Register | Sign In


Understanding through Discussion


EvC Forum active members: 65 (9162 total)
4 online now:
Newest Member: popoi
Post Volume: Total: 915,817 Year: 3,074/9,624 Month: 919/1,588 Week: 102/223 Day: 0/13 Hour: 0/0


Thread  Details

Email This Thread
Newer Topic | Older Topic
  
Author Topic:   Implicit Bias
mike the wiz
Member
Posts: 4752
From: u.k
Joined: 05-24-2003


Message 46 of 52 (286412)
02-14-2006 10:14 AM
Reply to: Message 36 by macaroniandcheese
02-13-2006 1:14 PM


Re: Bren's new rules of chess
My main point is that there is the passive conformist who cannot be the instigator, unless moved to, and then there is the person of self-motivated bias. The latter will be the instigator in any move, ten times out of ten. For he is the leader, and will be out of two people. The more he's self-biased, the higher he goes in the group. The other sheep will remain mere deatheaters.
If you think it's right to protect hate-speech then that's your opinion. I don't do politics so I'll leave that to you. My only interest is seeing all the angles baby, so my inner-Columbo can sleep at night.
IMHO, my point remains intact; There is a bias which is a mere cordial and harmless inclination or preference, as with the cat and dog example, and there is a prejudice which is ethically wrong, forceful, self-motivated bias.
I think the passive racists are more dangerous because, chances are, they don't know they are biased
Oh come off it. Do you think scarface the gangster leader isn't as dangerous as the pawn he uses? Lol. They're dangerous, but if they didn't have a leader they wouldn't be. So the leader with the strong form of bias is the one in control. Popping off a thousand pawns won't do anything, getting the leader will do everything. You can spend energy popping of pawns all year, or you can focus that energy into removing the real threat.
I'll admitt that soldiers are dangerous. But there's no way they're as dangerous, or as guilty, as the radical Hitler who orders them to kill and be killed. The soldier's first move to join up, is his last individual thought process, IMHO. Any viability in his soldiering, will only be fueled if he conforms. Otherwise he's outed immediately.
The pawn only does it to fit in. That's this motive(he wants a purpose). He doesn't know it, but it doesn't matter what the cause is, because his necessity is purpose.
Most people have general bias of all sorts, and it is simply a result of environment. It is harmless, as with the cat and dog example.
But it's the one who is self-motivated. He's the danger - he's the leader. He doesn't do it to go along with the group. He does it because he's an arrogant arshole that has this self-willed bias I've been talking about. He has hatred for sure. But it's his propensity to turn that hatred into active violence. It's his violent religiousity and belief in his own ego.

This message is a reply to:
 Message 36 by macaroniandcheese, posted 02-13-2006 1:14 PM macaroniandcheese has not replied

  
melatonin
Member (Idle past 6209 days)
Posts: 126
From: Cymru
Joined: 02-13-2006


Message 47 of 52 (286494)
02-14-2006 1:32 PM
Reply to: Message 44 by Silent H
02-14-2006 5:34 AM


