouldn't make eye contact and would act coldly toward someone whom they were told was gay. In Germany, the test was given to people with regard to Turks and then they were asked to determine if a suspect in an ambiguous situation was guilty or innocent. Those with a strong bias against Turks were more likely to find the person guilty if presented as Turkish.Another example was a group who were told details about an assault on a woman where she had a concussion and required stitches. When told the perpetrator was "David Edmonds from Canada," the sentence given was much lighter than the one given when told the perpetrator was "Juan Luis Martinez from Mexico" among those who showed implicit bias against Hispanics.

quote:

---

I might add that one of the factors I raised regarding a potential factor in the association test, could be a factor in the employment studies. Although the realworld effect would be detrimental treatment of minorities, rather than bias against race it could very well be bias against names that are less familiar.

---

But it isn't the name. It's the association of the name to the race.

quote:

---

Once again, I'd appreciate it if you read through my post before answering

---

Once again, I would appreciate if you could keep a coherent train of thought. You made two arguments that contradicted each other, hoping we wouldn't notice.

quote:

---

or at the very least go back and erase points that are made moot by later elements within my post.

---

So you admit you contradicted yourself. Why should we pay attention to any of your arguments, then, if they naturally lead to a contradiction which you admit to?

quote:

---

I'm not sure which you are suggesting is the horse and the cart.

---

Neither.

quote:

---

I don't understand the contradiction.

---

You whined about the name. So when the name factor was removed and replaced by people with the same results, you're whining about the faces. The face test contradicts your claim about the name test.

quote:

---

I said all the tests may actually involve confounding factors.

---

But we removed the confounding factor and achieved the same result. Ergo, your claim of "confounding factors" is shown to be false. You don't get to have it both ways, claiming that the face confounds the name and the name confouds the face.

quote:

---

I gave the example of a word test and then stated that the example works for pictures as well. By that I meant similar types of confounding factors.

---

Incorrect. Instead, there are no confounding factors. By achieving the same results regardless of the icon used to signify race, we find that there is no confounding factor and the test is accurately measuring implicit bias with regard to race.This is part of the reason they kept mixing up the test with regard to its presentation. Did it matter if you tested one axis before the other? No. Did it matter if you paired the axes first one way and then the other or the other way around? No. Did it matter if you put one pole of the axis on the right as opposed to the left? No. No matter how you fiddle with the test, it returns the same results. If there were these "confounding factors" of yours, we would see changes in the result. Since we do not see any changes, we necessarily conclude that there are no "confounding factors."

quote:

---

You will please relate how the test works against the specific factors I mentioned.

---

I already did. Go back and read the thread.

quote:

---

I meant cognitive processing of visual images. How it would "affect all sides equally" I am not sure how that could be done or assured.

---

Because the stimuli is associated across the axis. It isn't as if blacks are represented by pictures while whites are represented by words. Instead, the axis is represented in precisely the same way for both poles. Therefore, any issues with regard to processing would affect all points along the axis equally.

quote:

---

As an example a study was done on facial recognition based on race. It was found that familiarity with a race allowed one to pick up on (identify) the features which most distinguished members of that race from each other. Such features are not the same for every race, and identifying the features was not racially inherent. Thus a black person could find blacks to "look alike" if they grew up around primarily white populations.

---

But we're not gauging the ability to distinguish two people of the same race from each other. We're gauging the ability to distinguish people of different races.You're complaining that the ability to distinguish between squares and rectangles is more difficult to someone who has only really been exposed to circles and ellipses is somehow affecting the ability to distinguish ellipsoids from quadrilaterals. The distinguishing characteristics are not the same.

quote:

---

Thus something which seems familiar, or involves patterns one can recognize, may be easier to process and so result in a different reaction, though have nothing to do with bias against a group.

---

(*blink!*)You did not just say that, did you? That's the entire point of the test! Familiar associations are easier to process and result in a different reaction which is the hallmark of bias.That's amazing, holmes. You really are a Republican, aren't you? What an amazing ability to claim that black is white, you have.

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

riVeRraT responds to me:

quote:

---

I found it confusing to associate whites with weapons, after I had spent the last 1.5 minutes trying to match up blacks with weapons.

---

No, you found it confusing to associate whites with weapons because you have an implicit bias against such an association. That's the point behind reversing everything during the test. All actions should then be distributed equally across everything. If you had no bias, all actions would be affected equally.

quote:

---

I am sure that the makers of the test did not account for that.

---

That's because it's irrelevant. You can try to justify yourself all you want, but the test is blowing past your conscious attempts and is looking at your automatic responses.

quote:

---

So I stick to my original statement and say you can't make a test like that, and account for all the variables, when you just don't know them all.

