Context-Based Belief Revision

In credibility-limited (CL) belief revision, an agent may reject new information if it is considered not credible relative to its current beliefs. A core principle of CL revision requires that all consequences of a credible formula must themselves be credible. We propose a new framework, context-based (CB) belief revision, which generalizes CL revision by relaxing this requirement. In CB revision, a formula may be deemed credible because it strengthens one of its non-credible consequences by providing sufficient supporting context, a situation that CL revision does not allow. We introduce an axiomatic framework for CB revision operators, identify specific subclasses, provide representation theorems, and examine the relationships between CB revision operators, their subclasses, and CL revision operators.