During the time the agent is carrying out actions in support of event handling, the world does not stand still. Information continues to stream into the agent from outside, the world continues to change, and others continue their tasks. The agent cannot, then, suspend event handling while waiting for actions to complete; new events will demand attention, and should be handled.
Two possibilities suggest themselves. One solution is to make the state of event handling explicit, so that the agent can interrupt the handling of one event, then later resume where it left off after the necessary actions have been taken. A problem with this is that until the actions complete, the agent has actually done nothing about the event, which could be catastrophic: consider an event that suggests an AUV may be drifting below its crush depth. The second solution is to allow event handling to finish without waiting for actions, communication, and/or negotiation to finish-but to make only a provisional diagnosis, assessment, and action recommendation. Later, when additional information arrives as a result of the actions or communication, the agent would reevaluate the event in light of the new information. We will investigate the latter solution as this work progresses.