Complex Behaviour

“Does the flap of a butterfly's wings in Brazil set off a tornado in Texas?”

Description

Many of the scope-related factors describe ways in which a system can have a complex structure. However, the behaviour of a system can also be complex. Complex behaviour means that the system's outcomes change in many different and/or irregular ways. If we can't determine patterns in system behaviour, it can seem random to us. In extreme cases, the system's behaviour is completely unpredictable; we call this chaos.

Sensitivity

Complex behaviour isn't always completely weird. The initial movement of the double pendulum (above) is unsurprising. However, such systems tend to be very sensitive to their initial conditions. For example, if we started the pendulum from a slightly different position, its behaviour after ten seconds could be completely different. This can be a problem in practice because we often don't know a system's initial conditions accurately enough to be able to forecast its behaviour for very long.

Transients

A complex system can have periods during which its behaviour is significantly different to the overall pattern. For example, in the double pendulum animation above, the pendulum with the fixed end usually swings from one side to the other (albeit not smoothly). However, occasionally it swings to one side, stops moving, and then swings further to the same side. If we saw this behaviour in isolation, we might assume that this is what the pendulum normally does. Or, when the pendulum first does that, we might assume that its behaviour is entering a new phase and it will continue to act in that manner.

Transient behaviour is especially problematic if the system we're contemplating has a long time-scope. If the system's behaviour changes slowly, we're more likely to notice transients and more inclined to think that they're more significant than they really are. For example, one cold year — or even several in a row — doesn't prove that climate change can't be happening.

Emergent Behaviour

A saving grace can be emergent behaviour. Even if we can't tell what the exact temperature will be on a particular date a few decades away, the overall temperature trend may still be discernible. We can also take into account known cycles, such as seasons and ice ages. This sort of limited predictive ability may be sufficient to allow  us to manage the problem we're dealing with. Dealing with emergent behaviour is often more important than influencing any one incident.

Prevalence

We need to keep complexity in perspective. Most systems aren't so complex that prediction is impossible. If they were, our instincts and intuition would be wrong much more often. However, some systems are complex and have surprising or unpredictable behaviour. We should not assume that we can always guess a system's behaviour; sometimes we need to work harder or find alternative ways of dealing with it.

Examples

Coin Toss

A coin and its environment aren't a structurally complex system. However, the system is very sensitive to initial conditions that we can't create or assess accurately enough to allow us to determine the outcome of a coin toss. An imperceptibly small variation in the rate at which we set the coin spinning can make the difference between a head and a tail.

Sand Pile

If we try to drop grains of sand over the same spot, the grains don't balance on top of each other and form a tall pillar. Instead, most newly-added grains fall off the top of the pile and tumble some way down one of the sides. Despite our best efforts to drop the grains from the same position, we can't accurately predict whether a grain will tumble to the north, or to the east, or any particular direction. Nor can we accurately predict how far down the pile it will tumble before coming to rest. So, even though we're trying to do the exact same thing with every grain, the outcomes can be totally different and unpredictable.

However, we can predict the emergent behaviour fairly accurately (although not precisely). If we drop enough grains of sand, we'll get a roughly cone-shaped pile. Even with the occasional minor avalanche, the overall shape of the pile will remain similar. It will just get bigger as we add more sand.

Weather

Weather forecasting is the original context of the 'butterfly effect' quotation. Chaos in the weather system makes research into global warming more difficult.

Supply Chains

Dependencies between suppliers form a complicated web of relationships. For example, a semiconductor shortage in Taiwan could halt truck production in Detroit, which could then delay food deliveries to grocery stores weeks later.

Road Networks

Complicated road networks and decisions by individual drivers form a complex system with occasionally surprising behaviour. For example, adding a new road can actually make traffic worse because of how drivers change their behaviour (see Braess's Paradox).

Related Issues

This section describes similarities and differences between this issue and related problem factors and thinking traps.

Element and Relationship Scopes

Large numbers of elements and/or relationships in a system give a kind of complexity called detail or structural complexity, which can make the system hard to understand. However, the focus of complex behaviour is on how the system responds rather than what is in it. As the double-pendulum example shows, complex behaviour can arise from a system with only a few elements and relationships.

Unclear Causality

Complexity is similar to unclear causality in that aspects of system behaviour can be hard to understand. The focus of unclear causality is the relationship between two elements within a system, whereas the focus of complex behaviour is on the system as a whole. In the examples above, the relationships between elements within each system are well understood (i.e., there is no unclear causality), but the systems' behaviours are still complex.

Probability

Probability and uncertainty can be consequences of complexity. A coin toss seems random because the exact motion is too complex for us to predict, even using a computer. The outcome is highly sensitive to conditions we can't accurately measure, such as how the coin is launched and caught. Even variations in air currents can affect the result. For systems like this, where we can't accurately predict the behaviour of the system, we have to accept that the system might produce a range of different behaviours, with each possibility having some probability of occurrence.

If we can solve our problem based only on the system's emergent behaviour, it may not be necessary to consider the probabilities of the possible outcomes. A risk here is that tipping points in the system can result in the system having more than one emergent behaviour.

Intuition

Intuition can be misled by complexity. Intuition assumes that systems behave simply or obviously. Before you saw how a double pendulum can move, would you have predicted it to move as it does?

Unconfidence

If the problem system's behaviour is complex — and especially if it is chaotic — we should have a justifiable lack of confidence in our ability to manage the system. This doesn't necessarily mean that we should do nothing or just guess.

Related Engine Processes

Later, we will describe a thinking framework that comprises multiple processes. This section points forward to the processes that deal with this page's topic.

If a system's behaviour is totally unpredictable, proactive attempts to manage it may be no better than guesses. Any actions we initially take could do more harm than good. Adaptation is especially important for such problems: we can change our actions as we learn more about what the system is doing in reality and how it responds to our previous actions.

Further Reading

---

To Do

Murdered Darlings