Software Design, & Imperfect Knowledge: Types, Models, & Challenges

Abstract

The imperfection of human knowledge is, on one hand, an inevitable fact in view of (i)the finite and discrete human mental faculties, and(ii)the infinite and continuous Nature all around us, the knowledge of which the human beings are required to have for survival.  On the other hand, the imperfection is a problem for scientists to understand properly the various natural phenomena, and for engineers to design and develop appropriate technologies. Also, for appropriate design of software, knowledge of various aspects of the relevant available knowledge of the problem domain and its environment, including the types of imperfection in the available knowledge is essential. Inappropriate design may lead to disastrous consequences, as has happened in the case of some critical applications.

        In this paper, first of all, we enumerate various significant types of possible imperfections in human knowledge, along with the types of problem domains in which each of these frequently occurs. Then discuss different mathematical/computational models developed to handle the imperfection of knowledge of respective problem domains. Knowledge of appropriate imperfection-model match is essential for designing robust software, and to avoid disasters in critical applications.

        The mathematical approaches/models/theories for handling these different types of imperfections in knowledge include probability theory, fuzzy set theory; fuzzy measures and its special cases belief measure, plausibility measure, probability measure, possibility and necessity measures, Modal logics, belief models, Rough Set Theory, Chaos theory, neural networks, genetic algorithms, and swarm intelligence etc.  Despite availability of these models, the solving of many types of problems is still challenging. All these issues are discussed in the paper.