A recent paper in the Proceedings of the National Academy of Sciences (PNAS) claims to have developed a machine learning model that can infer over 30 personality or psychological traits of a person from simply looking at a picture of them. They used deep generative image models to create photorealistic pictures of different faces and combined that with over one million judgements to infer physical traits such as age and happiness, but also personality traits such as trustworthiness, smart, liberal/conservative, Middle-Eastern, gay, and dorky.
One of the authors (Joshua Peterson) announced the paper in a Twitter thread. He noted:
Note that we study attribute *inferences* (impressions), which have no necessary correspondence to the actual identities, attitudes, or competencies of people whom the images resemble or depict. Put another way, our dataset not only contains bias, it deliberately reflects it.
He also pointed out that they can use their model to manipulate images by trait, so one could take a photo and increase its perceived trustworthiness without changing any of the other features, and he invites people to upload their own photos for a demonstration.
I'm sure my fellow Soylentils will agree that this kind of research will never be used out of context nor exploited in any untoward manner (oh, by the way, one of the traits is "electability").
I wonder how this relates to Resting Bitch Face?
Journal Reference:
Joshua C. Peterson, Stefan Uddenberg, Thomas L. Griffiths, Alexander Todorov, and Jordan W. Suchow,Deep models of superficial face judgments [open], PNAS, 119, 117, 2022.
DOI: 10.1073/pnas.2115228119
Related Stories
Humans are very perceptive to facial expressions and face-to-face interactions are strongly influenced by detecting very subtle changes in those expressions. For instance, we know someone is pulling our leg when we see a "twinkle in their eye", or when you're talking to someone and you see a very small change in position of their eyebrows, that can indicate a change in their mood from indifference to annoyance. We can form initial impressions of people based solely on their expressions, and some people just rub us the wrong way because they might have RBF: Resting Bitch Face (also called Bitchy Resting Face). These are people who always seem to look annoyed or pissed-off, but it might not be their fault. A couple of behavioral researchers think they have figured out why we react so negatively to these expressions and they've posted their research.
We all know the face. No, not just the face, but that face. That look that she swears is not a look. She says she's not angry; she reassures you she's having fun. But her face has been "throwing shade" all night – without saying anything, that face is indicating that she is not happy; more than not happy, she's about to make your night miserable too.
They ran thousands of pictures through a facial recognition program designed to detect emotions. What they saw was that most neutral expression faces registered low indicators of emotion, but fairly equally distributed across emotions; however, pictures of people who are said to exhibit RBF all seem to consistently register strong indicators of contempt. People seem to develop stronger opinions against those who look like they're showing contempt rather than those who look angry or grumpy.
While on the same continuum as anger, the philosopher Robert C. Solomon and others, such as researcher John Gottman, treat contempt differently – it's more a blend of anger and sadness directed at those deemed below you. Contempt is maladaptive in that it is particularly toxic when it develops between people in a relationship.
Do you have RBF? Remember, admission is the first step to recovery.
(Score: 3, Insightful) by Anonymous Coward on Tuesday April 26 2022, @03:49AM (2 children)
> manipulate images by trait, so one could take a photo and increase its perceived trustworthiness
> will never be used out of context nor exploited in any untoward manner
It should've been obvious from the first quoted line this is research the Advertising Industry wants.
(Score: 2) by ikanreed on Tuesday April 26 2022, @02:42PM
Manipulate?
No, way to time consuming.
Create a GAN that generates trustworthy people? Done in 2 hours.
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @08:40PM
No no, this one will also read your palm and tell the first name of your future husband. You gullible idiot.
(Score: 3, Insightful) by Runaway1956 on Tuesday April 26 2022, @04:22AM (8 children)
in with the new
Oh, the resting bitch face? What about those bitches who never rest?
Abortion is the number one killed of children in the United States.
(Score: 1, Touché) by Anonymous Coward on Tuesday April 26 2022, @04:33AM (5 children)
Can the algorithm detect INCELS?
(Score: 2) by Runaway1956 on Tuesday April 26 2022, @04:52AM (4 children)
It will probably work as well as phrenology. The advantage with this method is, you don't actually have to touch a person to decide that you don't like him.
Abortion is the number one killed of children in the United States.
(Score: 0, Spam) by Anonymous Coward on Tuesday April 26 2022, @06:30AM (3 children)
We got a Full Frontal Runaway1956 pic, from the dastardly doxxing done upon him.
But, I am sure that any competent soylentil could discover an actual picture of the Runaway, complete with shit-eating grin, before he realized that posting incinderary right wing disinformation might be a threat to his livelihood, his family, and his manhood. But, now he does, after reporting a false phishing attempt, and reportage to his employer (which, if he keeps on, someone really ought to do. Post for details.) and so here we are. Ugly pic of the real Runaway1956. Can you read his character from his picture, without considering all the right-wing racist crap he has posted? I mean, the man looks like the archtypical Right-wing Nut-job. We have to take his word on the guns, and, his word, not too reliable.
(Score: -1, Troll) by Anonymous Coward on Tuesday April 26 2022, @06:35AM
I hear the University of Arkansas, such as it it, has a brain-pan mold of Runaway, from back when they were trying to prove the Neathanderal cross-breeding hypothesis. Runaway was a good candidate, since he retained the gene to fuck anything in sight, including farm animals. So, now we know.