---

But that's why the test is designed as it is. You take it as quickly as you can so that you don't have to consider those other aspects.

quote:

---

It's when something as silly as this test is even considered to be used in a court room, is when I have a problem with it.

---

Who said it was being used in a courtroom? In fact, Banaji has stated that she would testify against the use of the test in such a manner.

quote:

---

quote:

---

It's telling you that you have a race bias.

---

​

No it's not rhain.

---

Why not? Because you say you don't have such a bias? Why should we believe you? People say things about themselves all the time that aren't true. That's the biggest problem about bias: Very often we don't even realize that we're doing it.

quote:

---

I have never had a problem with any race, as long as I lived. We get along fine.

---

It seems that you are interpreting the results to mean that you are some sort of raving bigot. Since it has been repeated over and over that this is an incorrect interpretation of the results, one wonders why you have latched onto it.I think it's safe to say that Jesse Jackson does not treat black people poorly compared to white people, yes? And yet, he admitted in a speech about 10 years ago about the pain he felt when he realized that he would "walk down the street and hear footsteps and start thinking about robbery. Then look around and see somebody white and feel relieved." He's black. He fights for equal rights for black people. And yet he, too, recognizes that he has an inherent bias against black people having grown up in a culture that is biased against them.While inherent bias is a very good indicator of behaviour, it is not a perfect indicator and people can consciously override their automatic response.

quote:

---

quote:

---

Isn't that what your holy book tells you to do? Stop worrying about the mote in your neighbor's eye when there is a great plank in your own?

---

​

Huh???
Can you please explain how I am pointing out a mote in someone’s eye?

---

You whine about how the test can't possibly be true without even pausing to consider it. You're so certain that this test has a mote that you have ignored your own plank.

quote:

---

They should examine why most of the tests put the stereo typical answers first.

---

But they don't. I've already shown that to you.

quote:

---

quote:

---

Again, I'm sorry you are upset that you found out something about yourself you don't particularly like.

---

​

Argue the topic, not the person. How many times must you be reminded of this?

---

I'll stop when you stop.You're the one insinuating political agenads on the part of the test designers. If you don't like having your personal psychology analyzed over the internet, perhaps you should stop doing so to others.Isn't that also in your holy book? Do unto others as you would have them do unto you? If you don't like it when it's done to you, what makes you think it's OK for you to do it to others?

quote:

---

Since nothing in science is proven, I guess I am right.

---

Prove it.

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

I've only done one and, it seems, have no particular preference for Meg Ryan or Julia Roberts. However, it may well be true statistically that one or the other can come first and statistically it washes out but what if one individual gets a particular order? I'm not clear how that is handled for that individual. I'll be interested to see what occurs in other tests. I do know I have some biases that I try to put aside. I'll see how well the test does at pulling them out.ABE an update
The next one I took was white/black people. It connected black to positive first. I came out with a slight (shit I wrote slight and went back to check) a MODERATE bias for white over black. Slight is what I want to think but ...This message has been edited by NosyNed, 02-13-2006 12:45 AM

NosyNed responds to me:

quote:

---

However, it may well be true statistically that one or the other can come first and statistically it washes out but what if one individual gets a particular order?

---

No. If you reverse the order and give the same test to the same person, the results wind up being the same. That's how we know that we're detecting actual bias and not a "confounding factor," as holmes likes to call it, or some sort of policital agenda, as riVeRraT is insinuating. If we test you and just you, moving all the variables around, the results wind up being the same.When Banaji received the test from her PhD supervisor, Greenfield, she thought the same thing: It must be an artifact of how the test presented the items. So she altered the test to reverse the order of the names, reverse the association order, even switching which one was on the left and which one was on the right, running it on the same people. The results didn't change.

---

Rrhain

---

Thank you for your submission to Science. Your paper was reviewed by a jury of seventh graders so that they could look for balance and to allow them to make up their own minds. We are sorry to say that they found your paper "bogus," specifically describing the section on the laboratory work "boring." We regret that we will be unable to publish your work at this time.

I just ran the Sexuality test, and the initial pairing was gay+bad and straight+good. I apparently scored a strong preference for straight. Which is unusual, because I'm gay. I took it again, and this time the initial pairing was gay+good and straight+bad (so it looks like they do change the order from test to test). This time I scored no automatic preference. Now, I guess I might have been 'subconsciously' trying to correct the bias the first test reported, but I tried to do it as fast as I could.

