Complexity, Coupling, Catastrophe

Watching different countries and governments respond to the Covid pandemic has motivated me to (morbidly) read more about disasters. To see how humankind dealt with the unknown unknowns of epidemics or pandemics, I recommend The Pandemic Century: One Hundred Years of Panic, Hysteria, and Hubris. Published in 2019, the author Mark Honigsbaum takes you on a case-study-style tour from the Spanish Flu to Zika (with “parrot fever” in between). The chapter on SARS is particularly interesting, given the present crisis – the same hysteria and hubris echo strongly. Are we learning the lessons of history? Or are we simply being human?

But that’s not the book I want to discuss in today’s post. While breezing through Pandemic Century, I’m also slowly working my way through Normal Accidents by Charles Perrow. The title is misleading, given the impetus for this 1984 book was the analysis of the Three Mile Island nuclear power plant “disaster”, it’s anything but normal. Perrow explains that “normal” here means “inherent to the system”, and that this is an ever-present danger in interactively complex and tightly coupled systems. Things will go wrong. It’s not a question of if but when.

Perrow is a sociologist who studies organizations. I’ve dabbled in this literature as part of my interest in complex systems. Sure enough, garbage can theory makes its appearance, but my focus today is on Complexity and Coupling, two key factors Perrow introduces in his analyses. The title of today’s post comes from the title to Chapter 3 in Normal Accidents.

Complex interactions have numerous subsystem linkages, some seen some unforeseen, multiple paths, feedback loops, and “connections… multiply as other parts or units or subsystems are reached”. They are distinguished from linear interactions (serial, one thing after another), but they are not labeled non-linear, because there are non-linear systems that are comprehensible. Complex systems, according to Perrow, have hidden interactions reducing their comprehensibility. He argues that this is particularly true of processes involving “transformation”, i.e., they “can be described, but not really understood”. He groups nuclear power production, chemical plants, and recombinant DNA technology in this category. Interestingly, these are all underlying processes where you can’t quite “see” what is going on.

Systems can be coupled tightly or loosely. Tightly coupled systems can respond quickly to changes in lockstep; they tend to be classified as more efficient. Loosely coupled systems, on the other hand, “allows certain parts of the system to express themselves according to their own logic or interest” but are more robust to recovery if something breaks. Perrow goes into detail with many examples of what it means to be tightly or loosely coupled not just from an engineering perspective but also from human organizations.   

Besides nuclear power, Perrow analyzes several systems of interest to explore the landscape of complexity and coupling. These include petrochemical plants, aircraft design and air-traffic control, marine accidents, dam and mine accidents, outer space exploration, and recombinant DNA technology. He then places these, and other systems, along a two-axis chart as shown below.

Universities are in the bottom right quadrant. They are interactively complex, but very loosely coupled. Or at least that’s how Perrow views them in 1984. (You can also see trade schools and junior colleges on the chart.) That seems fair. Universities have, in Clark Kerr’s words, become “multiversities”. There are many stakeholders, seemingly divergent goals, and seemingly slow-to-change behemoths. Traditional universities are not known for their nimbleness. In 2020, there is even more multi to the multiversity. The entrance of educational upstarts into an increasingly competitive space has brought about higher education’s own hubris and hysteria. As the “crisis” heightens, university administrations increasingly insist on more centralization to avert seemingly looming disasters. Covid-19 will accelerate the trend towards the All-Administrative University. Never let a good crisis go to waste.

Perrow’s primary polemic is highlighting the serious dangers of nuclear power plants and weaponry, from the technical but also the organizational point of view. These sit in the top right corner of his chart. It has to do with how these systems could and perhaps should organize themselves to deal with the inevitable crises, black swan events notwithstanding. Should these organizations centralize or decentralize? That is the question! Let’s take each of the quadrants in turn.

TOP LEFT: In interactively linear and tightly coupled systems such as dams, power grids, and rail transport, centralization is recommended. You want maximum efficiency out of these systems, and tight-coupling helps with that. Also since the system is not complex, it is possible to respond quickly and concertedly through being centralized. Basically, you want centralization for tight-coupling, and centralization is “compatible” with non-complex systems.

BOTTOM RIGHT: Let’s look at the opposite quadrant, applicable to mining, university, R&D. When you have an interactively complex system, Perrow argues (with the many examples in his book) that decentralization is desirable. When problems crop up, it’s often advantageous to have frontline people (preferably with experience and expertise) be highly involved and empowered to come up with solutions. Loose coupling allows for this, and information can move back and forth between operators and management without severe time crunch constraints.

TOP RIGHT: The problem in this quadrant is that centralization is needed to cope with tight coupling. But interactively complex systems are better solved through decentralized approaches. Thus, neither approach is optimal and Perrow predicts that the organizations of these systems will ping-pong back and forth with a mixture of approaches that will constantly evolve depending on what disaster hits. Organizational structure thus varies reactively.

BOTTOM LEFT: Since the interactions are not complex, and they are loosely coupled, it doesn’t matter which you choose – both centralization and decentralization are compatible. Perrow notes however that “elites” (the bosses) tend to favor centralization over decentralization in most cases. That’s no surprise. One theme that Perrow emphasizes as a sociologist is power relations within the system.

In my recent post on collegiality, my bias towards decentralized “organic” subunits within the university system is apparent. As a mid-level administrator, one of my tasks was to fend off higher-administration from its tendency to recommend centralized solutions in a one-size-fits-all approach. Financial fears have made these tendencies more acute. Administrators want more tight coupling. At the same time administration is increasing in scope and manpower with increasing interactive complexity of the multiversity. This combination will move universities towards the upper right quadrant. More fail-safe systems must be put in place, further increasing interactive complexity. It’s a positive feedback loop.

Even if we were not in Covid crisis, and the university isn’t like a nuclear power plant with catastrophic potential, the move towards increasing centralization and top-down approaches is very worrying. And it’s all in the name of supposedly running leaner, better, more efficient, whatever. Humans are involved. I close by quoting Perrow.

Organizational theorists have long since given up hope of finding perfect or even exceedingly well-run organizations, even where there is no catastrophic potential. It is an enduring limitation – if it is a limitation – of our human condition. It means that humans do not exist to give their all to organizations run by someone else, and that organizations inevitably will be run, to some degree, contrary to their interests. This is why it is not a problem of capitalism; socialist countries, and even the ideal communist system, cannot escape the dilemmas of cooperative, organizational effort on any substantial scale and with any substantial complexity and uncertainty. At some point the cost of extracting obedience exceeds the benefits of organized activity.

P.S. As one studying the origin-of-life, I've also been reflecting on the interplay of complexity and coupling as I learn biochemistry! More about that in a future post.