What Can Foucault Tell Us About AI

This short essay examines the connections between Foucault’s thought and AI, questions we will pursue further in our online course on Foucault’s The Archaeology of Knowledge in January 2027.

It seems that the world Michel Foucault described throughout his work is still unmistakably ours. We speak of power relations, discourse and surveillance - often without realising how much this vocabulary owes to him. These terms have traveled far beyond philosophy because the forces they describe permeate our political and everyday lives. Even those who have never read a word of Foucault find themselves speaking his language. His work also raises a question that artificial intelligence brings to the fore: what happens when shifts in the organisation of knowledge alter the place of human beings within it?

We ask whether AI’s answers are correct. But what happens when those answers are a professional opinion, an assessment of a job applicant, or the basis for a lending decision? Who checks them, what gives them authority, and how can they be challenged?

Foucault offers a way to understand this transition: how what is said becomes knowledge that people act on. The same sentence can appear in a private conversation, a chatbot response or a medical report, acquiring a different status each time. To understand the difference, we need to examine who is authorised to make a diagnosis, what tests support it, where it is recorded and what decisions follow from it.

This is the central concern of The Archaeology of Knowledge. Foucault examines diagnoses, reports, classifications and records to understand how our ways of knowing the world take shape. How does something come to be recognised as a diagnosable illness? How does a human characteristic become something measurable? And how does someone acquire the standing of an expert authorised to interpret the data?

To classify someone as a “high-risk borrower”, for example, we need to choose which information about them matters, how to measure it and how to connect it to their likelihood of repayment. We need to understand how it was constructed: which aspects of their life became relevant, and which circumstances never entered the calculation.

AI changes every stage of this process: the data selected, the connections identified and the explanations given for the result. We can also imagine a situation in which bankers or lawyers use recommendations from a complex system and check some of its results, yet struggle to explain how it reached them. What would expertise mean then, and who would be able to scrutinise the decision?

Foucault’s approach allows us to ask what exactly changes when we use AI: are we answering familiar questions with a new tool, or is the way we define job suitability, risk or sufficient evidence also changing? To understand this change, we need to examine the assumptions on which the answers are based. A system can generate countless answers that all rest on the same assumptions. More answers do not, in themselves, guarantee that we will question whether those assumptions are justified, offer alternative explanations or recognise which aspects of people’s lives are being left out.

These changes also affect our own position: as people seeking to understand, as experts making decisions and as individuals whom others assess and classify. Foucault helps us investigate how such arrangements came into being, recognise when they are changing and consider how we might intervene.

In our eight-week online course, Michel Foucault’s The Archaeology of Knowledge: A New Archivist, we will engage in a contemporary reading of the book. Through examples from medicine, education and the digital world, we will learn to use his approach to understand how knowledge acquires authority and how it shapes the possibilities open to us.