The Rumsfeld Matrix

Higher Ed is at war, after all

How should universities think about AI as a problem of knowledge and awareness during battles over budgets, legitimacy, and the authority to define, steward, and validate knowledge? AI is shifting the balance of power away from higher education and this is a problem for future knowledge production.

In a recent conversation with Jeff Giesea about accelerating scientific discovery I brought up the Rumsfeld Matrix—Donald Rumsfeld’s famous taxonomy of known knowns, known unknowns, unknown unknowns, and unknown knowns from a Pentagon briefing in February 2002. Rumsfeld was describing the problem of acting under conditions where the cost of ignorance is catastrophic.

Reports that say that something hasn’t happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns — the ones we don’t know we don’t know. And if one looks throughout the history of our country and other free countries, it is the latter category that tend to be the difficult ones.

And so people who have the omniscience that they can say with high certainty that something has not happened or is not being tried…can do things that I can’t do.

Higher education faces a version of that problem. No president has the omniscience to say with certainty what isn’t coming for the sector, from enrollment cliffs to AI competition. I’ve found this matrix incredibly helpful for thinking about what universities could become if they stopped organizing themselves mainly around known knowns. The better goal is accelerating students toward knowledge production.

The Rumsfeld Matrix for higher education

In the Rumsfeld Matrix for higher education, the “knower” is the institution and its subunits: the library, research centers, departments, and its own archives. Each unit has a sense of what is settled knowledge, where the named gaps are, and where the invisible gaps are. The matrix makes clear the university’s over-investment in simple knowledge transfer; what it knows it should do to produce more knowledge; how it should be a better steward of the knowledge it has; and why it isn’t producing more knowledge. Ideally students would be working in all four quadrants.

Known knowns

Most states mandate “general education,” which, from the standpoint of curriculum designers, is a bundle of known knowns: canonical topics and skills that lawmakers think matter for thoughtful and productive citizens. Putting aside the skills question for a moment, most of the public debates are over what should be in the bundle. (Lately it’s more pro-American civics and less painful American history.) The squabbles are never over whether the “known knowns” quadrant is the right place to invest. So hundreds of millions of dollars are spent each year moving a fixed body of knowledge—whether Plato or Sally Hemings or the periodic table—into student heads.

I want to emphasize that “known knowns” refers to what the institution knows and considers settled. But conceptualize a place inside the “known knowns” quadrant called “unknown to students.” Keeping that relationship in view matters in the AI era, because universities must decide which parts of known-known transfer need classroom time and which parts can be handled by cheaper, more flexible LLM systems. (I’ve written about this here, here, and most recently here.)

The arguments about the social utility of doing known-knowns transfer in a classroom will go on for a long time, I expect. The skills of judgment and discussion matter in digesting information about Plato or Sally Hemings or basic chemistry. Character development might happen in these classrooms, with faculty members charged with shaping it. Plus there is a real case to be made that teaching keeps faculty current in their domain knowledge and its standards, and students need concepts, vocabulary, methods, and examples before they can see what is missing, ask better questions, or interpret what they find.

The known-known quadrant is also necessary to prepare students for knowledge production in the other three quadrants. But are universities too focused on this part of their task when LLMs could do knowledge transfer more economically?

If universities are to justify their cost in the AI era, the strongest case will come from the quadrants with “unknown” in them. The critical job in the known-known quadrant should be setting standards for future competence in addressing unknowns. The matrix helps clarify where the work is actually happening, where it is merely claimed, and where it is blocked by the way resources remain tied up in known-known delivery and auditing.

Unknown knowns

Unknown knowns, for a university, are largely internal. They are the university’s buried assets and underused capabilities.¹ Knowledge may be unknown because it was never catalogued or exists in poor archives. Every university has datasets no one is currently maintaining. Departmental expertise and institutional memory are often invisible outside a narrow circle. Unspoken rules guide dissemination. A great deal of knowledge in the humanities (particularly history) is “known somewhere” yet unavailable in practice because it is gated, fragmented, untranslated, or lost in obscure venues. The library and archive function sits here, though the quadrant is larger than the library. “Unknown knowns” are knowledge the university could surface, connect, and make usable but regularly fails to.

Here’s where investing in LLM-enabled systems will have a payoff, especially when paired with strong indexes, metadata, and access controls. These tools can support search, translation, summarization, and cross-literature pattern finding, which lowers the friction of discovery. Students can spend less time in the “known known” quadrant and more time creating and maintaining reliable indexes, preserving primary sources, establishing provenance, documenting context, and enforcing norms around citation and traceability. Used this way, AI helps the university become more aware of what it already has and more able to make that knowledge accessible and defensible, while graduating students who have practiced knowledge production.

Known unknowns

Known unknowns are recognized knowledge gaps: causes of a terrible disease, the best design for a stable fusion reactor, the mechanisms linking poverty to educational outcomes. Universities have long bet their prestige on this quadrant. Money is already being spent here. AI can speed up hypothesis generation, literature review, experimental design, but the back end of running the experiment, collecting the field data, and validating the result can be the work of faculty and students, producing graduates who better understand knowledge production.

Unknown unknowns

Unknown unknowns are questions no one has yet thought to ask, including blind spots created by the current question-generation process. Universities often claim a special relationship to this quadrant, yet few systematize it in a way that includes students. Imagine investing in structures that put students in settings where anomalies are likely and then requiring them to record what happened, protect it from being dismissed as noise, and maintain channels that translate 'something strange happened' into a future research agenda. In this quadrant, students can participate in knowledge production by helping create conditions where surprise appears and by maintaining processes that keep surprise available for investigation.

What does Rumsfeld Matrix thinking enable?

A university focused on and invested in the known-known quadrant is doing the bare minimum for its students, deciding what students need to know before they can work in the other three quadrants. The reality is, most students never get out of that quadrant. They consume knowledge. They never produce it. And now that AI can deliver known knowns for cheap, the task is deciding whether the bare minimum is sufficient to survive.

A university that invests in unknown knowns is more efficient at bringing to light knowledge it already has, building new infrastructure to find, trace, connect, and defend knowledge buried in its own archives, scattered across its own departments, locked in formats no one is stewarding. AI can accelerate that work. Teaching undergraduates to help with this work is vastly better than keeping them in the known-known quadrant.

A university that invests in known unknowns believes that curing a rare disease or building a new technology will pay off in the long run. It invests in cutting-edge faculty and makes time for research that may not pay off for years. In the meantime, students are working in this quadrant, after learning known knowns via AI.

A university that invests in unknown unknowns is a university focused on the future, giving top researchers and top students time and funding to ask new questions. The task is problem formulation, documentation of surprise, and forging paths from anomaly to question. No industry will fund this. No government agency will structure it. Yet the institution organized to encounter the unexpected is more likely to survive than organized entirely around known-known delivery.

The problem is resources spent on the known-known quadrant, demanded by state legislatures. Could there be better incentives to support the “unknown” quadrants? The matrix gives a president language for seeing where resources are going and where they aren’t, and in the AI era, the institutions that survive will be the ones that shifted investment toward the quadrants focused on knowledge production.

In this war, the decisive advantages are speed in turning uncertainty into verified knowledge, credibility in what you claim, and discipline in how you allocate scarce money and attention.

1. Slavoj Žižek famously called Rumsfeld’s forgetting of this category ideology: “What he forgot to add was the crucial fourth term: the ‘unknown knowns,’ the things we don’t know that we know-which is precisely, the Freudian unconscious, the ‘knowledge which doesn’t know itself,’ as Lacan used to say.”