Approaching algorithmic power

Contemporary power manifests in the algorithmic. Emerging quite recently as an object of study within media and communications, cultural research, gender and race studies, and urban geography, the algorithm often seems ungraspable. Framed as code, it becomes proprietary property, black-boxed and inaccessible. Framed as a totality, its becomes overwhelmingly complex, incomprehensible in its operations. Framed as a procedure, it becomes a technique to be optimised, bracketing out the political. In struggling to adequately grasp the algorithmic as an object of study, to unravel its mechanisms and materialities, these framings offer limited insight into how algorithmic power is initiated and maintained. This thesis instead argues for an alternative approach: firstly, that the algorithmic is coordinated by a coherent internal logic, a knowledge-structure that understands the world in particular ways; second, that the algorithmic is enacted through control, a material and therefore observable performance which purposively influences people and things towards a predetermined outcome; and third, that this complex totality of architectures and operations can be productively analysed as strategic sociotechnical clusters of machines. This method of inquiry is developed with and tested against four contemporary examples: Uber, Airbnb, Amazon Alexa, and Palantir Gotham. Highly profitable, widely adopted and globally operational, they exemplify the algorithmic shift from whiteboard to world. But if the world is productive, it is also precarious, consisting of frictional spaces and antagonistic subjects. Force cannot be assumed as unilinear, but is incessantly negotiated—operations of parsing data and processing tasks forming broader operations that strive to establish subjectivities and shape relations. These negotiations can fail, destabilised by inadequate logics and weak control. A more generic understanding of logic and control enables a historiography of the algorithmic. The ability to index information, to structure the flow of labor, to exert force over subjects and spaces— these did not emerge with the microchip and the mainframe, but are part of a longer lineage of calculation. Two moments from this lineage are examined: house-numbering in the Habsburg Empire and punch-card machines in the Third Reich. Rather than revolutionary, this genealogy suggests an evolutionary process, albeit uneven, linking the computation of past and present. The thesis makes a methodological contribution to the nascent field of algorithmic studies. But more importantly, it renders algorithmic power more intelligible as a material force. Structured and implemented in particular ways, the design of logic and control construct different versions, or modalities, of algorithmic power. This power is political, it calibrates subjectivities towards certain ends, it prioritises space in specific ways, and it privileges particular practices whilst suppressing others. In apprehending operational logics, the practice of method thus foregrounds the sociopolitical dimensions of algorithmic power. As the algorithmic increasingly infiltrates into and governs the everyday, the ability to understand, critique, and intervene in this new field of power becomes more urgent