Epistemic Accessibility Relations

Core Idea

An accessibility relation R relates possible worlds to possible worlds, with wRw' meaning that world w' is epistemically possible relative to w. The logical properties of R determine the modal axioms: reflexivity (wRw) gives K ⊢ p → p; transitivity (wRw' ∧ w'Rw'' → wRw'') gives positive introspection. Different epistemic situations (justified belief vs. knowledge) correspond to different relational properties.

Explainer

From your work with possible-worlds semantics, you understand that modal claims like "S knows that P" are analyzed by looking across sets of worlds. Roughly, S knows P in world w if P is true in all worlds that are epistemically relevant to S from w. But what determines which worlds count as "epistemically relevant"? This is exactly what an accessibility relation specifies. An accessibility relation R is a binary relation on a set of possible worlds: wRw' means "from world w, world w' is epistemically possible for the agent." A world is epistemically possible from w if, as far as the agent knows in w, that world might be the actual one. The set of worlds accessible from w is the agent's epistemic range — what they cannot rule out.

The power of this framework comes from the connection between the logical properties of R and the modal axioms that hold. Think of properties of relations you may know from logic or mathematics: reflexivity (every world accesses itself), transitivity (if w accesses w' and w' accesses w'', then w accesses w''), symmetry (if w accesses w', then w' accesses w), and Euclidean structure (if w accesses both w' and w'', then w' and w'' access each other). Each property corresponds to a different modal axiom about knowledge. Reflexivity gives the T axiom: Kp → p (if you know p, then p is true). This is the basic distinguishing mark of knowledge over mere belief — you cannot know something false. Reflexivity says the actual world is always in the agent's epistemic range, so whatever is known must hold in the actual world.

Transitivity corresponds to positive introspection (the 4 axiom): if you know p, then you know that you know p (Kp → KKp). If w accesses w' (w' is epistemically possible from w), and w' accesses w'' (w'' is epistemically possible from w'), then transitivity means w also accesses w'' directly — the agent in the actual world can "see through" to everything accessible from the worlds they consider possible. The Euclidean property gives negative introspection (the 5 axiom): if you don't know p, then you know you don't know it (¬Kp → K¬Kp). Together with T and transitivity, the Euclidean property produces the S5 system, which is the standard logic of knowledge in formal epistemology. S4 (T + transitivity) corresponds to a weaker notion that lacks negative introspection.

The philosophical significance is that different epistemic situations can be modeled by choosing which structural properties R satisfies. A skeptical hypothesis challenges reflexivity: the evil-demon scenario tries to insert a world that the actual world accesses — one where P is false — to undermine knowledge claims. Justified belief might correspond to a non-reflexive accessibility relation (the believed worlds need not include the actual world), which is why you can have justified belief in a falsehood but cannot know a falsehood. Formal epistemology uses this machinery to precisely distinguish and argue about these concepts. When philosophers debate whether knowledge requires positive or negative introspection, they are really arguing about whether the accessibility relation for knowledge is transitive or Euclidean — which is a precise, tractable formal question rather than a vague intuitive dispute.