By Leena Ilmola-Sheppard, IIASA Advanced Systems Analysis (ASA) Program
Crisis management problems are getting more complex and complicated, but at the same time, governments have less and less resources for their management. How can research help decision makers plan for the unplannable?
Last week in Geneva, I took part in a crisis management workshop for national decision makers organized by the OECD High Level Risk Forum and the Swiss Federation Chancellor While the meeting was very specific to national security and crisis management, I found some takeaway messages that are relevant to us researchers as well, especially for those of us that hope that to help decision makers make better decisions through modeling.
Mads Ecklon, Head of the Centre for Preparedness Planning and Crisis Management of the Danish Emergency Management Agency, used the figure above as a framework to explain crisis management. His message can also be applied to the development of any social system. Picture 1 describes the standard starting point of the modeling exercise. We are modeling one behavior and then analyze how the system performance develops in a controlled situation. Ecklon explained that potential futures are not so predictable: the crisis in hand can either be solved, solved only partially, not solved at all, or in the worst case the problem may escalate (you never know how a social system will react in the crisis situation—a small incident can turn into a massive riot). The challenge for both national level crisis managers and modelers is same; you have to take all of these potential developments into a consideration.
But what happens if a new, unexpected crisis pops up while all attention is focused on the initial problem? Such hard-to-predict events are often referred to as “black swan events.” Eclon said that their team has more frequently been seeing situations where, when attention is focused on the current crisis, a new, different or related, crisis develops and no one notices it. For example, in the UK in 2007, just when all the crisis management resources were invested in flooding crisis, foot and mouth disease broke out among cattle. The new phenomenon, Ecklon claimed, is that these crises are piling up and even if they are independent from each other, the joint impact can be disastrous.
Modeling black swan events
I think that this message is important for modelers as well. We may be very happy to model all the four windows of our comic strip. But how can we include new surprises and crises into an ongoing model? We should develop models that include different development trajectories triggered by a change in one of our variables, but simultaneously we should be able to account for several overlapping surprises.
In the meeting, national risk managers spoke about ”unknown unknowns,” low probability high impact risks–strange unforeseen animals like a black swan that jump on the plate just when we think that the situation is in some kind of control.
This kind of modeling challenge is fascinating from an academic perspective, but researchers’ intellectual hunger should not be the only reason to develop methods for these kinds of situations. From decision makers’ perspective, this is exactly the case where useful models are needed. The multiple simultaneous developments of the complex systems are difficult to capture even for the brightest of the crisis teams, but a model could manage a job very well.
Most of the IIASA models are large, integrated models that cover global systems. These models are not designed for digesting black swan sandwiches. The Danish crisis management team has a solution worth for benchmarking for this problem as well. They have a specific small team that is called a Pandora’s Cell. Pandora’s Cell is dedicated to anticipating, imagining, and scanning for potential not-so-obvious developments that should be taken into consideration in decision making. This dedicated team is needed because all the other resources available have been focused on the obvious events, as described in the square one of our comic strip.