Algorithmic Sovereignty: Machine Learning, Ground Truth, and the State of Exception

This thesis examines the influence of algorithmic security technology on political theories of sovereignty and the state of exception. In computer science, an exception is a special circumstance that prevents a program from terminating normally. In political theory, the state of exception, as theorized by Carl Schmitt and Giorgio Agamben, cannot be encoded within the normal liberal legal order. With states’ increasing turn toward algorithmic answers to questions of ‘security,’ which, per Agamben, is the long-term form of the political state of exception, the relationship between the concept of the exception in these two disciplines comes to the fore. In this thesis, I argue that the specifically algorithmic nature of these techniques of security have fostered a proliferation of two aspects of sovereignty and the political state of exception: the obfuscation and diffusion of the sovereign decision, and a contradictory relationship with time captured by what I call the fractalization of the present. Moreover, I engage with Theodor Adorno’s negative dialectics to interrogate the inherent and contradictory claim to truth that machine learning algorithms make through their model of ‘ground truth,’ which describes the image of reality that a machine learning model uses to make decisions. I conclude that most machine learning technology, and all algorithmic security technology, makes a claim of identity between itself and bourgeois reality – and thus inherently reinforces and reproduces the relations of domination entailed in that image of the world. However, space still exists for machine learning and its ground truth to instead operate within spaces of political non-identity, or exceptions to the bourgeois totality, and therefore play a role in revolutionary politics today