Impreciseness and Its Value from the Perspective of Software Organizations and Learning

Abstract

When developing large software products many verbal and written interactions take place. In such interactions the use of abstract and uncertain expressions is considered advantageous. Traditionally, this is not the case for statements which are imprecise in the sense of being vague or subjective. In this paper we argue that such statements should not only be tolerated but, often, they can be very useful in interaction. For this purpose we relate abstraction, uncertainty and impreciseness to each other by investigating the differences and common properties. We also discuss the relation to the use of common sense implementations in Artificial Intelligence. The introduction of degrees of impreciseness leads to the question of finding an optimal level. This is interpreted as a learning problem for software organizations. The success can be measured in terms of different cost factors. The design of evaluation experiments is shown as an interdisciplinary task.