What can system thinking bring to anticipation?
The concepts of system thinking and system dynamics were popularized in the ’70s after the publishing of “Limits to Growth” by Donella H. Meadows and “World Dynamics” by Jay Forrester. System thinking focuses on the detection of the underlying structures that generate events. In a sense, it tries to generate a model of the world. Its main characteristic is “non-linear” thinking. That is, a system, in which variables can influence each other, creating self-reinforcing loops.
Classic system thinking literature represents the causality ladder in the shape of an iceberg. It emphasizes that the events we witness have roots that are further than we immediately realize. However, system thinking adds relevance when generating anticipations because it considers the interplay of variables.
It occurs at different levels:
The first level is that of the event. It refers to what is actually happening. Most media fluxes and news feeds stay at this level. Following events is useful to build a dataset, based on which you can start to build an intuition and move on to the next levels of analysis.
A fox has eaten 5 chickens.
As we identify patterns in what is happening at the event level, we reach the level of trends. This is where we find correlations and simple statistical analysis. Machines have become extremely good at this game and are able to detect even ultra-weak signals, such as: “If on Tuesday at 19h57, the temperature is between 9 et 10 C, then there is 20% more chance that the babies born on the following Friday will be called John”. This is already useful for anticipation because you know that if you see the cues, you can prepare yourself for the event.
However, this is dangerous too, because correlation does not always mean causation (even if you need correlation for a causation, the link is not automatic at all). Some cues might be linked together through an invisible third variable. Suppose you have found a strong correlation between ice cream sales and the mortality rate of people over 70. Ice cream does not necessarily kill old people. But a heatwave might explain the behavior of these two variables.
If you observe chickens and foxes over time, you notice that the more the chicken population grows, the more that of the foxes grows, until a point where the number of chickens decreases.
The graph displays the correlation between the number of foxes and the number of chickens over time.
The next level focuses on the structure that is causing the pattern. At this level, we seek to expose causality. And for that, we need to propose and build a model.
Given that variables are interconnected in system thinking, there are constant flows and loops among them. By modeling such feedback loops within a system, you can explore deeper the mechanism of supposed cause and effects.
Here, you build a simple model where more eggs support the growth of more chickens, which then supports the growth of more foxes. However, more foxes mean fewer chickens, which in turn means fewer eggs. Fewer chickens mean fewer foxes. Fewer foxes mean more chickens and so on.
Model building is key for anticipation work because it is where you start articulating your current understanding of the world. By drawing the logical steps you imagine, you are making your hidden assumptions visible. And by making them visible, they become easier to debunk and modify. In this way, you are on the road to being less wrong.
The model explains the causation relationship between the change in the number of chickens, foxes, and eggs.
Mental models and intentions
In order to refine our analysis, we need to question what series of beliefs, worldviews, and intentions are behind the structure. The level of intentions is not relevant if we assume chickens and foxes have little more intention than eating. However, the anticipation of crises and human affairs deals with humans and their complexness. No crisis happens in a vacuum: they are always attached to actors. Therefore, understanding such actors is key to anticipating. By articulating the goal of a system, we can often better understand its behavior and better predict its next moves.