What Is the History of Ideas?

This short essay examines Michel Foucault’s critique of the history of ideas, which we will pursue further in our online course on Foucault in the age of AI in January 2027.

What do we mean when we say that an idea has a history? We can trace how the idea of “freedom” changes from one philosopher to another, how “progress” becomes a political ideal, or how a medical theory enters everyday language. The history of ideas examines these movements through books, institutions and beliefs, exploring how concepts emerge, circulate and acquire meanings their earlier users might not even recognise.

Its materials extend beyond celebrated thinkers, through newspapers, medical manuals, religious instruction and anonymous publications. These can reveal ideas that never formed a coherent theory but nevertheless shaped how people understood their lives. A history of ideas can follow the exchanges between these materials and the works that eventually became famous.

Michel Foucault challenges the underlying assumptions used to organise these histories. Why should a single author’s body of work form the default unit of investigation? What justifies our habit of tracing a single theme across centuries? And what do we overlook when we treat past beliefs simply as incomplete stepping stones toward our current state of knowledge?

Consider a history of ideas about madness. We could follow successive explanations of its causes, comparing treatments and identifying discoveries. But the phrase “ideas about madness” can already suggest that a stable object lies beneath these changing accounts. Foucault asks how madness became available to particular forms of knowledge: which behaviours could be recorded as symptoms, how someone’s words became evidence of their condition, and who acquired the authority to interpret them.

In place of a linear history of ideas, Foucault introduces a vast, shifting archive: the underlying rules that dictate how objects are categorized, how concepts become usable, and how statements gain authority. These rules have their own histories, and when they shift, society begins investigating entirely different objects, recognising new forms of evidence, and asking questions that previously made no sense.

Foucault calls this shift problematisation - the process by which familiar, unquestioned practices suddenly become matters of intense concern and dispute. The focus is not just on the solution, but on how a previously accepted aspect of life becomes a problem that demands an explanation, and how framing it that way dictates the kinds of solutions people can propose.

Consider our idea of “human intelligence”. To measure it, we need to decide how memory, speed, verbal fluency and problem-solving count towards an assessment. These choices bring different activities together, assign them relative importance and make people comparable. We also draw conclusions about the person performing the tasks: a convincing explanation is taken as evidence of understanding, and a successful solution as evidence of reasoning. Through these procedures, intelligence takes a particular form as an object of knowledge.

Those same tasks can become benchmarks that guide the development of an AI system. When a machine succeeds, it may unsettle the connection between the performance and the capacity we thought it revealed. Does a convincing explanation justify attributing understanding to the system? Or does its success give us reason to reconsider what such an explanation demonstrates, including when a person produces it? A procedure we used to judge intelligence has itself become something that requires examination and justification.

This is how AI can prompt a problematisation of intelligence: we might respond by recognising understanding in the machine, changing our tests or distinguishing abilities we previously treated as one. These responses arise from a shared difficulty: we can no longer take for granted what a successful performance allows us to conclude about whoever, or whatever, produces it. The question of whether machines are intelligent therefore leads back to the practices through which we recognise thought and understanding in human beings, and to the possibility that those practices may have to change.

In our eight-week online course, Michel Foucault: A New Archivist in the Age of AI, we will read The Archaeology of Knowledge closely to investigate these relationships. Through Foucault’s early work on madness, medicine and the human sciences, we will explore how ways of knowing take shape, and how their transformation changes the place human beings occupy within knowledge.