Interval Uncertainty as the Basis for a General Description of Uncertainty: A Position Paper

Uncertainty is ubiquitous. Depending on what information we have, we get different types of uncertainty. For each type of uncertainty, techniques have been developed for efficient representation and processing of this uncertainty. However, the plethora of different uncertainty techniques is often confusing for practitioners. The situation is especially difficult in frequent situations when we need to gauge the uncertainty of the result of complex multi-stage data processing, and different data inputs are known with different types of uncertainty. To avoid this problem, it is necessary to develop and implement a general approach to representing and processing different types of uncertainty. In this paper, we argue that the most appropriate foundation for this general approach is interval uncertainty. Uncertainty is ubiquitous. All the data comes either from measurements or from expert estimates. Neither measurements nor expert estimates are absolutely accurate, so we always have to deal with uncertainty; see, e.g., [8]. The situation is especially critical for dynamical and/or spatial data: • For dynamical data, we not only have uncertainty about the corresponding values, we also have temporal uncertainty, i.e., uncertainty about the moment of time. • For spatial data, we not only have uncertainty about the corresponding values, we also have spatial uncertainty, i.e., uncertainty about the spatial location. Different types of uncertainty. Depending on what information we have, we have different types of uncertainty.