Operative Contexts: Belief Revision and Memory in Agentic AI

Persistent memory is becoming central to agentic AI systems, yet it is often evaluated as retrieval: can the system recall past information? We argue that this framing misses a deeper interpretive problem: agents with long-term memory do not merely store user information, but construct an operative context through which future requests are understood and acted upon. Drawing on common ground, cognitive control, and belief revision, we distinguish stored memory from operative context: the information that actually guides an agent’s behavior. We define silent contextual misalignment as a failure in which outdated, uncertain, wrongly scoped, or contextually inappropriate information guides action while remaining hidden from the user. We propose desiderata and diagnostic stress tests for evaluating context-sensitive memory control, and argue for contextual inspectability as a mechanism for collaborative repair. More broadly, we frame persistent memory as an interpretive infrastructure that should remain contestable, revisable, and sensitive to the social and situated contexts in which users interact with AI systems.