Big data, the collection, storage, and analysis of data at unprecedented scale, is the material foundation of contemporary AI. It is also a Foucauldian object: a system of power-knowledge that defines who is knowable, what counts as knowledge, and who has the authority to know.
The production of knowable subjects. Big data produces knowable subjects, people who are legible to data systems in ways that enable prediction, classification, and intervention. But this legibility is not equally distributed. The people most thoroughly tracked by data systems, the poor, the sick, the incarcerated, the migrant, are not the people with the most power. They are the people with the least. Their data is collected and used to make decisions about them, decisions over which they have little control and minimal recourse.
The rich are also tracked, their financial transactions, their social media activity, their digital footprints. But they are tracked in different contexts, for different purposes, by different institutions. The data collected on poor communities by welfare systems is used to surveil, control, and punish. The data collected on wealthy consumers by luxury brands is used to personalise and serve. The asymmetry is not incidental to big data. It is structural.
Normalisation. Foucault analysed how disciplinary institutions produce norms, standards of what counts as normal, healthy, competent, well-adjusted, and how deviations from those norms are identified, treated, and managed. Big data and AI extend this normalisation function massively. Credit scores, risk scores, behavioural profiles, health scores, all define what counts as normal and pathologise deviation. The definition of normal is not neutral. It reflects the data it was derived from, which reflects existing social hierarchies, historical patterns, and dominant cultural assumptions.
When AI systems trained on data from one population are applied to another, when models trained on Western medical data are applied to patients in the global South, when hiring algorithms trained on data from elite universities are applied to graduates of state universities, they impose one population’s normal on another. This is a form of epistemic colonialism, the extension of one group’s knowledge standards over another group, without acknowledgment or consent.
The political economy of data. Foucauldian analysis of power-knowledge connects with Marxist analysis of data as raw material. The knowledge produced by big data systems is not merely descriptive. It is productive, it shapes decisions, allocates resources, and produces subjects who understand themselves in terms of the categories that data systems create. The power exercised through data-knowledge is simultaneously economic and political, shaping markets, institutions, and individual life chances.
Reflection question: In what contexts is data about you collected? Who has access to it? What decisions does it influence? How does the power-knowledge relation Foucault describes apply to your own data situation?