Re: IAT
Hi Holmes,
I'll answer what I can, forgive me if I miss anything...
1) familiarity has been extensively tested and the IAT has been shown to be quite impervious to such effects. The only problems arise when the stimuli are extremely unfamiliar. The studies mentioned used pseudo-categories with nonsense words, whereby the participants needed to learn the category and exemplars.
2) I'm not too sure what to answer here. Is it a specific IAT experiment that fits this profile? I don't know of them all, there are so may of them now. But I would say that such an effect, if it was consistent, would counterbalance over the sample.
3) Cross-category associations - this shows the need for careful experimental construction. There is a need to ensure the stimuli are good exemplars of a particular category if they fall into fuzzy categories we would expect problems.
I think the underlying issue your criticisms focus on are a problem with psychological testing in general. We generally form and assess an experiment at the group sample level, rather than the individual level. There are many variables that we control for and maybe 95% of the participants fit the methodology. However, this means that 5% do not. As I mentioned in another post, I wouldn't draw to many conclusions from an individual result. In fact, I disagree with the interweb IAT reporting performance, only if it shows a strong preference is it significant.
So, if we take name familiarity, they are tested to ensure that most participants would be familiar with them, but by no means does this include all participants, just a significant level. In my own research I am using a custom-designed task whereby I select stimuli according to the individual and have found even better results with this approach (although this was not the IAT). Sadly, we submitted a manuscript with this approach and a reviewer questioned why we did, lol. This approach could be readily used for the IAT as it has been shown that as few as two stimuli in each category can produce the effect.
Personal preference - if a person actually liked to associate black/guns compared to white/guns then the test would be measuring what we hope to. Same goes for female/nurse, male/engineer etc. These are what we are interested in, it is a stereotype and/or preference. It can measure both. The reason for assessing the black-weapon association was due to an event in the US were a police officer shot a black guy holding what he thought was a gun. It has been shown many times that people use race as a cue in a shoot/no-shoot experimental situation (i.e. they will shoot faster and exhibit more errors for a black person).
Feature bias - this is known to be important and it has been shown that blacks with less african features are the targets of less bias and whites with a high degree of african features are more likely to attract bias. Again, it is important to ensure the stimuli are good exemplars of a particular category.
The Bodenhausen studies seem very good, tbey used computerised black/white faces and clearly show that black faces are viewed as showing anger earlier than white faces, and an angry face is more likely to be viewed as black. The icing on the cake - these effects were significantly correlated with IAT performance. Was it familiarity? I doubt it. Even if it was, it still shows an in-group/out-group effect.
Does the IAT measure real-world behaviour? I think it does, however, with prejudice we use controlled behaviour to negate effects of implicit bias. Only under certain circumstances would such biases raise themselves (complexity and/or ambiguity). Dovidio & Gaertner found such an effect in job applications - if there was no comparison between a black and white person, then the job went to the best applicant. But when there was ambiguity (i.e. almost facsimile profiles), then racial biases seemed to be used. A good example of IAT effect in real-world behaviour is in consumer psychology, where they readily show that the IAT can predict brand preference and consumer behaviour (e.g. macdonalds vrs burger king). There is a big slice of emotion in the IAT effects I feel, and the IAT race effect has been shown to correlate with amygdala activity (Phelps et al, 2000).
But as soon as we try to model real-world behaviour in the lab, we do lose a degree of ecological validity and of course no methodology is perfect.
Believe me, this is a hard area to research, presentations to a non-socal psych crowd are uncomfortable. I'm moving into affective biases in sport soon, lol.
If I missed anything just restate it and I'll answer it soon.

This message is a reply to:
 Message 44 by Silent H, posted 02-14-2006 5:34 AM Silent H has replied

Replies to this message:
 Message 49 by Silent H, posted 02-14-2006 6:25 PM melatonin has replied

  
melatonin
Member (Idle past 6209 days)
Posts: 126
From: Cymru
Joined: 02-13-2006


Message 48 of 52 (286503)
02-14-2006 1:49 PM
Reply to: Message 45 by riVeRraT
02-14-2006 7:17 AM


Re: Implicit Association Test
Hi riverrat,
It does measure differences on the millisecond scale. It doesn't assess individual responses, it will average responses over the trials and it only uses certain blocks, some of them are practice blocks. The high number of trials increases reliability.
As I mentioned before, reaction times are used extensively, the speed of response, if the experiment is performed to instructions, will be closely associated to speed of cognitive processing. The stroop test is based on the same processes and is used to measure cognitive abilities (it can show cognitve dysfunction - frontal lobe damage etc).
Honestly, I wouldn't worry about an individual result. It means nada.

This message is a reply to:
 Message 45 by riVeRraT, posted 02-14-2006 7:17 AM riVeRraT has not replied

  
Silent H
Member (Idle past 5820 days)
Posts: 7405
From: satellite of love
Joined: 12-11-2002


Message 49 of 52 (286641)
02-14-2006 6:25 PM
Reply to: Message 47 by melatonin
02-14-2006 1:32 PM


