Automating the managerial gaze: critical and genealogical notes on machine learning in personnel assessment

This paper takes a Foucauldian approach to current discussions on the use of machine learning in personnel selection, with a focus on pre-selection assessment. It problematizes two central themes in this debate: the objectivity and the accuracy of machine learning–supported personnel assessment. A Foucauldian perspective is employed to offer an alternative and critical approach to these themes and to allow for a reformulation of their underlying questions. From this perspective, the paper analyzes how ML-based personnel assessment tools reflect broader developments in governmental practices and technologies, conceptualized as elements of algorithmic governmentality. Drawing on the empirical example of HireVue, it critically examines HireVue’s operations. It also traces the historical development of 20th-century personnel testing and statistical procedures to show how contemporary ML-based personnel selection systems are embedded in longstanding practices of data-driven governance in the workplace. By situating the discourse on ML in personnel selection within the broader context of algorithmic governmentality and its prehistory, the paper highlights key implications for the study and practice of personnel assessment.