Toward Considering Risk Attitudes in Engineering Organizations Using Utility Theory

Abstract

Design projects within large engineering organizations involve numerous uncertainties that can lead to unacceptably high levels of risk. Practicing designers recognize the existence of risk and commonly are aware of events that raise risk levels. However, a disconnect exists between past project performance and current project execution that limits decision-making. This disconnect is primarily due to a lack of quantitative models that can be used for rational decision-making. Methods and tools used to make decisions in risk-informed design generally use an expected value approach. Research in the psychology domain has shown that decision-makers and stakeholders have domain-specific risk attitudes that often have variations between individuals and between companies. Risk methods used in engineering such as Failure Modes and Effects Analysis (FMEA), Fault Tree Analysis (FTA), and others are often ill-equipped to help stakeholders make decisions based upon risk-tolerant or risk-averse decision-making conditions. This paper focuses on the specific issue of helping stakeholders make decisions under risk-tolerant or risk-averse decision-making conditions and presents a novel method of translating engineering risk data from the domain of expected value into a domain corrected for risk attitude. This is done by using risk utility functions derived from the Engineering-Domain-Specific Risk-Taking (E-DOSPERT) test. This method allows decisions to be made based upon data that is risk attitude corrected. Further, the method uses an aspirational measure of risk attitude as opposed to existing lottery methods of generating utility functions that are based upon past performance. An illustrative test case using a simplified space mission designed in a collaborative design center environment is included. The method is shown to change risk-informed decisions in certain situations where a risk-tolerant or risk-averse decision-maker would likely choose differently than the dictates of the expected value approach.