gci

Artificial Racial Bias and Beauty

Contact Author
Close
Fill out my online form.

If an algorithm receives an imbalanced data set to learn from, can it produce unbiased results?

Youth Laboratories develops digital tools, specifically algorithms, to study aging and discover effective anti-aging interventions using advances in machine vision and artificial intelligence. But can the technology be trusted to be unbiased? 

Youth Laboratories' Beauty.AI recently unleashed a suite of algorithms upon 6,000 photo applications from more than 100 countries and judged the relative beauty of the human faces based on wrinkliness relative to actual age, pigmentation and pimples, similarities to their own racial group, facial symmetry, and the gap between chronological and perceived age.

While the technology worked, its results favored overwhelmingly Caucasian entrants, with a much smaller portion allotted to Asian entrants. Few, if any other races were selected by the algorithms as having particular beauty.

According to one recent report, the results may have been skewed by the fact that the algorithms "learned" facial patterns from an overwhelmingly Caucasian data set, meaning darker complexions may have been flagged as less than ideal.

It just goes to show, the technology can only be as fair as the data it's based upon.

 

Related Content