A semantics for updating in the context of causal constraints
Nico Roos · Research Publications (Maastricht University) · 2001
An agent acquiring information about its environment, either by observation or by communication with other agents, must update its beliefs about the environment. In this way the agent keeps track of the changes in its environment. Usually the agent will not be able to perceive the environment completely. Since most environments do not change very fast over time, the agent may assume that everything about which the agent does not learn that it has changed, will still be the same. This minimal change assumption results in counter intuitive conclusions if an agent has background knowledge describing general (causal) relations between aspects of the environment. To avoid these counter intuitive conclusions, this paper proposes instead of the minimal change assumption, a minimal intervention assumption. Here, an intervention is an action of an agent that changes the environment. A Kripke semantics that formalizes the update process based on the minimal intervention assumption, will be specified.