Your post contained personal insults. If you want to discuss this issue then keep it clean and about methodology as related to results. I feel my point has been made so I will not be answering any further posts which stray from direct, sober discussion. I stated in my post that those with high bias scores might very well have indications of a real world bias. That's one of the assumptions of the test which I have raised a question about. I will state again, the face test does not contradict my claim regarding the name test. I said in the same post that they both suffer from the same problem. That the test can measure something other than bias toward a race, do to confounding factors within the test itself due to its nature.I mentioned them both in the same post, so one could not have been removed and replaced. I also did not whine. Simple criticism of methodology and results is not "whining". These are both kinds of personal attacks I don't want to deal with in the future. ??? I never said this at all. I'm not even sure how that can be read into what I wrote. I did not criticize the method of mixing up the presentation. Indeed that looks like a good way to remove some possible confounding factors. Unfortunately mixing up presentation cannot remove the factors I discussed.If you have addressed the factors I discussed, then please repeat the explanation. I did not see them addressed before, or now. This does not affect familiarity or cognitive cues for distinguishing objects. I will add that imagery was not consistent, at least not for tests that I took. You had different backgrounds, color schemes, and expressions on faces. That's correct. What I was using that study to point to was how much familiarity plays in cognitive processing. Since cognitive processing is part of the issue at hand, lack of familiarity with a subject may produce OTHER, secondary issues which can confound the test. Thank you for underlining that my criticism was correct. Familiarity is largely what is being tested here, with an ASSUMPTION that testing lack of familiarity is equivalent to testing real world bias. I am challenging that very assumption.Yes those that are biased may lack familiarity, or those lacking familiarity may be prone to bias. But it is incorrect to claim that either are fundamentally related such that existence of one means the other must be there. There are other cognitive aspects related to familiarity that can confound such a connection being made by such a test.I also made specific criticisms regarding the lists. Obviously one may not be familiar with as many items on one list as another with no bias based reason. And in some tests there may also be an issue of diversity v rarity which can drive a different reaction. You have not responded to either of those points.It appears you feel this testing method is somehow beyond question regarding its limits. To me it is the same as any other type of psychological testing. It makes some assumptions and so creates limits by allowing for some measure of confounding factors not able to be tested.

---

holmes
"What you need is sustained outrage...there's far too much unthinking respect given to authority." (M.Ivins)

Oh yea?
I had a hard time telling the difference between weapons, and non-weapons. I guess that means everything is actually a weapon for me.Or maybe I view guns as not a bad thing, since I own a few.So maybe in truth, I was associating whites with stuff that I actually consider bad, and not blacks with guns.I wonder if they accounted for that? That's my whole argument, I do not think the test is good enough to do that. Especially if I can fake the test out. Don't take it wrong rrhain, I do not think the test is a complete failure. I just don't think it is a fool-proof method. The fact that it was even mentioned. I agree. That's like circle jerk.
I suppose he blames white people for that?
Or is it that more black people commit crimes in that time and place where it happened, and he knew that. I am sure he blames the whites ultimately for the crimes anyway.And he would probably be right. I have been close to many black people in my life, and I have seen first hand at what goes on. I have even been the victim of it, just because I associate with black people. Of course I considered it. I took it didn't I?
But the tests whole function is to point out a mote in your eye. That made me realize that the test has a plank.But it would seem, someone else mentioned that the order can be switched, and it is not a set order. That makes it better, but I am still not buying into it.The reason is, I just plain had trouble distinguishing pictures, let alone associations. The time spent to figure out what a picture was, interferes with my actual biases. You are relying on a few pictures, and words to spell out a whole way of thinking.It was like a game more than anything to me, not actual psycho analyses. I could care less what was in the pictures, my subconscious too. I am a gamer at heart, and that’s how I looked at it.Take for instance, a person who actually has no bias. No the whole test result relies solely on the persons ability to distinguish objects from one an other. The person may be able to recognize guns faster than non guns, and blacks faster than whites, so that person was able to associate those quicker. The test results would show that the person has bias towards blacks, when in fact he doesn't. Case in point. Have I insulted you by putting down the test, and it's designers?Do you keep science that close to your heart, and put that much faith in it?I see no need for you to defend the test or it's designers rrhain. You should be doing just the opposite and be keeping an open mind in the spirit of true scientific method, and logical thinking. But again you fall short of what you claim to be.I think your an intelligent person rrhain, and you have a purpose in life, I hope one day you find it. It can't be what is represented in this conversations we have, but that's just my opinion. Still I have enjoyed these dialogs, as frustrating as they may be. When did I say political agenda? I just think its a bad test, no agenda. They probably are genuine in their attempts.But they put the test out there, now it is subject to scrutiny. Isn't that the scientific method? Why do I have to explain that to you? Don't make me go back and quote you, unless you've had a change of heart, and a new definition of the word science.

Did you see a similar study in the same vein where they discovered that HR managers were half again as likely to respond to a white resume with a felony conviction - even for on-the-job theft - than a highly qualified black resume?Truly amazing.