(Inter)faces of predictions

by Sheung Yiu

‘What is a Face?’ In the age of pervasive facial recognition algorithms and Tiktok filters, the question is anything but apolitical. The answer to this deceptively easy question is ultimately complicated and depends on the observer and the apparatus through which they observe. A child, with a crayon, sees a face as a solar head and draws a circle with two eyes, a nose and a mouth. A 19th-century French police officer, with a camera pointing towards suspects, might say facial features predict criminality. A writer, holding a picture of his deceased mother, might look deeply into the image, searching for her ‘truth’. A face reader, riffing off of literature of Chinese Physiognomy, might insist that the face is an indicator of personality traits and fortune. A Tiktok engineer working on face filters already decided people want their cheeks slimmer and their jawbones more pronounced. A facial recognition software, trained by machine learning, deconstructs the faces as pixel patterns of different shades and features. GANs researchers might answer that to the deep neural networks, a face, or rather the digital image of a face, is a high dimensional data point, the pattern used to generate other faces computationally. Countless answers all point towards a fact: a face is much more than what is physically there. Each answer is representative of a knowledge system. Especially when a face is seen through technology, whether analogue or digital, faces become an interface for predictions.
Contrary to the common wisdom, human tends to ‘judge a book by its cover.’ Across Eastern and Western cultures, society has developed ways to predict a person’s character through facial features. In East Asian cultures, the esoteric practice of face-reading promises the power to see into one’s future through facial analysis. Though face reading remains largely a folk belief, many continue to seek the occult power of predictions from face readers. In the West, the forgotten pseudo-science of physiognomy, combined with statistics and machine learning, re-enters our modern lives as facial recognition algorithms, where societal biases and individual prejudices continue perpetuating.
In the project, I put my face through various processes of creating predictions, namely Western physiognomy, face reading, face recognition, facial phenotyping, facial generation, and synthetic facial data. I mix the visual language of the occult in face reading with the ‘scientific’ in facial recognition in an attempt to blur the line between practices from the East and the West. The visual study reveals the similarities between two predictive regimes centered around the face: one remains folklore, while the other is extensively applied to almost every aspect of our daily life. The project challenges the automation bias of facial recognition (or what I call Western face reading). It unveils the deeper, often-unexamined meta-narratives underlying these practices.
A face is more than just a face. It is an interface of predictions trapped between two analytical frameworks and cosmological views that are not as different as they initially seem.