Higher-Order Knowledge and Iteration

Core Idea

Higher-order knowledge concerns whether an agent knows that she knows. In epistemic logic, positive introspection (KₐKₐp → Kₐp) and negative introspection (¬Kₐp → Kₐ¬Kₐp) are properties of the S5 axiom system. Transitive accessibility (wRw' ∧ w'Rw'' → wRw'') formalizes positive introspection: if you know something in all accessible worlds, you know that you know it. Not all epistemological positions accept full introspection.

Explainer

You have learned to work with the knowledge operator Kₐ: Kₐp means "agent a knows that p." In possible worlds semantics, Kₐp is true at world w just when p is true at all worlds accessible from w — all worlds the agent cannot distinguish from w given her evidence. This framework lets us ask not just what an agent knows, but what she knows *about her own knowledge*. These are called higher-order epistemic questions.

Positive introspection is the principle that if you know something, you know that you know it: Kₐp → KₐKₐp. In possible worlds terms, this holds whenever the accessibility relation is transitive: if world w can access world w', and w' can access w'', then w can access w''. Think through why this works. If Kₐp is true at w, then p holds at every world accessible from w. If accessibility is transitive, then from any world w' accessible from w, the worlds accessible from w' are also accessible from w — and p is true at all of them. So at w', the agent still knows p. Since this holds at every w' accessible from w, the agent at w knows that she knows p. Transitivity corresponds to the S4 axiom system in modal logic.

Negative introspection is the stronger principle that if you don't know something, you know that you don't know it: ¬Kₐp → Kₐ¬Kₐp. Adding this to S4 yields S5, the strongest standard epistemic logic, in which iterated knowledge operators collapse: KₐKₐp and Kₐp become equivalent, and you can always "see all the way down" your knowledge stack without loss of information. S5 is mathematically elegant but epistemically demanding. It requires that your ignorance is always transparent to you — that you can never fail to know p without also knowing that you fail. Real agents are frequently and unknowingly ignorant about the limits of their own knowledge; experts in one domain routinely overestimate their knowledge in adjacent ones. The debate over which introspection axioms to accept maps directly onto substantive philosophical questions about whether self-knowledge is privileged, whether introspection is reliable, and how we should model the epistemic situation of agents whose knowledge of their own knowledge is itself imperfect.