The Azimuth Project
Risk management (changes)

Showing changes from revision #2 to #3: Added | Removed | Changed


Risk management consists of assessing and prioritizing risks and then taking steps to monitor them, minimize their probability, and minimize their negative effects. See:

One place to get started for risk in ecosystem restoration is:

Among the most active people in risk management in environmental contexts is Igor Linkov:

  • Linkov et al., eds, Real-time and Deliberative Decision Making, Springer, Berlin, 2008.

Deep uncertainty refers to situations where the parties to a decision do not know or cannot agree on how to model a system and estimate the probabilities of events. In the presence of deep uncertainty it is often preferable to seek policies that are robust, meaning that they work well (if not perfectly) over a range of possible assumptions, without having to explicitly assign correct probabilities to all outcomes. Rob Lempert at RAND, among others, has worked on robust decision-making with respect to climate change:

Abstract: Many commentators have suggested the need for new decision analysis approaches to better manage systems with deeply uncertain, poorly characterized risks. Most notably, policy challenges such as abrupt climate change involve potential nonlinear or threshold responses where both the triggering level and subsequent system response are poorly understood. This study uses a simple computer simulation model to compare several alternative frameworks for decision making under uncertainty—optimal expected utility, the precautionary principle, and three different approaches to robust decision making—for addressing the challenge of adding pollution to a lake without triggering unwanted and potentially irreversible eutrophication. The three robust decision approaches—trading some optimal performance for less sensitivity to assumptions, satisficing over a wide range of futures, and keeping options open—are found to identify similar strategies as the most robust choice. This study also suggests that these robust decision approaches offer a quantitative, decision analytic framework that captures the spirit of the precautionary principle while addressing some of its shortcomings. Finally, this study finds that robust strategies may be preferable to optimum strategies when the uncertainty is sufficiently deep and the set of alternative policy options is sufficiently rich.


See also this article by Azimuth Forum member Curtis Faith:

The ensuing discussion turned up a number of interesting references which were used in this article. Tim van Beek pointed out an interesting book about common mistakes that people make when they try to control complex systems:

  • Dietrich Dörner, The Logic Of Failure: Recognizing And Avoiding Error In Complex Situations.

Dietrich Dörner is a professor of cognitive behavior and psychology. In his book he describes among other things an experiment, where the subjects had to come up with a plan to improve the situation in a poor country, much like a minister of development aid of some rich western country would today. Behind the curtain was a complex system of coupled differential equations modeling some typical aspects of the target country, with computer operators feeding the model with the input of the subjects.

Of course one could engage in endless discussions about the validity of the computer model, but it is nevertheless possible to identify typical mistakes made by most subjects. For example: After they had developed a strategy, most subjects did not ask about the current development of the target country (they were free to ask for any kind of information they wanted at any time). Instead they waited for the model to complete the model run (some decades of simulated time) and were shocked when they were told that their decisions had some unwanted long-term consequences. These were obvious long before the model run was completed, but the probands simply did not ask about them.

The mathematical statistics of extreme values has been extensively developed:

In brief, the Gumbel distribution is to extreme values what the Gaussian distribution is to average values. But even so, expected maxima are difficult to calculate properly.

The science of decisions has also been highly developed, mostly through operations research:

The most used concept involves the receiver operating characteristic or ROC curve, originally developed for military reasons:

category: methodology