Tool Development for Risk Management Decisions under Strong Uncertainty

We present experiences and developments from cases of utilizing multi-criteria decision analysis in societal risk management decision problems. Of particular interest is the pursuit of a common framework for computer support, which has been developed from learnings accommodated by participation in societal flood risk management and renewable energy pathways projects.

In general, quantitative methods are desired for discriminating between various risks and management options regarding their severity and efficiency. Probabilistic approach determines the risk in terms of frequency and consequences, and various methods for quantitative risk management have been developed over the years, e.g., extreme weather and operability (HAZOP) analysis, failure modes and effects analysis (FMEA), event tree analysis (ETA), and safety management organization review technique (SMORT). However these are often demanding when it comes to data requirements, and approaches for relaxing these requirements in probabilistic methods have been proposed, such as sets of probability measures, interval probabilities and utilities, fuzzy probability, and possibility theory. In actual practice though, surprisingly little has been done to take such perspectives into consideration when performing risk analyses and using them to inform decision makers. In many cases, this seems to depend on that these more realistic models are perceived as too difficult to use for practitioners. Also, the computational complexity often introduced by extended methods can become problematic.

In our cases, large uncertainties were involved, calling for an approach capable of providing quantitative risk assessment in a numerically imprecise domain. We therefore propose new methods for analysing risk management options, built upon analytical approximations of second-order belief functions where estimates can be made by using intervals, rankings, and valued ranking statements in a common model, without introducing any further complicating modelling aspects into the problems. The ambition is that this allows for intuitive interpretation of results, less complicated to interpret for decision makers and stakeholders, and to significantly increase the applicability of decision analytical risk management in complex and uncertain domains.