Re: IAT
1) familiarity has been extensively tested and the IAT has been shown to be quite impervious to such effects. The only problems arise when the stimuli are extremely unfamiliar.
I'm sorry but this suggests a lack of imperviousness. At the very least on the individual level, which is what I was discussing.
2) I'm not too sure what to answer here. Is it a specific IAT experiment that fits this profile?
Yes, I gave as my main example during criticism the arab/muslim test. It put a group of arabic sounding names up against a set of names from all different nationalities. This cannot be counterbalanced as is. A person could react to the rarity/abundance of a kind of word or image if it involves a set of similar things against a set of inherently different things.
3) Cross-category associations - this shows the need for careful experimental construction.
Agreed, yet there is no way of stopping this from cropping up, which it looks like you agree with...
We generally form and assess an experiment at the group sample level, rather than the individual level. There are many variables that we control for and maybe 95% of the participants fit the methodology. However, this means that 5% do not. As I mentioned in another post, I wouldn't draw to many conclusions from an individual result. In fact, I disagree with the interweb IAT reporting performance, only if it shows a strong preference is it significant.
This makes sense to me. Indeed it is very close to what I was trying to get across to readers. Your closeness to these tests makes your statements a bit more specific, but that's cool.
if a person actually liked to associate black/guns compared to white/guns then the test would be measuring what we hope to.
I'm not sure if that's totally true, the example I gave would certainly throw off what you would have been trying to measure due to the even you stated. I was suggesting a positive/positive reaction and thus a cop would have been less likely to shoot thinking a black person had a gun (or one that would be used to harm him), even if they associated a blacks with guns. Do you see the difference? Cultural context and even personal context can mean a world of difference.
The Bodenhausen studies seem very good, tbey used computerised black/white faces and clearly show that black faces are viewed as showing anger earlier than white faces, and an angry face is more likely to be viewed as black.
Yes you had mentioned these studies earlier. However I do not see them avoiding the same issues as above. Were these studies done cross culturally (over many different nations) and looking at differences between people raised around one type of face as the majority, versus another?
A good example of IAT effect in real-world behaviour is in consumer psychology, where they readily show that the IAT can predict brand preference and consumer behaviour
I believe this, but it is different than human behavior which involves human interaction beyond simple issues like word and facial image association. I guess where I see this best being used, or applicable, is where we are faced with pure impulse choices on scant information available in a time sensitive setting. Shopping is a good case.
IAT race effect has been shown to correlate with amygdala activity (Phelps et al, 2000).
Do you have a link to this article? I am interested in seeing that (neuro is more my interest than cog). If it was in the greenwald site I missed it.
Thank you for your responses. While I might not be as enthusiastic (or lets say I'm a bit more skeptical), I think we are on the same page with what kinds of limits this test has, particularly regarding reported results for individuals. In any case, your style is clear and concise, and I'll be interested in watching more from you at EvC.

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

This message is a reply to:
 Message 47 by melatonin, posted 02-14-2006 1:32 PM melatonin has replied

Replies to this message:
 Message 50 by melatonin, posted 02-14-2006 8:44 PM Silent H has replied

  
melatonin
Member (Idle past 6209 days)
Posts: 126
From: Cymru
Joined: 02-13-2006


Message 50 of 52 (286675)
02-14-2006 8:44 PM
Reply to: Message 49 by Silent H
02-14-2006 6:25 PM


