Review of “Machine Habitus: Toward a Sociology of Algorithms”

In Machine Habitus: Toward a Sociology of Algorithms (2022), Massimo Airoldi provides a very timely theoretical contribution to the study of “machine learning” algorithms, which are at the core of what is often referred to as artificial intelligence (AI). Airoldi presents these algorithmic systems in a way that is approachable for sociologists, particularly those familiar with a Bourdieusian set of theoretical tools. Although this helps demystify technologies that are often assumed to be so complex they can only be understood by technologists, the value of Airoldi’s argument ends up being limited in regards to algorithms more generally, as it depends on certain correspondences between the human and the machine. Recent years have seen growing concerns over harmful forms of algorithmic discrimination. Sociological contributions by scholars including Safiya Noble and Ruha Benjamin have drawn attention to the ways that algorithms and machine-learning-based classification and prediction systems frequently provide outputs that are racist, or reinforce other intersecting axes of oppression and inequality. Among the developers of such systems—data scientists or AI researchers—such problems are typically discussed through the language of “bias”. This vague and flexible concept fills the gap of social theory for AI practitioners, whose understanding is that society contains various “biases” which then end up finding their way into algorithmic systems. Airoldi provides a much more productive way of thinking about these problems, by relating them to sociological theories about the reproduction of culture, structure, and inequality, specifically through the concept of “habitus”.