AI and the sociology of knowledge

We used to ask a simple question: Who knows?

Sociology has never treated knowledge as something that simply floats above society. Knowledge is produced, organized, authorized, and remembered within social arrangements. Karl Mannheim made this central to the sociology of knowledge: what people know is connected to where they stand in society.

When we ask ChatGPT to explain Philippine politics, summarize a book, or tell us what a concept means, we are not merely using a faster search engine. We are entering a relationship with a system that can organize fragments of a vast cultural archive and return them in a coherent voice.

Generative artificial intelligence (AI) does not create knowledge from nowhere. It is trained on human-produced data. Its answers are shaped by what was written, digitized, selected, translated, classified, and made available.

Consider the seemingly innocent prompt: “Explain Filipino culture.”

The answer may sound authoritative. But whose culture is being described? Is it Manila-centered? Middle class? English-speaking? Urban? Academic? Is the Philippines being described through categories Filipinos use for themselves, or through categories accumulated from outsiders writing about us?

The French philosopher Michel Foucault taught us to examine the relationship between knowledge and power. Institutions do not merely possess knowledge; they help determine what counts as legitimate knowledge. The French sociologist Pierre Bourdieu similarly reminded us that cultural authority is unequally distributed.

Former gatekeepers included universities, libraries, publishers, newspapers, governments, churches, and schools. New gatekeepers increasingly include technology companies that build systems through which people retrieve and organize information.

Inequality already shapes whose voices become part of the national archive. Much scholarship is published in English. Historical and cultural knowledge is unevenly digitized. Indigenous and local knowledge can be poorly represented in formal databases. Rural experiences are less visible online than urban experiences. Access to AI also depends on connectivity, devices, money, language, and digital literacy.

AI can therefore democratize knowledge while reproducing inequality at the same time.

A student can ask AI to explain a philosophical idea, draft an outline, translate an article, or clarify a statistical concept. A researcher can use it to navigate unfamiliar literature. A teacher can generate examples and materials in minutes.

But convenience can quietly become dependence.

If students stop asking, “How do we know this?” because the machine supplied an answer, education loses the habit of inquiry.

The problem is not merely that AI gets things wrong. It can also make uncertain knowledge appear settled. A fabricated citation can look scholarly. A distorted or partial account can sound comprehensive.

AI literacy cannot simply mean knowing how to write better prompts.

We need epistemic literacy: the capacity to ask where knowledge comes from, who produced it, what interests shaped it, what evidence supports it, whose voices are absent, and why one interpretation has become more authoritative than another.

That is precisely the kind of questioning sociology has always demanded. If AI changes how we know, sociological thinking becomes more necessary. We should ask the same questions of AI.

Who built it? Who trained it? What data did it learn from? Who owns the infrastructure? Who decides what counts as a reliable source or an acceptable interpretation?

The future of knowledge depends not only on AI becoming more accurate, but on whether societies retain the capacity to question the systems that mediate what they know.

We should welcome AI without surrendering our intellectual agency to it. The sociology of knowledge reminds us that reality is constructed through institutions, relationships, language, authority, and interaction. AI does not abolish this process. It adds a powerful new actor.

The question before us is not whether machines will know more than we do.

It is whether, in allowing machines to mediate what we know, we will remain curious enough to ask how they came to know what they appear to know.

It is about keeping alive the social capacity to really question them today.

The stakes are high when an answer arrives without the visible trail of people, institutions, and choices that produced it.

AI is not an oracle. It is a social artifact.

—————-

Prince Kennex R. Aldama is a sociologist in the Department of Social Sciences, University of the Philippines Los Baños. He was president of the Philippine Sociological Society.