Re: IAT
Hi Holmes,
here's a link to the phelps article...
http://www.psych.nyu.edu/...lab/abstracts/phelps_oconnor.pdf
glad to see you prefer neuro, I'm in the social neuroscience realm, it's a bit of a buzz at the moment. I don't particularly like social psychology but the neuro aspect makes it worthwhile.
"1) familiarity has been extensively tested and the IAT has been shown to be quite impervious to such effects. The only problems arise when the stimuli are extremely unfamiliar."
I'm sorry but this suggests a lack of imperviousness. At the very least on the individual level, which is what I was discussing.
Well what would be found (I'll stick to black/white names as that is what I use and know best) is that a participant would be more familiar with some names, less familiar with others. This should balance across the task. Greenwald has shown that for equally familiar names the effect holds (i.e. the effect is not due to familiarity). But I did leave my qualification till after this point - I guess I should have noted it there and then - drawing individual conclusions can be unreliable. It is doubtful that people would be completely unfamiliar with black names, enough to confound the experiment. I'll also point out, that AFAIK Greenwald et al. have not published data for the weapons and Muslim task.
If an individual was completely unfamiliar with the stimuli then I would see a problem. Like nosyned noticed with the TV category, if you don't know TV programmes there could well be cross-category issues. But with the names it would be less a problem, the fact that you call them arab-sounding is a good sign they fit the category.
Edit: just found a paper that talks about the familiarity issue, it shows that IAT effects are actually greater when equally familiar black/white names are used than familiar white vrs unfamiliar black (i.e. unfamiliarity of the out-group stimuli reduced the IAT effect).
http://faculty.washington.edu/...s/DasguptaEtAl.JESP2000.pdf
"2) I'm not too sure what to answer here. Is it a specific IAT experiment that fits this profile?"
Yes, I gave as my main example during criticism the arab/muslim test. It put a group of arabic sounding names up against a set of names from all different nationalities. This cannot be counterbalanced as is. A person could react to the rarity/abundance of a kind of word or image if it involves a set of similar things against a set of inherently different things.
Again, I would have thought Greenwald would have balanced for familiarity for the stimuli. I'll check this out but I assume the task assesses muslim out-group vrs other out-groups. He would probably predict a greater association of bad/muslim-good/other compared to the reverse. If the category is clearly muslim vrs other foreign names then the categories are sound. It would only be a relative comparison. If he defined english vrs american names then there would definitely be cross-category problems.
"3) Cross-category associations - this shows the need for careful experimental construction."
Agreed, yet there is no way of stopping this from cropping up, which it looks like you agree with...
Yes I agree, it's a problem but I think it can be readily ameliorated. If we tried to assess american names vrs UK names then we would ahve a problem. we need to ensure the group are distnictive and well-defined.
"if a person actually liked to associate black/guns compared to white/guns then the test would be measuring what we hope to.2
I'm not sure if that's totally true, the example I gave would certainly throw off what you would have been trying to measure due to the even you stated. I was suggesting a positive/positive reaction and thus a cop would have been less likely to shoot thinking a black person had a gun (or one that would be used to harm him), even if they associated a blacks with guns. Do you see the difference? Cultural context and even personal context can mean a world of difference.
I understand what you mean, but the task would be measuring the strength of association, so the task does what it is expected. I don't think the weapon task makes a good/bad distinction. This would cause problems for Greenwald drawing a strong conclusion that it is due to violence association - just that, generally, people associate weapons with blacks more than whites with no good/bad conclusion.
Cultural differences and beliefs are important, hence the effect in the politics IAT. There was a very interesting neuro study on politics and bias by Drew Weston (if you're interested I'll pass a link on, it's only a report of his presentation though). So for the Bodenhausen task we can only draw conclusions for the groups tested, we cannot really extrapolate to even europeans, as he tested in the US, but I see no reason to not allow this generalisation. The IAT works as well here in the UK, as it does in the US. As Bodenhausen made no assessment of out-group contact, we can make no conclusion about this, just that the greater bias exhibited via IAT, the greater race prejudice behaviour in face perception.
"A good example of IAT effect in real-world behaviour is in consumer psychology, where they readily show that the IAT can predict brand preference and consumer behaviour"
I believe this, but it is different than human behavior which involves human interaction beyond simple issues like word and facial image association. I guess where I see this best being used, or applicable, is where we are faced with pure impulse choices on scant information available in a time sensitive setting. Shopping is a good case.
But this is when we expect such implicit race biases to be important, snap judgements made with ambigious information and/or in situations of complexity. If the decision is easy or you have excessive time to assess the info, controlled processes will be greatest and implicit effects minimised. Implicit bias is fast, unconscious, and automatic (and in the situations I suggest, uncontrollable).
Thank you for your responses. While I might not be as enthusiastic (or lets say I'm a bit more skeptical), I think we are on the same page with what kinds of limits this test has, particularly regarding reported results for individuals. In any case, your style is clear and concise, and I'll be interested in watching more from you at EvC.
Cheers Holmes, I don't spend much time on forums, I generally just lurk but I thought I could make a contribution on this topic. I make this clear to the undergrads I teach research methods to - we can be limited in the conclusions we draw on an individual basis, results can be applied generally, and our methods are never perfect. Psychology does seem to ignore the outliers and focus on the imaginary "statistical mean" individual. But we do well, even with the limitations of method.
What I would suggest is that you actually e-mail Anthony Greenwald, he would probably be very interested in these criticisms of the arab/weapon task, he could probably give you more extensive answers -I only use the race task and it is a sound, reliable, and robust experiment
This message has been edited by melatonin, 02-14-2006 08:55 PM
This message has been edited by melatonin, 02-14-2006 09:40 PM

This message is a reply to:
 Message 49 by Silent H, posted 02-14-2006 6:25 PM Silent H has replied

Replies to this message:
 Message 51 by Silent H, posted 02-15-2006 5:33 AM melatonin has replied

  
Silent H
Member (Idle past 5820 days)
Posts: 7405
From: satellite of love
Joined: 12-11-2002


