Can artificial intelligence (AI) algorithms – the wave of the future – be trusted or do they show bias?They are being introduced into almost every aspect of our lives including the automatic estimation of age from a person’s face, a technology that could be used in the future to determine who can enter a bar or other venues where age is a factor, as well as in a range of other applications.But what biases are there in AI processing? Researchers from Ben-Gurion University of the Negev (BGU) in Beersheba and Western University in Ontario, Canada, tested a large sample of the major AI technologies available today and found that not only did they reproduce human biases in facial age recognition, but they also exaggerated those biases. Their findings were published recently in Scientific Reports under the title “Biases in human perception of facial age are present and more exaggerated in current AI technology.”“Our estimates of a person’s age from their facial appearance suffer from several well-known biases and inaccuracies such as overestimating the age of smiling faces compared to those with a neutral expression,” said Prof. Tzvi Ganel from the psychology department, who conducted the research with Prof. Carmel Sofer from BGU’s department of brain and cognitive sciences in collaboration with Prof. Melvyn Goodale from the Western Institute for Neuroscience at Western University.
Age bias in AI
The growing interest in age estimation using AI technology raises the question of how it compares to human performance and whether it suffers from the same biases. “Here, we compared human performance with the performance of a large sample of the most prominent AI technology available today. The results showed that AI is even less accurate and more biased than human observers when judging a person’s age – even though the overall pattern of errors and biases is similar,” they wrote.
“Thus, AI overestimated the age of smiling faces even more than human observers did. In addition, AI showed a sharper decrease in accuracy for faces of older adults compared to faces of younger age groups, for smiling compared to neutral faces, and for females compared to male faces. These results suggest that our estimates of age from faces are largely driven by particular visual cues rather than high-level preconceptions.”In addition, the pattern of errors and biases they observed could provide some insights into the design of more effective AI technology for age estimation from faces, the researchers wrote.The data about AI performance was collected from 2020-2022, using a representative set of 21 current commercial and non-commercial AI age estimation platforms. AI performance was compared with the performance of 30 undergraduate students at BGU. “The AIs tended to incorrectly estimate the age of young people by as much as two-and-a-half years,” Ganel said. “Interestingly, whereas in human observers, the aging effect of smiling doesn’t affect their estimate of middle-aged adult female faces, it was present in the AI systems.”At this stage, the researchers can only speculate about why these biases occur – perhaps because of the photo sets used to train the AIs or perhaps because of a statistical phenomenon called regression to the mean – which results in an overestimation of the ages of young people and an underestimation of the ages of older adults.