Marius Guenzel and Shimon Kogan, Wharton finance professors, teamed up with researchers from Yale and Indiana University to investigate whether artificial intelligence (AI) can infer personality traits from facial images in ways that help predict labor market outcomes.
The connection between facial features and personality traits may seem far-fetched, but Guenzel and Kogan considered various biological and social mechanisms that could link the two. Genetics and hormone exposure can influence both craniofacial structure and personality development. Additionally, how a person is perceived (and how they present themselves in photos) can reflect or reinforce certain traits. For example, studies have found that people with more symmetrical faces tend to score higher on extraversion, while “babyfaced” features are correlated with perceptions of warmth and naivety, influencing behavior towards higher levels of agreeableness.
“As researchers, we’re excited about the potential to study personality at scale and gain deeper insights into how it shapes career outcomes,” said Guenzel. “But as a society, we need to be mindful of how this technology is applied, particularly when it relies on traits that individuals can’t easily change. In addition to ethical concerns, there’s also a real risk that meaningful personal growth could be overlooked — and ultimately disincentivized — simply because it doesn’t show up on someone’s face.”
The researchers emphasize that their study should not be viewed as support for using facially inferred personality data in employment decisions. Rather, it highlights the ethical concerns such technologies raise and calls for deeper examination and public discourse around the ethical, practical, and societal implications of their use.
“As a society, we need to be mindful of how this technology is applied, particularly when it relies on traits that individuals can’t easily change.”— Marius Guenzel
Here are the key findings from their research:
AI can be trained to infer personality traits from a single photo.
To estimate personality traits from facial images, the researchers used a convolutional neural network (a type of deep learning model designed to process and understand visual data) trained on a separate dataset where individuals had provided photos of themselves and completed standard personality assessments, known in psychology as the “Big 5.” This pretrained model was then applied to LinkedIn photos of more than 96,000 MBA graduates from top U.S. programs. The model extracted what the researchers call “Photo Big 5” personality traits based solely on facial features.
The Photo Big 5 personality traits extracted from the AI model include Openness (e.g., curiosity, aesthetic sensitivity, imagination), Conscientiousness (e.g., organization, productiveness, responsibility), Extraversion (e.g., sociability, assertiveness, energy level), Agreeableness (e.g., compassion, respectfulness, trust), and Neuroticism (e.g., anxiety, depression, emotional volatility).
Though the individuals in the MBA sample did not take personality assessments themselves, the accuracy of the model was validated on other large datasets. The resulting Photo Big 5 traits were then matched with career data (salary, job title, tenure, and school rank) to identify statistical relationships between personality traits and job outcomes.
These traits strongly predict career outcomes.
AI-derived personality scores were linked to a broad set of career outcomes, including salary, job seniority, mobility, and even the rank of the MBA program attended. For example, the predicted pay gap between individuals with the most and least “desirable” personality scores exceeded the compensation gaps by race in some cases.
AI-derived personality adds predictive value beyond test scores.
The Photo Big 5 had low correlation with cognitive measures like GPA and GMAT, meaning they offer distinct, incremental insight into likely career success.
Extraversion and conscientiousness drive career advancement.
Extraversion was the strongest predictor of higher salary and seniority. Conscientiousness predicted compensation growth, especially for men, and was associated with lower turnover. Further, the study shows that personality fit, as measured by position and industry, matters for turnover and advancement.
Personality matching is linked to higher wages.
Using a mapping between the functional description of a job position and personality, the researchers find that personality/work-style matching influences job selection and helps to explain wage variation within occupation categories. Those better matched along personality earn higher wages.
Ethical implications are real and urgent.
These findings suggest that personality-based insights can complement cognitive measures like GPA or GMAT in predicting career success. However, they also underscore how, in practice, firms could adopt appearance-based AI in high-stakes contexts such as hiring and promotions, potentially without individuals’ knowledge or consent.
While the technology is already feasible and may become more widespread, its use raises serious concerns about fairness, privacy, and discrimination. Since facial features are not easily changed, and plausibly correlate with how individuals are perceived by others, relying on them in hiring decisions could reinforce biases, the researchers note.
This article was partially generated by AI and edited (with additional writing) by Kyle Kearns and Knowledge at Wharton staff. Read our AI policy here.
