Knowledge and Belief Operators

Core Idea

Formal operators K and B represent knowledge and belief; Kₐp reads 'a knows p' and Bₐp reads 'a believes p'. These operators have distinct axiomatizations: K satisfies S5 (including Kₐp → p and Kₐp → KₐKₐp), while B typically satisfies only S4 (no factivity). Combined with quantifiers and propositional operators, they enable formal representation of complex epistemic attitudes and their logical relationships.

Explainer

From your study of epistemic accessibility relations, you learned that modal logic models knowledge by asking which possible worlds an agent can "see" from a given world — the worlds she cannot rule out given what she knows. Knowledge and belief operators are the syntactic tools that bring this semantic idea into formal logic. The operator K (knowledge) and operator B (belief) work like modal operators □ and ◇: they attach to a proposition and an agent to produce a new, complex proposition. Kₐp says that agent a knows proposition p; Bₐp says that agent a believes p.

The crucial difference between K and B is factivity. Knowledge is factive: if you know p, then p is true. Formally: Kₐp → p. This axiom (called the T axiom) has no analogue for belief — you can believe something false. This is not a trivial technicality. It is why the systems used to axiomatize K and B diverge. The knowledge operator K satisfies S5, which includes not only the T axiom but also the positive introspection axiom (Kₐp → KₐKₐp: if you know p, you know that you know p) and the negative introspection axiom (¬Kₐp → Kₐ¬Kₐp: if you don't know p, you know that you don't know p). The belief operator B typically satisfies only S4, which includes positive introspection but drops factivity and usually drops negative introspection, since it is far less clear that beliefs about your own ignorance are always accessible to you.

Combining these operators with standard logical connectives lets you express the fine structure of epistemic situations with precision. "Agent a knows that agent b believes p but does not know it" becomes Kₐ(Bᵦp ∧ ¬Kᵦp). Such formulas appear in multi-agent epistemology — particularly in the analysis of common knowledge and coordination problems. A key application: two agents may each know p, each know the other knows p, and yet fail to have common knowledge (mutual knowledge iterated to all depths), which matters for explaining why coordination sometimes fails even when both parties are well-informed.

The reason to axiomatize these operators rather than describe them informally is systematicity: once you fix the axioms, you get logical consequences for free. If you accept S5 for K, then you cannot have a situation where Kₐp is true but p is false — such a model is simply excluded. This lets you reason about knowledge rigorously without analyzing every case from scratch. The tradeoff is idealization: real human agents are not logically omniscient, do not have perfect introspective access to their own beliefs, and can hold contradictory beliefs simultaneously. The formal operators model an idealized epistemic agent, and the gap between the ideal and the real is itself a productive source of questions in epistemology.