A Control Structure for Time Dependent Reasoning

by afternoon having changed to another state. In many domains it is important to reason about processes that change over time. Unfortunately, in many situations relevant data may not be available when the reasoning is done; there may be corrections to the data; or the state of the process may be changing. These problems are particularly evident in the reasoning involved in patient management. Physicians tend to reason about the patient state as a series of states. Therefore, a system attempting to capture their expertise must be able to find the appropriate time intervals in which to do the reasoning, handle incomplete data and handle changes to data even when the data relates to a state that has since changed. We present a data-driven control structure for reasoning processes in a domain in which updating or changes can occur. This mechanism implements two abstractions for these processes: the abstraction of data from a continuous process and the abstraction of decision making in a static state context. We will illustrate the use of the system with an example from a medical expert system for patient assessment, but the techniques are also applicable to other domains such as business decision making with result tracking and sensor interpretation.