About the Author

Beyond UQ: Dealing with Deep Uncertainty

By Hans Kaper

Most SIAM News readers are familiar with the concept of uncertainty quantification (UQ). But when I received an invitation to participate in a workshop on “Decision Making Under Deep Uncertainty,” I was unsure of what made “deep uncertainty” special.

The workshop, organized by the Society for Decision Making Under Deep Uncertainty (DMDU) and officially the society’s 2016 Annual Meeting, was hosted by the World Bank in Washington, D.C., last November. It was preceded by a day of training for novices (like me) who wanted to learn more about the concepts and tools of DMDU.

DMDU differs from UQ. It plays an important role in the development of policy and business strategy under an extreme degree of uncertainty, i.e., when multiple plausible alternatives exist that cannot be ranked in terms of their perceived likelihood. This incapacity for ranking may be due to a lack of data, or a lack of knowledge of the mechanisms or functional relationships that govern the behavior of the system under consideration. In the worst-case scenario, all we know is that we are dealing with unknowns. But—and this is the essential element in DMDU—this ignorance is factored into the decision-making process. Think of a so-called “black swan” event — an event that lies outside the realm of regular expectation and is explainable only after the fact. Such events are more common than we think, and their impact can be catastrophic. One example is the level 9.0 earthquake that hit Japan in 2011. A tsunami and a nuclear catastrophe followed, which then led to supply chain disruptions (of automobile parts, for example) around the world.

UQ’s premise is that uncertainty can be reduced, e.g., by gathering more information. With DMDU, probabilities are fundamentally unknowable and unpredictable. Yet decision-makers must make decisions under this level of uncertainty, and these decisions often concern major infrastructure projects that have long life spans and require significant investments. Consider the World Bank and infrastructure projects in developing countries.

When faced with uncertainty, decision-makers generally emphasize one of the following:

  • Resistance: plan for the worst conceivable case
  • Resilience: develop a strategy that results in quick recovery after an unanticipated event
  • Robustness: develop and implement a policy that will perform reasonably well in all conceivable situations.

A policy based on resilience does not account for black swan events and may therefore be costly, while one focused on recovery may lead to possibly significant short-term losses. A robust policy, on the other hand, yields outcomes that are satisfactory across a wide range of scenarios, according to some predetermined assessment criteria. This is in contrast to an optimal policy, which may achieve the best results among all possible plans but carries no guarantee of doing so beyond a narrowly defined set of circumstances.

Robust strategies are appropriate when uncertainty is deep or decision-makers face a rich array of options. Instead of attempting to characterize uncertainty in terms of probabilities, as is done in UQ, deep uncertainty explores the possible effects of different assumptions about future values of the uncertain variables for the decisions actually at hand. As one of the workshop speakers noted, if something seems worth doing, it is worth doing first superficially. An exploratory approach might reveal options and provide an initial assessment of pathways for future considerations.

Workshop participants presented several case studies concerning major infrastructure projects where decision-makers faced deep uncertainty. The projects ranged from water and energy systems planning, flood risk management, infrastructure development, transportation networks, forest management, and public health to policy negotiations and security cooperation. The most important drivers of uncertainty in these projects were found to be climate change, rising sea levels, population growth, technology breakthroughs, and social choices.

A pitfall in all of DMDU is so-called “presentism” — a bias toward the present that too often results in “regrets” when prior commitments block decision paths. Another disadvantage, especially in the design of dynamic adaptive strategies, is a focus on the wrong indicators to monitor progress. Presenters gave examples where indicators did not align with a project’s overall objectives, resulting in unintended consequences.

The workshop program also featured hands-on sessions about hypothetical decision-making problems and a fascinating plenary TED Talk by Andrew Revkin titled “Conveying Wicked Climate and Energy Realities in an Uncertain Communication Climate.” Revkin was trained as a biologist but made a career shift to investigative journalism. He writes the Dot Earth blog about science and environmental issues for The New York Times and recently moved to ProPublica, where he focuses on how countries and companies are—and are not—responding to climate change.

The workshop left me with some food for thought. Did I learn any new mathematics? No, but I discovered that DMDU touches on several areas of interest to the SIAM community, namely UQ and the mathematics of planet Earth. Thus, here is a new opportunity for collaboration with our community, with potentially significant benefits for society.

More information about the Society for Decision Making Under Deep Uncertainty can be found on their website.

Acknowledgments: Some of the material in this article is based on the document “Deep Uncertainty” by Warren E. Walker, Robert J. Lempert, and Jan H. Kwakkel (2016), distributed at the “Decision Making Under Deep Uncertainty” workshop.

Hans Kaper, founding chair of SIAG/MPE and editor-in-chief of SIAM News, is an adjunct professor of mathematics at Georgetown University.