Strategic planners, engineering designers and decision-makers in general often face severe lack of information. In this talk we discuss a non-probabilistic paradigm of uncertainty which quantifies the information gap between what is known, and what needs to be known for an optimal decision. These "info-gap" models of uncertainty underlie a decision procedure which is particularly suited to extremely unstructured and deficient information.
We begin with a brief discussion of the axiomatic distinction between probability and the info-gap perception of uncertainty. We then formulate the basic decision functions of info-gap decision theory Robustness function protection against adverse uncertainty; Opportunity function exploitation of propitious uncertainty. We consider simple examples from engineering design, economics, and management.
This talk develops a theory of reliability which is based on info-gap models of uncertainty, and is entirely devoid of probabilistic information. This theory is particularly suited to structural engineering design and reliability analysis under severe lack of information about loads, operating conditions or geometrical and material properties.
This talk discusses info-gap decisions in economics and management. We consider specific examples such as risk assessment in project management, and principal-agent negotiation under uncertainty. We also discuss the info-gap analogs of classical concepts in the economics of uncertainty risk-sensitivity, gambling and the value of information.
The need for value judgments is widespread in engineering analysis and design. Frequently the engineer is called upon to choose between competing considerations and to make trade-offs between conflicting interests. It is sometimes difficult to decide that a given option or outcome is desirable. Knowing how desirable it is, or how much better than a given alternative, is often far more challenging.
What we need is an explicit linguistic evaluation, in natural language whose meaning we can relate to everyday values, of the results of an engineering analysis. The format for this is REASONING BY ANALOGY. We will apply analogical reasoning to two classes of problems (1) design certification of safety-critical systems and (2) calibration of reliability assessments. Our discussion will be made in the context of information-gap decision theory.
Send e-mail to author at yakov@aluf.technion.ac.il