Ignorance Analysis for Discovery and Knowledge Construction.

摘要

Engineers and scientists are increasingly required to design, test, and validate new complex systems in simulation environments with limited experimental results due to international and/or budgetary restrictions. Examples include space missions, certifications of missile stockpiles, and economic forecasting. These new trends require new practices of rigorous analyses of knowledge, information, uncertainty, and ignorance. This paper deals with knowledge construction by emphasizing both available information and ignorance. Knowledge can be constructed based on ignorance analysis. Ignorance analysis for knowledge construction ensures that our models and simulations do not assume and utilize implicitly or blindly more information than what is available, and accounts for confusion and conflict in available information. Using the concepts and definitions from evolutionary knowledge and epistemology, ignorance is examined and classified in this report. Two ignorance states for a knowledge agent are identified: (1) non-reflective, or blind, state (i.e., the person does not know of self-ignorance, a case of ignorance of ignorance); and (2) reflective state, (i.e., the person knows and recognizes self-ignorance). The paper also examines an algebraic problem set as identified by Sandia National Laboratories to be a basic building block for uncertainty propagation in computational mechanics. Solution algorithms are provided for the problem set for various assumptions about the state of knowledge about its parameters. (3 tables, 9 figures, 28 refs.).