Can a computer really decide what is beautiful? A new study suggests the answer is not as simple as many people think.
Researchers from the University of Virginia School of Data Science have found that attempts to use artificial intelligence to measure human beauty may reveal more about bias in data than any true universal standard.
For years, some people have believed that beauty can be explained using the Golden Ratio, a mathematical formula that appears in nature and art.
It has often been used to suggest that certain facial proportions are more attractive than others. In this study, scientists tested whether this idea holds up when analyzed using modern AI tools.
The research team used computer vision and statistical methods to examine a large set of facial images.
They compared real human faces with the proportions suggested by the Golden Ratio. At the same time, they applied techniques such as regression analysis and clustering to look for patterns in how the AI evaluated attractiveness.
The results were clear. Instead of finding a universal rule for beauty, the researchers discovered that differences between demographic groups played a much larger role.
In other words, the AI’s judgments were strongly influenced by the types of faces it had been trained on, rather than by any mathematical ideal.
This means that what the AI identified as “beautiful” depended heavily on the data it learned from. If certain groups were overrepresented or underrepresented in the dataset, the model’s output reflected those imbalances.
The findings suggest that widely accepted ideas about beauty may not be as objective as they seem, but instead shaped by cultural and social influences.
According to the researchers, this highlights an important limitation of artificial intelligence. AI systems are often seen as neutral or unbiased, but in reality, they learn directly from human-created data. If that data contains bias, the AI can repeat or even strengthen those biases.
This issue becomes more important as AI is used in everyday tools. For example, social media filters, facial recognition systems, and other technologies that analyze human appearance could unintentionally promote narrow or unfair standards. When these systems are used widely, they can influence how people see themselves and others.
The study also reminds us that beauty is a complex and personal experience. While mathematics can describe patterns in nature, human perception is shaped by culture, background, and individual taste. What one person finds beautiful may be very different from another’s view.
In the end, the research suggests that beauty cannot be reduced to a single formula or number. Instead of relying on AI to define it, we may need to better understand the diversity of human perspectives.