Message 51 of 52 (286770)
02-15-2006 5:33 AM
Reply to: Message 50 by melatonin
02-14-2006 8:44 PM


Re: IAT
glad to see you prefer neuro, I'm in the social neuroscience realm, it's a bit of a buzz at the moment. I don't particularly like social psychology but the neuro aspect makes it worthwhile.
I minored in sociology/anthro which obviously comes at things from a soc.psych angle. I was left pretty skeptical and very hardnosed about methodology (maybe my methodology teacher was better than I thought). It sort of left me with an idea that studying the brain itself would be more interesting and fruitful. My gf is in neuro which is cool for me as it allows me a backdoor method to info I'd like to have but at this point can't afford to shell out money for.
I don't think there's much left to debate on this issue between us. Perhaps a few points of clarification...
have not published data for the weapons and Muslim task...
If the category is clearly muslim vrs other foreign names then the categories are sound.
It wasn't weapons and muslims, it was actually muslim names vs a diverse list of names from other ethnicities. That to my mind cannot produce the desired effect as the lists of other foreign names from any particular ethnicity is by necessity shorter. One might have a preference (that is a positive or negative association) with different members of the second list though not all of them, creating a skew. Thus one cannot say it measures muslim vs ALL others. Indeed one could vastly prefer muslim vs hebrew or japanese, but because so many other names are on the second list it gets drowned out.
Then there is also the issue of diversity/rarity. Humans do tend to get tired of something that is repititous. It can become relatively irritating. Thus the increased frequency of muslim names as compared to any particular other sounding name could have a negative effect. Conversely one might start looking forward to another "cool" sounding name from the other list and its rarity in the test make one prize it and so reward it with a more positive association.
This goes for image tests as well. An example of that would be the jewish vs other religions. It would key on those same factors.
I guess this is to say it seems to one might also end up "testing" associations of homogenous v heterogenous lists, rather than anything on those lists.
I forgot to mention another issue with regard to imagery and that could play a role and that is aesthetic quality of the image itself. Color schemes and positioning should match as well as resolution and clarity. Both the Jewish v Other Religions and Bush vs FDR seemed problematic to me based on those qualities.
I know I am (at this point) pretty negative about Bush, and while I was glad to see something saying I had absolutely no preference (or bias), my thought was how much that might have been effected by the fact that I couldn't stand the quality of imagery for FDR.
This would cause problems for Greenwald drawing a strong conclusion that it is due to violence association
Ah yes, I guess the way to put this is that an association may be valid, but an assumed implication of an association may not be.
I generally just lurk but I thought I could make a contribution on this topic.
You certainly did for me, so I suspect it helped others too. Very good supporting info as well.

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

This message is a reply to:
 Message 50 by melatonin, posted 02-14-2006 8:44 PM melatonin has replied

Replies to this message:
 Message 52 by melatonin, posted 02-15-2006 12:23 PM Silent H has not replied

  
melatonin
Member (Idle past 6209 days)
Posts: 126
From: Cymru
Joined: 02-13-2006


Message 52 of 52 (286908)
02-15-2006 12:23 PM
Reply to: Message 51 by Silent H
02-15-2006 5:33 AM


Re: IAT
Hey Holmes,
I seemed to have a problem with plurals in that post, lol.
What I meant is - the taskS weapons/race and muslims/others have not been published to my knowledge; I certainly don't recognise them from the literature, many of the experiments on the harvard site are not published.
The criticisms you raise on the muslim/others are definitely valid and the conclusions that can be drawn from the weapon/race study are limited. Same for the stimuli issue, they should be well matched in terms of features/composition. I'll probably go and complete some of these other IATs sometime and see these issues for myself.
Methodological rigour is most important in psychology, behaviour is very complex, and we need to control as many extraneous factors as possible. Although, we can never cover everything.
Edited to add: ah, the dreaded participant fatigue problem, again we try to control for such problems by counterbalancing across the sample. So half the sample will get bored/tired on one block, the other half on the other block. It is an issue, I did enough 2 hour computerised perception experiments as an undergrad to know this from experience.
Cheers Holmes
This message has been edited by melatonin, 02-15-2006 01:42 PM

This message is a reply to:
 Message 51 by Silent H, posted 02-15-2006 5:33 AM Silent H has not replied

  
Newer Topic | Older Topic
Jump to:


Copyright 2001-2023 by EvC Forum, All Rights Reserved

™ Version 4.2
Innovative software from Qwixotic © 2024