AI face analysis used to predict MBA pay, researchers claim
(2026/02/10)
- Reference: 1770758698
- News link: https://www.theregister.co.uk/2026/02/10/ai_face_analysis_mba_pay/
- Source link:
A picture is worth a thousand words or, perhaps, a hundred thousand dollars in extra salary. Academics claim that personality traits inferred using AI photo analysis can predict how depicted individuals will fare in the labor market.
They emphasize that they don't advocate doing so because personality extraction from facial images is fundamentally discriminatory.
Even so, they say, personality screening is already commonplace among admissions and HR committees, and AI tools that offer personality assessment are seeing rapid adoption. So they argue that an academic evaluation of the technology is necessary.
[1]
In [2]a paper titled "AI Personality Extraction from Faces: Labor Market Implications," authors Marius Guenzel (Wharton), Shimon Kogan (Reichman University), Marina Niessner (Indiana University), and Kelly Shue (Yale University) describe how they used the LinkedIn facial images of over 96,000 MBA graduates to extract subjects' Big Five personality traits – Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism.
[3]
[4]
The machine learning algorithm used was originally described in [5]a 2020 Scientific Reports paper titled "Assessing the Big Five personality traits using real-life static facial images," one of about two dozen papers cited as "ML-laundered junk science" in [6]a 2024 paper titled "The reanimation of pseudoscience in machine learning and its ethical repercussions."
The algorithm "uses facial features to predict self-reported personality, rather than others' perceptions of personality based on visual appearance," according to Guenzel, Kogan, Niessner, and Shue.
[7]
By applying this algorithm, the authors found "that personality traits inferred from facial features provide substantial incremental predictive power for labor market outcomes."
[8]Anthropic's Claude Opus 4.6 spends $20K trying to write a C compiler
[9]Oracle Java licensing worries are percolating through the userbase
[10]OpenAI introduces ads...for the people!
[11]Microsoft boffins figured out how to break LLM safety guardrails with one simple prompt
The researchers determined that applying machine learning to infer personality traits from facial images produced accurate predictions for the rank of undergraduate and MBA programs attended by the depicted individual, initial compensation, salary trajectory, and job transitions.
So, based on these results, were an HR company to use a similar technique to assess the personality of managerial applicants, the result could serve as a forecast of the job applicant's future performance in the labor market – biased though it may be. And that appears to be happening.
Co-author Marina Niessner, assistant professor of finance at Indiana University, told The Register in a phone interview that companies like banks already use personality surveys in hiring and promotion decisions and that AI hiring companies are starting to use technology like Big Five personality trait analysis on video interviews.
"The regulatory environment, as you probably know, is very uncertain," said Niessner. "And so we don't think this is necessarily a valid way to do it [or] that companies should be doing it. But I think it's really important to have an academic evaluation of these methodologies if there's even going to be a regulatory discussion around this."
[12]
The paper argues that the AI-based screening needs to be considered in conjunction with the alternative, which is human decisions based on physical appearance that may also be inconsistent or biased. ®
Get our [13]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://www.nber.org/papers/w34808
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.nature.com/articles/s41598-020-65358-6
[6] https://pmc.ncbi.nlm.nih.gov/articles/PMC11573791/
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://www.theregister.com/2026/02/09/claude_opus_46_compiler/
[9] https://www.theregister.com/2026/02/10/oracle_java_licensing/
[10] https://www.theregister.com/2026/02/10/openai_ads/
[11] https://www.theregister.com/2026/02/09/microsoft_one_prompt_attack/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[13] https://whitepapers.theregister.com/
They emphasize that they don't advocate doing so because personality extraction from facial images is fundamentally discriminatory.
Even so, they say, personality screening is already commonplace among admissions and HR committees, and AI tools that offer personality assessment are seeing rapid adoption. So they argue that an academic evaluation of the technology is necessary.
