Jesse Singal writes:
This was presented, in Jennifer Eberhardt’s book Biased, as evidence to support the idea that even positive portrayals of black characters could be spreading and exacerbating unconscious antiblack bias. I did not see evidence to support that idea.
I replied:
I don’t understand what you’re saying here. I clicked thru and the article seems reasonable enough, for what it is. As you probably know, I’m not a big fan of these implicit bias tests. But I didn’t think the article was making any statements about positive portrayals of black characters. I thought they were saying that even for shows for which viewers perceived the black characters as being portrayed positively, a more objective measure showed the black characters being portrayed more negatively than the whites. I didn’t go thru all the details so maybe there’s something off in how they did their statistical adjustment, but the basic point seemed reasonable, no?
Singal responded:
Yeah, I didn’t include much detail. Basically it is this thing I see a ton of in social-priming-related research where people extrapolate, from results that appear to me to be fairly unimpressive, rather big claims about the ostensible impact of priming stuff on human behavior/attitudes in the real world. I think this table is key:
This was from when they edited out black and white characters and asked people unfamiliar with the shows how they perceived the characters in question. The researchers appear to have tested six different things, found one that statistically significant (but only barely), and gone all-in on that one, explanations-wise. Then by the time the finding is translated to Eberhardt’s book, where all the nuance is taken out (we don’t hear that in five of the six things they tested they found nothing), we’re told that it could be that even black characters who are portrayed positively on TV—the subject of this story—could be spreading implicit bias throughout the land.
I don’t really have a strong take on all this, but I thought it could be useful to post on this, just because sometimes maybe it’s a good idea to express this sort of uncertainty in judgment. In any sort of writing there is a pressure to come to a strong conclusion—less pressure in blogging than on other media, perhaps, but still there’s some pull toward certainty. In this case I’ll just leave the discussion where we have it here.
Tomorrow’s Post: Bank Shot
Quoting from the abstract of the linked paper:
“biased facial expressions and body language may resist conscious identification and thus produce a hidden social influence”
is this idea that a given facial expression has some universal meaning solidly established? (much less that some form of racial bias can be interpreted from it! What gnosh!) Pardon me for doubting this purportedly universal truth! Read any psychology self-help book. It will tell you that you can’t interpret what other people are thinking. It’s hard to imagine how a used car salesman could be successful if it were possible to do so. Much less Stalin or Hitler.
See, now what can you do statistically with this? Zippo. Mr. P has nothing meaningful to say, NHST or no. The data might as well be random numbers, and the results seem to confirm that.
Paul Ekman has certainly spent some time trying to establish exactly that. Was he successful in doing so?
https://www.paulekman.com/resources/universal-facial-expressions/
From the Ekman Group’s website:
“Dr. Ekman discovered strong evidence of universality* of some facial expressions of emotion as well as why expressions may appear differently across cultures.”
That statement leaves *A LOT* of open ground. It barely claims anything: a smile means happy a frown means sad, we all know that! It doesn’t even get close to claiming validity for the idea of microexpressions.
I like the page on microexpressions. Reads like a late night infomercial.
One can take that perspective and the implication is that there is no such thing as a valid finding in all of social science, including economics and political science, given the reality that there are almost no results for which context does not or cannot change an outcome.
However, I think that is a misreading of the claim. What the statement implies is that facial expressions of emotion are at the simplest level have been found to generalize across cultures (i.e., smile=happy, frown=sad, etc.) and that more complex expressions have been found to be valid within each culture and not across cultures.
“the implication is that there is no such thing as a valid finding in all of social science”
I don’t agree, although I fear much social science research is vulnerable.
I don’t doubt this general statement. Seems reasonable enough:
“Facial expressions of emotion are at the simplest level have been found to generalize across cultures (i.e., smile=happy, frown=sad, etc.) and that more complex expressions have been found to be valid within each culture and not across cultures.”
But the devil is in the details: How much more complex? How consistent is the association? It’s one thing to show that people generally associate a given facial expression with a general emotion. That’s pretty straightforward and conceivably reliable. It’s an immense jump to infer details about what the person who is displaying a given expression is feeling – much less that ms-variations in expressions have interpretable detailed meanings about things like racial bias. My guess is that such an idea would be extremely vulnerable to scrutiny.
Human beings are professional liars and – calling on anecdotal experience I admit – there’s every reason to believe their ability to lie extends to both body language and facial expressions.
My apologies. I think I gave the impression of Ekman’s research. Ekman did not establish that one could infer racial bias from facial expressions.
It looks like I muddied the water here by introducing Ekman’s research into a discussion about implicit bias research. Though there may be a bridge between the two, it is not as direct as my dropping the link into the discussion without explanation may have made it seem.
To your point, Ekman established universality for basic emotional expression and the human ability to correctly detect and distinguish among these. And later an individual difference in the skill to detect deception that generalized across contexts.
In contrast, implicit bias tests are typically conducted by the presentation of two pictures which differ by skin color to which respondents select one or the other picture in response to a second descriptive stimulus of some sort such as “reliable” or “attractive” or “intelligent”, etc. Conducted over repeated trials I assume that departures from the “no bias” = 50% are interpreted as a bias in one direction or the other.
