Implicit Bias and You. And Me. And Everyone..

[The Grand Nudger]

Actually been looking into that.  Apparently, people don't get the same results when they take them repeatedly.  Unless their bias matrix changes between instances (in some cases just minutes), that would suggest that there's a problem.  OFC it shows associations, it's an association game with a time and pattern impulse UI...but what associations?

In any case, yeah....biases definitely exist.  It's difficult to see how we could avoid them.  Ambiguity demands interpretation, when for example, sally is having her children for dinner.  Does the manner in which one interprets this sentence expose an implicit bias?  

Is there a cognitive/linguistic bias involved in assuming, for example, that someone named sally, a mother, wouldn't eat her children? Or is the implicit bias more an issue of what this sentence could refer to in the first place? Is cannibalism the minority report that our subconscious ignores? Or is it, itself, associative? If I had said that jeffrey is having his friend for lunch...do we subconsciously omit (or perhaps, now, include) the dahmer?

[Aroura]

Actually been looking into that.  Apparently, people don't get the same results when they take them repeatedly.  Unless their bias matrix changes between instances (in some cases just minutes), that would suggest that there's a problem.  OFC it shows associations, it's an association game with a time and pattern impulse UI...but what associations?

In any case, yeah....biases definitely exist.  It's difficult to see how we could avoid them.  Ambiguity demands interpretation, when for example, sally is having her children for dinner.  Does the manner in which one interprets this sentence expose an implicit bias?  

Is there a cognitive/linguistic bias involved in assuming, for example, that someone named sally, a mother, wouldn't eat her children? Or is the implicit bias more an issue of what this sentence could refer to in the first place? Is cannibalism the minority report that our subconscious ignores? Or is it, itself, associative? If I had said that jeffrey is having his friend for lunch...do we subconsciously omit (or perhaps, now, include) the dahmer?

I completely agree, it's not a perfect test. Ideally, there would be someone qualified administering it and more controls and such. But for a test someone can take at home, it's decent. Better than most self administered online tests anyway, lol.

I think psychotic testing is always more variable as well, as it relies on a lot of self reporting. I guess that is one of the reasons some people consider it a soft science?

[The Grand Nudger]

I think it's far, far less than perfect, as the ui takes the familiar form of a prompting input game.  Some amount of the "bias" reported by this test cannot be anything other than hand eye coordination, being bad at games.

I'm not willing to conclude that someone has an implicit bias against, say, people of color, on the basis of whether or not they can speedrun super mario bros. Nor should anyone accept that I have no such bias....because I can. A -much- better form of this test has itself been the subject of criticism...and this one introduces such an obvious fudge factor that it seem pointless.

[Aroura]

Psychotic! lololol.
Thanks autocorrect!!

Psychiatric. I should say psychiatric.

[Aroura]

This topic wasn't about the test, though. Why focus on that and not the meat?
Quantifying the exact level of bias is difficult, of course, but that we can dismiss the whole shebang.

What do we DO about passing on these sorts of bias? I was reading 538 yesterday, and it pointed out that nationally, Trump is ahead 11 points with men, and Clinton is ahead 33 points with women. Are many women suffering a positive confirmation bias, or just overly negative about Trump? Are many men suffering a negative bias about Clinton, or a positive one about Trump? Some combination of both? It seems that if only men were allowed to vote, Trump would win by a landslide.

http://fivethirtyeight.com/features/elec...ald-trump/

This is a real world issue that deeply affects everything from whether your daughter will chose to become a scientist or a nail artist, to the outcome of the presidential election, to one of the root causes of police violence against blacks, etc.

[The Grand Nudger]

They've -used- versions of that test to produce the meat (for 20 years now, no less), surely the one affects the other?  The bias meat, as it were, contains what might be gently referred to as "non-meat filler". How much non meat filler can be allowed before it's no longer accurate to call it a steak, before it's meat slice?

You;ve simplified the equation above, and granted, we have to, sometimes, to parse things. Trump is ahead with what demographic of men? Clinton is ahead with what demographic of women? How many families, do you estimate, are split down the middle with daddy voting trump and mommy voting clinton? So, while the simplification seems to express some sort of gender based bias, it;s only because it tacitly assumes such a bias. Which is confirmation bias.

It is a real world issue, it does deeply affect us. All the more reason to eliminate one's proficiency at video games from the bias conclusion. For my part, as regards Trump..if there's something about him that makes him more desirable to men, specifically, I'm missing it.

[Aroura]

The test lets each user set their own baseline. If you are slow and make a lot of errors during the first 4 parts of the test, then if you are slow and make a lot of errors during the 2 that determine the bias level, it takes that into account. And visa versa if you are fast and make few or no errors. I honestly don't see your "good at Mario brothers" point, as if you are good or bad at the game is accounted for in the first 4 stages of the test. Can you link me something that suggests such testing, particularly the formal versions conducted in a lab, are as utterly useless as you seem to think?

Also, did you read the article fromarketing 538 I linked? It answers a lot of the questions you asked.

[The Grand Nudger]

You could just check the wiki page for the test type to figure out all of the criticisms such a test, even in a better form, is subject to.  The point about mario brothers is that an input reversal and prompting is one of the ways that video game designers have made their games "difficult".  Getting you to press the "wrong" button at a familiar prompt (say you meant to hit I because you were thinking bad but because e was the previous "bad" button you hit that instead).  Some amount of my getting answers "right", particularly after reversals, comes down to how familiar I am with that mechanic, which is present both in the test and in just about -any- video game. Some amount of my getting answers "wrong" is motor control. It's present even in the baseline setting. It;s not merely accounted -for-, it's the very accounting of purported bias.

Trouble is, that these are the purported associative metrics for determining biases...and that;s just another wrinkle on top of the basic form of the test, applied in a different form or setting. May as well test my biases by asking me a series of questions related to beyonce while playing whack-a-mole - alternating between my left and right hands. While you may, with a deep enough data set, be able to tease out correlations (lies, damned lies, and statistics)...it just might be that I'm good at whack-mole, and that covers for my lack of truly felt/known beyonce knowledge.

Do you think that a fatty hater, to the very core, couldn't game that test? Is it inconceivable that his performance would indicate little to no implicit bias? If it is possible, -how-...how could he do that, how could it be done? Perhaps by realizing those instances in which the prompts put a "bad" word up before a "fat" sil, or a good word before a skinny sil, and every permutation between, and minding his i's, e's, and reversals. The test positively depends upon a person not realizing that's happening...so is it really a test of bias, or something else? if it is a test of bias, however imperfect, what measure of something else is included in the measure of bias. For example, could the difference between "slight to no bias" and "heavy bias" be accounted for by that "something else" even in an instance in which no conscious attempt to game the system is occurring? Important questions.

[GrandizerII]

The IAT (Implicit Association Test) is not without its flaws, obviously. But it's shown good reliability and validity in some cases at least, although it seems to suffer inadequate test-retest reliability at times (depending on the variables assessed, I guess).

Anyhow, there will always be some nuisance variables that need to be controlled for but are hard to disentangle, but it does seem like a lot of these variables have been taken care of, like the pairing order effect by randomising it with each participant.

[henryp]

The other weird considerations.

I consider things "black and white" and "good in bad" in that order. So if I have to think of them out of order, my brain gets confused.

On top of an actual association of the color black with bad, and white with good. I wonder how it'd play out with african american, or some other terminology that doesn't already have it's own bias.