I will be at the Western Political Science Association Conference in San Diego, 18-20 April 2019. Here’s the abstract for my talk:
Machine Learning and the Algorithmic Interpellation of the Future
In this paper I explore the impact of machine learning algorithms on the data they process. I make a two-part argument. First, that the inductive basis of machine learning functionality forecloses their capacity to produce outcomes with no ancestral relation to their training data. This has a protracted winnowing effect on content, which is a political concern due to the growing presence of machine learning algorithms in various facets of communal and individual life. I maintain that the a priori restrictions placed by algorithms on data constitute an emerging hegemonic order. Second, that an intervention in this scenario can be staged through an examination of non-digital, interpretative and self-reflexive methods in empirical science. The restrictions of machine learning, I offer, is drawn into high relief by exploring scientific studies in which it is deemed methodologically insufficient. This reading indicates predicates of intelligence which allegedly “intelligent” automation fails to self-generate.
I substantiate the first part of my argument with work from three philosophers of digital media. Antoinette Rouvroy writes on “algorithmic governance,” the automated retraction of possibility from probability in digital content. Matteo Pasquinelli argues that machine learning-based systems, including A.I., cannot be intelligent, as their inductive functionality does not compass the creativity which constitutes genuine intelligence. Adrian Mackenzie applies a Foucauldian reading to algorithmic determinacy. These thinkers provide me with materials for the second part, in which I use specific cases to theorize non-digital methodology in empirical science as a challenge to the algorithmic cancellation of the future.
In other news, my review of Andrew Feenberg’s Technosystem: The Social Life Of Reason has been published in the Humanities and Technology Review .