While I have conducted research in the area of indirect measures of a psychological construct and have taken IAT tests, I have never administered them and thus am not familiar with all of the scoring protocols. That said, there are a few things that concern me about the methodology as typically presented. As anyone who reads this blog regularly would know, the number of HEADS (streaks) that can come up in a row for a coin that has pr = 0.5 is as high as 12 in some simulations I ran using code Bob Carpenter posted here. Thus, drawing a strong inference within person would not be warranted given the number of trials I experienced when I took the test. The other issue is that of balancing pictures for what may seem like extraneous characteristics, but which may in fact be causally related to responses. When I took the test there were only two pictures used, thus none of those other characteristics could have possibly been balanced across trials.
You also mentioned the idea that it would not be possible to read mental content from facial expressions. I tend to think this is true as well, but I do remember reading some of Ekman’s work (and it’s been quite a while) where he was asserting that certain areas of the face related to certain types of content, but I cannot recall where. I think it was in his book:
Ekman, P., & Rosenberg, E. L. (Eds.). (2005). What the face reveals: Basic and applied studies of spontaneous expression using the Facial Action Coding System (2nd ed.). New York: Oxford University Press.
or maybe
Ekman, P. (2001). Telling lies: Clues to deceit in the marketplace, politics, and marriage (3rd ed.). New York: W. W. Norton.
I think nervousness can be reliably detected, but correctly identifying the actual cause of that nervousness would require far more information than could be provided by the facial response alone.
Curious,
Thanks, that seems generally within the scope of reasonable possibilities. I’m no expert on this topic. My background is in geology. But I read enough to know some things – like that reliable lie detection isn’t possible. So that says a lot about how well we can read what’s really going on in people’s minds.
I know some poker players who make a living proving you wrong.
Actually I’ve read a bit about top poker players. I read about one top player who, after becoming well established, went on a two year losing streak. Maybe her facial expressions were giving her hand away! But apparently she learned how to conceal them again when she got back on a winning trend?
Likely the key factor all top players share is their knowledge of the odds. But everyone can learn the odds – lots of work, but the most dedicated can all do it. After that, a key ability would be to conceal their emotions and read others’ emotions. This would be the main variation among players. But the streakiness of winning and losing suggests that they don’t have scientific knowledge of how to hide their own emotions or identify others’ emotions.
And of course the top player circuit is probably a few tens of players. They all know each other, have played each other many times so they can rely on that experience to assess others’ emotions.
It is also important to keep in mind the distinction between whether something exists in reality and whether the methods being used can reliably detect it. I believe racism in the year 2019 continues to be a widespread problem. I think the breadth of this problem can be assessed by directly measuring beliefs that:
1. Assert a superiority between races.
2. Deny historical racism.
3. Deny multi-generational effects of historical racism (Slavery ended over a hundred years ago).
4. Social norms reflected in popular culture that are racist (Black NFL Quarterbacks are a fairly recent phenomenon within my lifetime.).
5. Deny that current strains of white supremacy are actually instances of white supremacy.
6. Group differences on crude measures of crude constructs reflect a form of superiority between races.
7. Beliefs that performance as measured in academics or at work are devoid of racial bias themselves or that it can easily be adjusted for statistically to remove such bias.
The list of potential beliefs is much longer. And the ways in which social structures that result in continued disparate outcomes are justified gives a window into how a racist culture is sustained in the face of many well-intentioned people and why it is so difficult to change.
The framing of a bias as implicit suggests that there are people who do know or do not believe they have beliefs about other racial groups that are supportive of either personal or structural racism. I believe this does exist and I believe that it exists in the beliefs in idealized forms of meritocracy that do not actually exist in reality.
I too like to place the word ‘deny’ in front of the positions of people I don’t agree with.
While posting about framing of bias.
I haven’t heard of the book, but “even positive portrayals of black characters could be spreading and exacerbating unconscious antiblack bias” seems to be a taking a bit too far. For one, it is quite unclear that the supposed positive portrayals in the study were positive. Just from reading the list of characters they used and the scenes, there are a couple of cases where they used a “white” protagonist and a “black” antagonist, with corresponding positive and negative ratings. Even if the particular scene was somehow balanced, surely the faces of the other characters would show some alignment with that longer term role of the character? In addition, the overall negative average ratings associated with some characters would seem to indicate they are not portrayed positively? The supplemental material Table S1 does not inspire confidence in the result.
“even positive portrayals of black characters could be spreading and exacerbating unconscious antiblack bias” seems an uncontroversial almost unchangeable claim. I doubt, however, that the claim can be made into a well formed question that can be studied scientifically. Uncle Tom from Uncle Tom’s Cabin was meant by Harriet Beecher Stowe at the time to be a positive portrayal of a slave and to cause Northern whites to sympathize with the plight of slaves. It unquestionably worked. However, not a generation later, African American resented the portrayal. “Uncle Tom” became an insult among African Americans. “A positive portrayal can spread antiblack bias” has got to be true, but it is pretty empty absent an account of how the portrayal “spreads” the bias. Any story can be reinterpreted to have a different meaning. Any positive story about an African American can be reinterpreted by a racist to have some negative meaning. Social scientists need more rigor in the claims that they are making. If someone spent a few hundred hours thinking about what “spreads” means or what implicit bias was other than just thinking it is identical to the test results, than maybe we would make a lot more progress on these questions.
Good points.