In the 2016 published paper „Automated Inference on Criminality using Face Images“
researchers Xi Zhang and Xiaolin Wu introduce their new dataset ChronoNet as foll-
ows:“The approach is to let a machine learning method explore the data and reveal
the most discriminating facial features that tell apart criminals and non-criminals.“
This approach deeply stems from essentialist eugenic theories developed during the
19th century. Francis Galton, a British natural scientist, tried to find the essence
of a criminal face using "pictorial statistics." He also applied his supposed findings
on the inheritance of traits to the human mind and introduced the concept of eugenics,
which he defined as a doctrine aiming to increase the proportion of positively valued
human hereditary traits through "good breeding.“
In a research article published in 2021 in SN Applied Sciences, Iranian researchers
present a system for facial expression recognition closely connected with human emotions.
They claim: "There are seven main types of human facial expressions, such as joy or happi-
ness, sadness, surprise, neutral, disgust, fear, and anger. These expressions represent
our internal emotions, which in turn manifest on our faces." There is a far-reaching
criticism regarding the universal validity of a taxonomy of facial expressions linked
to emotions. The claim to truth of this rebirth of pseudoscientific disciplines like
phrenology and physiognomy within the fields of computervision and machine learning harbors
great dangers.
Media theorists LukeStark and Evan Hudson describe this development using the term "Physio-
gnomic AI" in their 2021 paper, „Physiognomic Artificial Intelligence“, as "the practice of
using computer software and related systems to infer or create hierarchies of an individuals
body composition, protected class status, perceived character, capabilities, and
future social outcomes based on their physical or behavioral characteristics.”
In the artwork „Asset Flip“, the fictitious company F-ACQ offers its data to the global security
and military markets. As part of their advertising campaign, they use two light boxes as advert-
ising media. The product being advertised is a newly developed system for the taxonomy of human
emotions based on facial micro-expressions, which is purportedly able to reveal a person's true
self. In its efforts to map the human face using deep neural networks, F-ACQ echoes the remnants
of the 19th century in contemporary computer sciences.