Can AI quantify beauty? New study suggests it can’t
AI analysis of facial features finds demographic variation, not mathematical proportion, shapes perceptions of beauty
image: AI analysis of facial features finds demographic differences, not mathematical ratios, shape perceptions of beauty, highlighting bias in training data.
Credit: Margaux Jacks
Attempts to define human beauty using artificial intelligence may reveal more about bias in data than universal standards, according to a new analysis from the University of Virginia’s School of Data Science.
Using computer vision and statistical modeling, researchers evaluated whether facial features align with the “Golden Ratio,” a mathematical formula often cited as an objective measure of attractiveness. Instead, the analysis found that demographic variation, not mathematical proportion, was the strongest factor shaping model outputs. This challenges long-standing assumptions that beauty can be quantified.
The study analyzed facial images using regression, clustering, and dimensionality reduction techniques to compare real-world data against Golden Ratio benchmarks. Results showed significant variation across demographic groups, suggesting that widely accepted standards of beauty may reflect bias in underlying datasets rather than any universal ideal.
“This project highlights a core limitation of AI models, which are only as objective as the data they learn from,” said Prince Afriyie, an associate professor of data science. “Attempts to quantify beauty often reveal more about representation and bias than any universal standard.”
The findings underscore a broader challenge in artificial intelligence: systems trained on human-generated data can reproduce and even amplify social and cultural biases, particularly when applied to subjective domains. Efforts to standardize or automate judgments about human appearance, from social media filters to biometric tools, risk embedding those biases at scale.
Researchers emphasize the importance of diverse, representative datasets and critical evaluation of model assumptions, especially when AI is used to interpret complex human characteristics.
While mathematical patterns can describe aspects of the natural world, the study suggests that human perception is shaped by culture, context, and experience. As a result, beauty does not reduce to a single formula, underscoring the limits of AI in defining inherently subjective traits.