(Score: -1, Spam) by Anonymous Coward on Tuesday April 26 2022, @07:02AM (1 child)
OK, for those without the Google-fu, here is the link: https://www.cx500forum.com/d1/avatars/m/4/4665.jpg?1580718039 [cx500forum.com]
Warned, have been you!
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @07:29AM
And here is a pic of aristarchus! So there!
https://en.wikipedia.org/wiki/File:Aristarchos_von_Samos_(Denkmal).jpeg [wikipedia.org]
Of course, no one ever erected a statue of Runaway1956. See, Jane of Mudtown, Firefly episode 7. He got a statue! Of course, slightly later, Jane betrays the crew, by storming the Capitol and stealing medicine. Oh, wait, different Runaway, there, and Firefly is not at all relevant. Except for the loyalty to crew, and fealty to oaths, all of which the traitor Runaway1956 has violated. Death is too good for scum like him. Out the airlock, Space 'em, I say!
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @04:09PM
Are you talking about yourself?
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @08:42PM
> Oh, the resting bitch face? What about those bitches who never rest?
Why not both? The lady with the resting bitch face that nonetheless works hard to be a bitch.
(Score: 1, Touché) by Anonymous Coward on Tuesday April 26 2022, @05:03AM
Vast improvement in accuracy when coupled with measurements of the skull.
(Score: 4, Informative) by mhajicek on Tuesday April 26 2022, @05:53AM
Being Middle-Eastern is a personality trait?
Also, from what I've read, the algorithm is not inferring personality traits such as trustworthiness etc., it's inferring how a human observer would rate that trait based purely on the picture. That's a very different thing.
The spacelike surfaces of time foliations can have a cusp at the surface of discontinuity. - P. Hajicek
(Score: 2) by stretch611 on Tuesday April 26 2022, @05:53AM (2 children)
Proof that Gay-dar does exist.
https://www.youtube.com/watch?v=X4XzCP8sQqU [youtube.com]
Now with 5 covid vaccine shots/boosters altering my DNA :P
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @06:14AM (1 child)
this model was performing Gaydar the worst Rsq~0.2 actually.
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @08:45PM
Should've trained on guys that are having buttsex. 100% gay.
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @06:25AM
https://en.wikipedia.org/wiki/Physiognomy [wikipedia.org]
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @12:58PM (3 children)
I could see this working only for a limited subset of traits and photos. If some has dyed their hair a shade of neon crayon color, it's a safe bet they are a left winger, and probably an angry one. If the face can be reliably recognized as female (and I have my doubts), and the hair is cut very short like a boy and she is wearing square glasses, she is probably a lesbian. I don't think this in general can be that accurate, but I would love to see the results of their classifications to compare to how I would rate someone by the photo. The most reliable tells are those the person chooses to make to their own appearance.
(Score: 4, Insightful) by HiThere on Tuesday April 26 2022, @01:33PM (1 child)
They weren't trying to be accurate, they were trying to identify the stereotypes that people use.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 4, Insightful) by acid andy on Tuesday April 26 2022, @02:53PM
You can see how easy and tempting it is for people to lose sight of that fact though. Convenient.
Master of the science of the art of the science of art.
(Score: 0) by Anonymous Coward on Tuesday April 26 2022, @08:50PM
> The most reliable tells are those the person chooses to make to their own appearance.
I mean, duh. Does it also help to read the writing on their red baseball cap as they drive their pickup truck to a political rally by President Donald J. Trump, the greatest of all time? There are clues there if you look hard enough.
(Score: 4, Interesting) by gznork26 on Tuesday April 26 2022, @03:35PM
There's a Michael Creighton movie from 1981 called "Looker" that had a marketing icon attempting a similar feat through far cruder methods. The focus character, played by Albert Finney, was a plastic surgeon who grew suspicious of all the requests by models for very specific changes, and the subsequent deaths of those models. What he discovered was that the models were given the specs for their changes so they could get work in the company's ads. It seems the marketing company, which used tech to track where viewers were looking during video ads, found that although the models with surgery worked perfectly in stills, their movements weren't perfect, and the eyeballs drifted, causing lost revenue. So the solution was to digitize the perfected models, kill them, and use the digital versions in the ads. What is amusing today is that Creighton imagined inserting the digital models into mechanized physical sets, rather than making the entire ad digitally. But this lack of tech also made it possible for the climax to be a fight between surgeon and evil marketer on that mechanized set with the cameras running.
(Score: -1, Offtopic) by Anonymous Coward on Tuesday April 26 2022, @04:59PM (4 children)
I looked at a photo of a Democrat and inferred that he was planning to raise my taxes.
(Score: 3, Informative) by Anonymous Coward on Tuesday April 26 2022, @08:16PM (3 children)
Are you wealthy? If nit then trump raised your taxes while cutting taxes for the wealthy. Of course he delayed some of the increases to be enabled after 2024 in case a dem won, that way they could make dems look bad for things the cons do thus keeping their rightwing herd in the ranch.
(Score: -1, Offtopic) by Anonymous Coward on Tuesday April 26 2022, @08:57PM (1 child)
Sounds like sour grapes. Why don't you try winning elections instead?
(Score: 0) by Anonymous Coward on Wednesday April 27 2022, @06:44PM
Let me guess, you're one of those stable geniuses who knows 1/6 was just a little fun amongst nazis and not an attempted coup of the United States Federal Government? You are? How shocking!!!!!!!!!!
(Score: 0) by Anonymous Coward on Wednesday April 27 2022, @03:37AM
Workers who make less and retirees are hit hardest, though.