In epistemic (modal) propositional logic, we have to set a collection of possible worlds in addition to the standard assignments of the predicate symbols to "true" or "false". However, the syntax does not require a set of possible worlds beforehand. For example, I can write...
$$K(\neg p\lor q)$$
...without having any knowledge about the possible worlds. Why can't I decide whether this is "valid" or "satisfiable" as in classical propositional logic without having to appeal to possible worlds semantics? Is there a formulation of epistemic logic semantics that doesn't involve appealing to some set of possible worlds?