Despite the clamor of criticism, the use of face analysis software continues to grow. Facebook tags photos automatically, smartphones recognize their owners, and security cameras profile people at airports and stadiums, in supermarkets, and on the streets. All too often, however, the neural networks being used to interpret the visual data are myopic, operating with flawed data, and are sometimes outright racist.
The website and installation Facework transforms this debate into a game. Visitors are challenged to fool a neural network that uses machine learning, and which is used in real life, too. The game begins innocently enough with identification of uncontroversial characteristics, such as a smile or sunglasses. But it’s not long before the focus shifts to race, criminality, or sexual orientation. In this way, users experience for themselves how we are seen by the algorithms that are increasingly running the world.