Strategies for Managing the Consequences of Black Swan Events

Behavioral scientists postulate that human experiences are often tainted with personal biases and formed from a rather limited time span of observations compared to the likelihood period of most extreme events. Consider the experience of people in the Indian Ocean region, who had likely never observed or even heard about tsunami, whether in normal discourse or in folklore. Except for a few experts almost no one was aware of their reality and considering the absence of any past observed occurrence or experience, no one had visualized this unusual event. The New Orleans residents had been through a lot of storms before hurricane Katrina and nothing that bad had ever happened. In hindsight, it is clear that there was no engineering redundancy that would save the residents if a large enough storm dumped water in short duration and the levees were breached. Could someone coming to work in the World Trade Center towers on the morning of September 11 have imagined the strange experiences of that day? Then, why are engineers often so dismissive of such scenarios until after they occur?

It may be because engineers do not like uncertainty and ambiguity, focus on specifics instead of generalities, and crave explicit explanations. Such thinking is shaped by their training in linear logic, whereas nearly all Black Swan events involve complex causal relationships. Taleb identifies five peculiarities of human behavior responsible for blindness to Black Swans:These peculiarities of human behavior are illustrated here with examples from science and engineering, as opposed to Taleb’s general historical and financial focus.

The following catastrophic out-of-the-ordinary events that happened in recent years were impossible to predict ahead of their occurrence and probably would never have been considered in risk assessment of critical engineering projects sited in these locations:

The first incident had a major impact on population centers, whereas serious consequences from the second incident were thwarted due to absence of engineering projects in the vicinity. “If a tree falls in a forest and no one is around to hear it, does it make a sound?” Hundreds of meteorites strike the earth every day, occasionally large ones, and even the cataclysmic Tunguska strike of 1908 in Siberia did not result in engineering consequences. But an ill-timed meteorite strike on a critical nuclear facility may have potentially huge consequences. A similar situation of being in the wrong place at the wrong time was experienced by people in littoral areas of the Indian Ocean during the monstrous December 2004 tsunami. And the effect of hurricane Katrina on New Orleans in August 2005 was catastrophic as storm surge caused major breeches in drainage canal levees, flooding 80 percent of the city with water as deep as $15\text{feet}$ in many places. Though not as widely publicized, a new type of serious hazard was uncovered in 1986 when a giant enveloping bubble of high concentration dissolved gases, estimated to be 1.6 million tons and $150\text{feet}$ thick, escaped from the bottom of Lake Nyos in Cameroon and asphyxiated hundreds of people in the surrounding areas. These dissolved gases are under high pressure in deep water and unstable density stratifications can lead to complete churning of lakes, similar to what may happen in liquefied natural gas storage tanks.

Often, the object is in the wrong place but timing is right, resulting in what one calls a providential escape or near-miss event. During the Indian Ocean tsunami, high waves hit the operating nuclear power plant on the east coast in India, and it was safely shut down. But the foundation pit of a $500\text{-}\mathrm{MW}$ fast breeder reactor under construction at a nearby location was inundated with water. Subsequently, the site location was moved to higher ground and the design was modified. It was providential that the timing was right; the plant was under construction and not operating.

Thomas Kuhn (1962), in The Structure of Scientific Revolutions, argues that evolution of knowledge and ideas follows saltatory growth through a series of epistemic paradigm shifts resulting from observed anomalies. Engineering design philosophy has certainly undergone such conceptual changes. Conventional design approach, using factors of safety based on value judgment and past experience, is reasonable for parameters that vary over a limited range and where failure is not generally anticipated. These are the so-called White Swans, completely predictable by well-defined procedures.

The need to model natural and accidental extreme events, with parameters varying over a wider range, led to the development of probabilistic risk-based approaches, where operational and life safety consequences of hazards are linked to their return periods. Nearly all of the risk assessment scenarios depend on some form of speculative extrapolation, where the future is expected to be an extension of the past and the likelihood is estimated by using the frequency of past similar events. These extreme event models rely on meticulously compiled databases for estimating the risk due to random uncertainty and represent the so-called Gray Swans (known unknowns). Such risk modeling certainly represents a paradigm shift from designing for safety to explicit consideration of failures in engineering design.

