Theoretical Knowledge

In the AI Story

Bell distinguished theoretical knowledge from tacit knowledge and from practical know-how, and the distinction matters for understanding what AI does and does not automate. Theoretical knowledge — the proposition that water boils at 100°C at sea level, the algorithm for sorting a list, the statutory requirements for a valid contract — is precisely the kind of knowledge that large language models absorb from training corpora. Tacit knowledge — the surgeon's embodied sense of when tissue is behaving anomalously, the teacher's intuition about which student is about to disengage — is precisely the kind of knowledge that current AI systems cannot yet produce.

This distinction reframes the AI transition as a selective automation event rather than a total one. The knowledge that was most prestigious in the post-industrial economy — the theoretical, credentialed, codified knowledge of the professional class — is the knowledge most vulnerable to automation. The knowledge that was least prestigious — the embodied, situated, contextual knowledge of the craftsperson, the nurse, the mentor — is the knowledge that survives. This inversion of the prestige hierarchy is itself a structural feature of the fourth transformation, and it produces some of the discourse's most painful confusions.

The commodification of theoretical knowledge changes the economic calculus of the institutions that produced it. Universities organized their curricula, their faculty hierarchies, and their credentialing around the scarcity of theoretical knowledge. When the knowledge is freely available through a chat interface, the justification for tuition, degrees, and professional gatekeeping requires reconstruction. The universities are not obsolete, but their value proposition must be reconceived around what remains scarce: mentorship, community, the cultivation of judgment, the friction through which tacit knowledge develops.

The question for individuals navigating the transition is which components of their professional capability were theoretical and which were tacit. The former is being commodified; the latter is becoming more valuable. The displaced expert's experience of vertigo often arises because she assumed her expertise was more theoretical than it actually was — or, conversely, because she undervalued the tacit components of her work that now command the premium.

Origin

Bell developed the theoretical-knowledge concept in dialogue with the philosophy of science literature of the mid-twentieth century, particularly the work of Karl Popper and Thomas Kuhn on the structure of scientific reasoning. Bell argued that what distinguished post-industrial society was the institutionalization of theoretical knowledge production through universities, research institutes, and R&D departments — a claim that remained controversial among industrial sociologists who emphasized continuing material production.

Key Ideas

Codified, teachable, transferable. Theoretical knowledge has these three properties, and they make it both powerful and automatable.

The professional-technical class rose on this resource. Bell predicted, correctly, that control of theoretical knowledge would define the dominant occupational class of the late twentieth century.

AI automates precisely this kind of knowledge. Large language models excel at reproducing codified propositions and struggle with embodied, situated understanding.

The prestige hierarchy inverts. Knowledge that was prestigious because it was theoretical is being commodified; knowledge that was undervalued because it was tacit is becoming scarce.

Institutional consequences. Universities, credentialing bodies, and professional hierarchies built on the scarcity of theoretical knowledge must reconceive their value propositions.

Debates & Critiques

The sharp distinction between theoretical and tacit knowledge has been challenged by scholars including Harry Collins, who argue that even apparently theoretical knowledge carries tacit components without which it cannot be successfully applied. If Collins is right, the automation of theoretical knowledge by AI may be more limited than the surface appearance suggests — the machines can produce propositions but cannot reliably apply them to particular cases without human judgment supplying the tacit glue.