The upshot is: Infrastructure cascades and catastrophizing about infrastructural failure have a great deal in common and this has major implications for policy and management.
An infrastructure cascade happens when the failure of one part of the critical infrastructure triggers failure in its other parts as well as in other infrastructures connected to it. The fast propagation of failure can and has led to entire systems failing over quickly, where “a small mistake can lead to a big fail.” The causal pathways, however, in the quick succession or chain reaction of interconnected failure are often difficult to pinpoint, let alone analyze, during the cascade, and even afterwards.
For its part, catastrophizing in the sense of “imagining the worst outcome of even the most ordinary event” obviously has some overlap with this notion of cascade, but by and large the imagining in catastrophizing is written off as irrational—the event in question is not as bad as imagined—while infrastructure cascades are real, not imagined.
We may want to rethink any such weak link when it comes to infrastructure cascades and catastrophizing failure across interconnected infrastructures. Consider the insights of the recent book by Gerard Passannante, Catastrophizing: Materialism and the Making of Disaster (2019, The University of Chicago Press).
In analyzing cases of catastrophizing (in Leonardo’s Notebooks, an early work of Kant and Shakespeare’s King Lear, among others), Passannante avoids labeling such thinking as irrational and favors a more nuanced understanding. He identifies four inter-related features to the catastrophizing from his material.
First (no order of priority is implied here), catastrophizing probes and reasons from the sensible to the insensible, the perceptible to the imperceptible, the witnessed to the unwitnessed, and the visible to invisible. Accordingly, the probing and reasoning involve ways of seeing and feeling as well. Second and third, when catastrophizing, an abrupt, precipitous shift or collapse in scale occurs (small scale suddenly shifts to large scale), while there is a distinct temporal elision or compression of the catastrophe’s beginning and end (as if there were no middle duration to the catastrophe being imagined). Last, the actual catastrophizing while underway feels to the catastrophizer as if the thinking itself were involuntary and had its own automatic logic or necessity that over-rides—“evacuates” is Passannante’s term—the agency and control of the catastrophizer.
If so, the features of catastrophizing take us much closer to the notion of infrastructure cascades as currently understood. In catastrophizing as in cascades, there is both that rapid propagation from small to large and that temporal “failing all of a sudden.” In catastrophizing as in cascades, causal connections—in the sense of identifying events with their beginnings, middles and ends—are next to impossible to parse out, given the quick, often inexplicable, processes at work. And yes, of course, cascades are real, while catastrophizing is speculative; but: The catastrophizing feels very, very real to, and out of the control of, the catastrophizer as an agent in his or her own right.
In fact, one of the most famous typologies in organization and technology studies—that formalized by Charles Perrow in terms of coupling and interactivity of complex technologies—sanctions a theory that catastrophizes infrastructure cascades. The typology’s cell of tight coupling and complex interactivity is a Pandora Box of instantaneous changes, invisible processes, and incomprehensible breakdowns involving time, scale and perspective. This is not a criticism: It may well be that we cannot avoid catastrophizing, if only because of the empirical evidence that sudden cascades have happened in the past.
The four features, however, suggest that one way to mitigate any wholesale catastrophizing of infrastructure cascades is to bring back time and scale into the analysis and modeling of infrastructure cascades. To do so would be to insist that really-existing infrastructure cascades are not presumptively instantaneous or nearly so. It would be to insist that infrastructure cascades are differentiated in terms of time and scale, unless proven otherwise. That, in fact, is what our research suggests. At the risk of tooting our horn:
“Much of the more sophisticated network analysis of interinfrastructural interconnectivity suffers from the same defect as sophisticated quantified probability assessments—both assume that if an infrastructure element (node or connection) is not managed, the system is not managed. One clear objective of recent network of networks modeling has been finding out which nodes and connections, when deleted, bring the network or sets of networks to collapse. Were only one more node to fail, the network would suddenly collapse completely, it is often argued…
“But ‘suddenly’ is not all that frequent at the [interconnected infrastructure] level. In fact, not failing suddenly is what we expect to find in managed interconnected systems, in which an infrastructure element can fail without the infrastructure as a whole failing or disrupting the normal operations of other infrastructures depending on that system. Infrastructures instantaneously failing one after another is not what actually happens in many so-called cascades, and we would not expect such near simultaneity from our framework of analysis.
“Rapid infrastructure cascades can, of course, happen….Yet individual infrastructures do not generally fail instantaneously (brownouts may precede blackouts, levees may seep long before failing), and the transition from normal operation to failure across systems can also take time. Discrete stages of disruption frequently occur when system performance can still be retrievable before the trajectory of failure becomes inevitable.” (E. Roe and P.R. Schulman, Risk and Reliability, 2016, Stanford University Press, pp. 28-29)
Let me leave you with another extension inspired by Passannante’s analysis. If infrastructure cascades, when catastrophized, have endings entailed in their beginnings (leaving in between only attenuated middles or no middles at all to speak of), the catastrophized cascade turns out to be the entailment of “just before” and “immediately after.”
That is, we are to believe we are in a state where disaster avoidance is no longer possible and disaster response has yet to start and remains unavoidably ahead. We are expected to experience cascade-as-disaster as all too close at hand for us to think about anything else. Closing time in the gardens of the West has always been, well, right now. But the point is: Every one experiences time as anfractuous, full of twists and turns–why else all the interruptions? No one experiences time and scale as an excluded middle; what is unimaginable are real-time operations without duration and depth.
To summarize: One criterion that distinguishes crisis scenarios apart from each other is less discussed today than more popular criteria that appeal to best available science and technology: How well does the scenario differentiate time? Is the duration under which the crisis unfolds discontinuous or continuous? A major storm is of set duration; a pandemic takes longer; while the half-life a thermonuclear war…well, that rupture and its aftermath could extend for all practical purposes indefinitely. I wish to suggest that many crisis narratives today say they are taking time seriously by insisting on the immediate urgency and action the crisis is said to demand, when instead had duration of the crisis in question been better identified and specified, so too would have been the urgency and the policy/management implications said to follow.