"The Sand Pile Effect" by Jean-Jacques Auffret
The term "normal accident" was coined to express the concept that some systems will sometimes present configurations so complex and tightly-coupled that the patterns of interactions between their components will inevitably lead to catastrophic results.
The bad news, according to the pioneer of this theory, Charles Perrow (1984), is that attempts to fix those systems will only make things worse, because they will result in more complexity, hence more accident sources. Accidents are in fact an intrinsic part of those systems, and not some kind of flaw that can be eliminated from them. In these systems, accidents are "normal," Perrow argues and not at all surprising.
Airliners are a very good example of such systems. Their stability, expected by all passengers, is only an illusion. How many tourists or business travelers know that the toilet system or the coffee machines have been known to blow some of them away in flight? Chaos theory shines a light onto such situations by explaining, on scientific grounds, how apparently trivial systems based on a handful of components and a couple of interactions, can actually produce very complex and, above all, unpredictable behavior.
Another example, from the natural world, is a mountain like Mount Hood, Oregon, which is considered an "easy" climb. However, this peak claims at least one life each year, and sometimes many more, as in the accident that occurred on May 30, 2002, where 9 men did slip across the slope, roped together, and finished their course into a crevasse where three of them died and another five were severely injured. Considered as a system, the mountain is bound to produce "normal accidents", as its track record shows. Problem is that this system being chaotic, it is impossible to predict when, how and to whom these normal accidents will happen.
The behavior of a sand pile helps to understand how such systems behave. Imagine sand pouring from an hourglass on a flat surface. That is a system with only one component - the sand grain, and one interaction force - gravity. One can think that such a simple system would display a deterministic behavior. It actually does, at macro level: past a certain critical height, the sand pile becomes self-organizing and maintains a certain triangle profile. But it does so, at micro level, by means of sand avalanches which occur in a totally unpredictable way. At some point, the new sand grains falling on the top of the pile from the hourglass will trigger small or great avalanches on the flanks of the sand pile. But nothing in the laws of gravity nor the shape or mass of the new grain can help predict if an avalanche will occur, and its magnitude in case it does.
The tiny points of contact between the climbers and the mountain act like the similar points of contact between the sand grains in the pile. At some point, those contact points, which in most normal case ensure the stability of the system, can actually play against it. It is also worth noting that the rope tying climbers together, which is originally designed as a safety device under normal conditions, can in no time turn into a deadly mechanism as it transmits the huge quantity of kinetic energy accumulated by the first falling men to the rest of the group.
In fact, the sand pile shows that such systems operate under permanent accident conditions. Sand grains roll onto each others. Crampons slip by a few inches. But once in a (generally long) while, one of those permanent accidents gets bigger, because of a domino effect with other system components around. As an engineer commenting on the space shuttle Columbia accident puts it: "Shit happens, and if we just want to restrict ourselves to things where shit can't happen, we are not going to do anything very interesting."