Risk is defined in terms of the likelihood of an event occurring coupled with resulting consequences, but likelihood for epistemic uncertainty due to lack of knowledge is difficult to estimate. Black Swan events are not foreseen from observed data and are about unknown unknowns, where absence of likelihood information makes risk management methods futile, since in this world, even knowledge is of no use because experts do not really know what they do not know. Black Swans are not the same as low-probability, high-consequence events because they are unpredictable by probabilistic means. One can not estimate probability of bombing for the Oklahoma City federal building, since the event is not a random process. These “left field” outliers can vitiate all attempts at prediction since there is no historical precedence or data for prognostications. Therefore, one needs to shift emphasis from risk management to devising strategies for dealing with the consequences of such unforeseen events.

For a particular subset of Gray Swans that are low-probability, high-consequence events, the likelihood information is available but designing for extremes is prohibitively expensive. Therefore, risk assessments usually screen out events if there is either absence of evidence or negligible chance of occurrence over long return periods. But “absence of evidence is not evidence of absence.” No known hurricane has ever hit the coast of Georgia even though Florida and South Carolina have been frequently impacted. Does this absence of evidence allow one to assume low risk from hurricanes for buildings in Georgia? How does one inspire public trust and confidence in risk assessments when once-in-a- million-years events like the Exxon Valdez oil spill seem to happen so frequently?

A meteorite strike has less than a one in a million chance of striking a critical facility, but there are still 80 places on earth that are struck every year. Even though there is a low likelihood of large seismic events in the midwest United States, such events have occurred. The ship Titanic is a prime example of a low-probability design failure without consideration of consequences. The vessel was built with a double-bottomed hull and divided into 16 watertight compartments and was claimed to be unsinkable—even with four of these filled with water—and, therefore, more life boats were not considered necessary. These low-probability, high-consequence events show the perils of discounting risk based on the low likelihood of occurrence, without adequate consideration of the consequences. Instead, it may be more appropriate to apply strategies devised for managing the consequences of Black Swans to low-probability, high-consequence Gray Swan events. However, unlike Black Swans, using such strategies is a matter of choice for Gray Swans. The chemical and process industries have long considered the consequences of adverse events through the development of systematic methods. Civil engineers need to ponder the development of similar consequence-based management strategies to cope with the potential or actual consequences of such large-impact, hard-to-predict events.

As Henri Poincaré propounded and chaos theory confirms, nonlinearity puts a limit on forecasting because small effects can often lead to serious consequences, and forecasting extremes is a treacherous pursuit. Consider the shocking collapse of the Ronan Point apartment building in England; an occurrence completely out of proportion to the minor triggering event of a kitchen gas cylinder explosion in an apartment of the 22-story structure. An inconsequential event leading to another sequentially caused a chain reaction with disproportionate consequences, dubbed as progressive failure. It is instructive to consider that such minor events, which are never a part of engineering design, can lead to complete building collapse despite being designed for all the extreme scenarios conjectured in building codes.

How should one account for such outlier events in important civil engineering projects? Because of the infeasibility of predicting such occurrences, no perfect risk management approach is feasible. As risk management techniques based on likelihood arguments do not necessarily address Black Swan outliers, which by definition are unknown, engineers must think through the likely consequences of such occurrences and use their experience and judgment for developing suitable design and management strategies. These strategies are also applicable for low-probability, high-consequence Gray Swan events. Remembering Voltaire’s aphorism that “perfect is the enemy of the good,” some reasonable stratagems for the prevention, detection, control, and mitigation of hazard consequences are discussed here. The desired safety goals can be achieved by choosing appropriate tactics or combination of tactics from among these or other strategies.

The metaphor Black Swan refers to the unexpected, an unlikely but not impossible catastrophic event that no one seems to plan for, the things one does not know or does not know that one does not know. As the likelihood of occurrence for these unforeseen events cannot be estimated from observed data or prognosticated by experts, the usual risk management techniques are inapplicable for addressing the impact of Black Swan events on vital engineering facilities. Therefore, engineers must think through the likely adverse consequences of such unpredictable events and use their experience and judgment to devise suitable strategies focused on managing the consequences of these outliers.