The growth of facial recognition (FR) technology is accompanied by consistent assertions (as catalogued by Georgetown University) that demographic dependencies could lead to accuracy variations and potential bias. Film and video technology has a well-documented history of bias in its treatment of skin tone. While engineers have benefited from the increased availability of AI and machine learning tools, allowing them to train their models to ever higher accuracy, the fairness and ethics of their algorithms have often been an afterthought. Modern digital algorithms are similarly susceptible to asymmetric representation and treatment of face types. This paper draws from the National Institute of Standards and Technology (NIST) report which dissects these demographic dependencies of over 100 FR algorithms and then details the strategies and techniques needed, to not only reduce bias, but deliberately design for fairness and socially responsible outcomes. Furthermore, it shows that a model trained for low-bias actually delivers higher performance.
Takeaways: Workshops attendees will
Reza Rassool, Chief Technology Officer, RealNetworks