Reanimation of pseudoscience in machine learning

Aug. 4, 2024

Andrews, Smart, & Birham (2024) delve into how machine learning has a pseudosciende problem. They take a deep view on the resurgence of deep learning-assisted physiognomic research. Physiognomy is the facility to identify from the form or constitution of external parts of the human body, signs of behavioral orientation, clinical conditions, interior states, or life outcomes. Cases of government and industry actors that have adopted such methods into technologies deployed on the public are cited, mentioning technologies like Faception, Hirevue, and Turnitin. Authors claim that:

"The use of ML or its products in the designing of real-world interventions or the architecting of public-facing technologies should, in general, be understood as cementing causal interpretations of the outputs of such models. Acting on model outputs is de facto causal interpretation. Misinterpretation of the causal mechanisms responsible for observed data in socially sensitive contexts should be expected to harm classified populations."

References:

• Andrews, Smart, & Birham (2024). The reanimation of pseudoscience in machine learning and its ethical repercussions. Patterns. https://doi.org/10.1016/j.patter.2024.101027