Belief contraction through safe formulas

1 In this paper we consider approaches to belief revision and contraction in modal logics that are based on a translation of an agent’s beliefs in modal logic to first-order logic, where revision or contraction is performed before translating the result back into the original modal language. Contraction presents a unique problem, as theorems of the modal language can be accidentally removed, which would destroy the correspondence between a system of modal logic and its first-order simulation. We propose a solution for this problem based on defining safe-to-remove-parts of target formulas for contraction.