When I read Hollis Robbins' sharp essay this week, "The Rumsfeld Matrix," applying Donald Rumsfeld's famous taxonomy of knowns and unknowns to the crisis in higher education, it caught my attention. Her argument: Universities are over-invested in the "known-knowns" quadrant, spending hundreds of millions of dollars each year transferring settled knowledge into students' heads, while neglecting the quadrants where knowledge actually gets produced. With AI now capable of delivering known-known content cheaply, the case for restructuring is urgent.

It is a genuinely useful framework that expands on her proposals that I have been reading about for some time. My own views on neuroscience and the cognitive architecture of meaning-making, and Robbins' structural lens maps onto that work in ways I find productive. She is asking where institutions should invest. I keep asking what human minds do that machines cannot. The two questions need each other.

Robbins writes that universities "must decide which parts of known-known transfer need classroom time and which parts can be handled by cheaper, more flexible LLM systems." The framing treats AI-delivered instruction as a settled alternative whose effectiveness is established enough to anchor a resource reallocation strategy. That assumption is everywhere right now: in policy documents, in board presentations, in the pitches that edtech companies make to provosts. Robbins herself has published a detailed technical roadmap for delivering California State University's general education requirements through an AI-driven microservices architecture, complete with competency rubrics and automated assessment pipelines. AI works for learning. The meta-analyses say so.

A recent unpublished study calls this assumption into........

© Psychology Today