(p.316) Appendix 5 A Non‐Symmetric Epistemic Model
(p.316) Appendix 5 A Non‐Symmetric Epistemic Model
A creature stores information in sentential form. Its language L has two atomic sentences H (‘It is hot’) and C (‘It is cold’), the logical constants ∼ and ∧ with their usual interpretations, and the unary sentence functor K (‘I know that . . . ’). Let <A> be the proposition expressed by the sentence A on this interpretation, and W = {w _{1}, w _{2}, x}. In order to specify which sentences are stored in which worlds, recursively define an auxiliary function φ from L to W:

# It is disposed in y to store A if and only if y∈ φKA.
We can argue plausibly that φA is the set of worlds in which <A> is true. The argument is by induction on the complexity of A. The only non‐routine case is the induction step for K. The induction hypothesis is that φA is the set of worlds in which <A> is true. Since the clause for φKA implies that φKA φA, # implies that A is stored only in worlds in φA. Thus, by the induction hypothesis, the creature is disposed to store A only when <A> is true. We can therefore reasonably suppose that if the creature is disposed to store A then it knows <A>. Conversely, if it is not disposed to store A, then it does not know <A>. But <KA> is true if and only if it knows <A>. Thus <KA> is true if and only if the creature is disposed to store A. It follows by # that φKA is the set of worlds in which <KA> is true. This completes the induction step. ▪
Given that φA is the set of worlds in which <A> is true, the definition of φ recursively specifies the truth‐conditions of sentences of L. One can easily check that its results coincide with those of a semantics in possible worlds style, using (p.317) the accessibility relation in the diagram in section 10.6. Since the accessibility relation is reflexive and transitive, every theorem of the modal system S4 is true in every world, when □ is replaced by K and propositional variables by arbitrary sentences of L. Consequently, the creature knows every logical consequence of what it knows; moreover, whenever it knows p, it knows that it knows p. But sometimes, when it does not know p, it does not know that it does not know p. That is because it cannot survey the totality of its knowledge. It is a failure of self‐knowledge, not of rationality in any ordinary sense.