35 Optimizing Risk Decisions with Imperfect Data

Abstract

In any integrity management program, there are always competing alternatives for any decision. Such as dig or not to dig, replace or not to replace. If perfect information were always available, like a math problem where everything except the answer is given, risk engineers would be an unnecessary expense. The next best alternative would be an exhaustive corpus of data with frequencies of every outcome and condition combination. However, in all but the most trivial cases, these don’t exist either and the engineer is dealt partial, imperfect information where the true state of nature is uncertain. To deal with this uncertainty in everyday life people develop heuristics, mental shortcuts that allow us to process this information with the least amount of effort and time. But when the probabilities are imprecise and the data imperfect, decisions based on these shortcuts can be fraught with biases and fallacies. All decisions carry some risk that is dependent on the (uncertain) true state of nature, and the potential loss associated with a given course of action.

This paper will discuss the fundamentals of decision theory and demonstrate an innovative application of them that incorporates existing knowledge and potential consequences. This will be used to quantify the tradeoffs of competing alternatives in an pipeline integrity management program to arrive at a decision that minimizes the risk based on the state of knowledge that is available.

Read/Download PaperGradeAvailable after the event