Machine Learning Work: Computing Orders, or the Production of the Conceptual and the Concrete in Labour Processes, from Scientific Management to Platform Capitalism

How do machine learning algorithms change how capital poses the problem of managing labour? And to what extent does such an approach break with the principles of scientific management, principles which many scholars argue platforms extend and radicalise? In pursuit of these questions, this thesis theorises platform management and scientific management as orders. To order means to command, regulate, and control, but also to organise, to classify, to sequence, and to hierarchise. Building on Harry Braverman’s analysis of the labour process as composed of a process of conceptualisation and a process of enaction, I theorise orders as a relation between a conceptual logic and a concrete logic, wherein management seeks to impose its conceptual logic onto the concrete labour process. When management gives a command, the statement presupposes a means of identifying elements within the labour process alongside their relations and capacities. Simultaneously, the statement relies upon a concrete economic relation that subordinates labour to management and ‘authorises’ management’s attempt to direct and control the concrete unfolding of the labour process. Through analysis of the conceptual and concrete logics of scientific management and platform management, I trace each regimes relation to the wider epistemic, economic, technological, and historical conjecture from which it emerged. Ultimately, I argue that where scientific management sought to stabilise the labour process, machine learning algorithms allow platforms to productively exploit the instability of the labour process. Thus, I claim that scientific management and platform management express different ordering logics for managing the capitalist labour process. Through analysis of this logic, I provide a theoretical framework that accounts for the emergence and interaction of conceptual and concrete processes in the capitalist labour process, as well as the failures of management to reduce the concrete labour process to management’s conceptualisation of it, while also elucidating how machine learning technologies function as part of wider epistemic logic, as a set of concrete apparatuses and practices, and in relation to the historical unfolding of the capitalist mode of production.