On the Semantics of Theory Change: Arbitration between Old and New Information