[1]
In [2]a paper titled "AI Personality Extraction from Faces: Labor Market Implications," authors Marius Guenzel (Wharton), Shimon Kogan (Reichman University), Marina Niessner (Indiana University), and Kelly Shue (Yale University) describe how they used the LinkedIn facial images of over 96,000 MBA graduates to extract subjects' Big Five personality traits – Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism.
[3]
[4]
The machine learning algorithm used was originally described in [5]a 2020 Scientific Reports paper titled "Assessing the Big Five personality traits using real-life static facial images," one of about two dozen papers cited as "ML-laundered junk science" in [6]a 2024 paper titled "The reanimation of pseudoscience in machine learning and its ethical repercussions."
The algorithm "uses facial features to predict self-reported personality, rather than others' perceptions of personality based on visual appearance," according to Guenzel, Kogan, Niessner, and Shue.
[7]
By applying this algorithm, the authors found "that personality traits inferred from facial features provide substantial incremental predictive power for labor market outcomes."
[8]Anthropic's Claude Opus 4.6 spends $20K trying to write a C compiler
[9]Oracle Java licensing worries are percolating through the userbase
[10]OpenAI introduces ads...for the people!
[11]Microsoft boffins figured out how to break LLM safety guardrails with one simple prompt
The researchers determined that applying machine learning to infer personality traits from facial images produced accurate predictions for the rank of undergraduate and MBA programs attended by the depicted individual, initial compensation, salary trajectory, and job transitions.
So, based on these results, were an HR company to use a similar technique to assess the personality of managerial applicants, the result could serve as a forecast of the job applicant's future performance in the labor market – biased though it may be. And that appears to be happening.
Co-author Marina Niessner, assistant professor of finance at Indiana University, told The Register in a phone interview that companies like banks already use personality surveys in hiring and promotion decisions and that AI hiring companies are starting to use technology like Big Five personality trait analysis on video interviews.
"The regulatory environment, as you probably know, is very uncertain," said Niessner. "And so we don't think this is necessarily a valid way to do it [or] that companies should be doing it. But I think it's really important to have an academic evaluation of these methodologies if there's even going to be a regulatory discussion around this."
[12]
The paper argues that the AI-based screening needs to be considered in conjunction with the alternative, which is human decisions based on physical appearance that may also be inconsistent or biased. ®
Get our [13]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://www.nber.org/papers/w34808
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[5] https://www.nature.com/articles/s41598-020-65358-6
[6] https://pmc.ncbi.nlm.nih.gov/articles/PMC11573791/
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[8] https://www.theregister.com/2026/02/09/claude_opus_46_compiler/
[9] https://www.theregister.com/2026/02/10/oracle_java_licensing/
[10] https://www.theregister.com/2026/02/10/openai_ads/
[11] https://www.theregister.com/2026/02/09/microsoft_one_prompt_attack/
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aYu4kR0_fDDBui0S-G-eawAAAk0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[13] https://whitepapers.theregister.com/
Re: Isn't that a less-US problem?
DS999
This is about predicting salaries. For the kind of roles MBAs are hired for you will have met in person at least once, perhaps multiple times, and their evaluation of your appearance (even on a subconscious level) will have been taken into account in the offer they make. Someone who "looks the part" might get tens of thousands of dollars more for their opening offer.
That would likely be based on more than just your face though. Height has also been shown to be a predictor of higher salary, more promotions etc. If you walk with a limp or you're fat or you slurp your coffee that's probably a negative, having a deeper voice (for a man) is probably a positive, etc. etc. There are tons of subconscious things that can bias such decisions even if they take great care to minimize the opportunities for discrimination.
Doctor Syntax
The ML barrel scrapers are steadily working their way down. I doubt they're at the bottom yet.
Isn't that a less-US problem?
As far as my limited knowledge goes: In USA it is not desired to base hiring on a photo sent along with the application to such an extend that it is even frowned upon sending in a photo. A side effect of anti racism and anti discrimination laws. So that first hurdle is at least taken, and when it comes to the personal or video interview - well that is a different topic, there you might have such an AI checking your face live.