Skip main navigation

Heuristics and dynamic systems

The representativeness heuristic dominates needed prior knowledge. This is particularly problematic in dynamic systems where prior knowledge is weak.
Increasing share prices with small fluctuations around trend.
© Moxnes

In the preceding step, you probably started to expect that heuristics and biases matter for the management of dynamic systems. The answer is yes, most of all because dynamic systems are complex and complicated to deal with.

The geocentric world view was based on observation of planetary movements seen from the Earth, which was assumed to be in the center of the universe. The observations were used to make predictions of how the planets would behave in the near future. Thus, they relied on an assumption that history would repeat itself, which is not a bad idea for planets. The model, however, became very complicated. Picture of a physical model that shows planetary movements, from the Science museum in Firenze. Picture of a physical model that shows planetary movements, from the Science museum in Florence

The historical development of the heliocentric world view illustrates complexity and the transition from observing to modelling. Already Plato observed that Mercury and Venus always stayed close to the sun. In 1543, Copernicus put the sun in the center of circling planets. In 1609 Kepler observed that the orbit of Mars was elliptic. This entire development from Plato to Kepler was based on observations.

However, there was no prior information that could explain why planets moved in elliptic orbits around the sun before Isaac Newton published his Principia in 1687. By formulating and analyzing a mathematical model, he was able to explain why the planets circle the sun in elliptical orbits. In other words, a very simple model provided the final proof of a new world view.

Now take a look at the above graph for the development of the price of shares for the Ford Motor Company. Imagine that you are as clever as Kepler, what pattern do you see? Imagine that you need a forecast of the price development, what do you expect that the price will be in January 2023?

If you assume that the historical development will repeat itself, the price is likely to exceed $30 per share. However, it is far more complicated to predict share prices from historical trends, than to predict the steady movements of planets. Still, people do that. Even Newton himself lost much of his family fortune after speculating in a continued increase in the share price of the South Sea Company. A tremendous increase in the South Sea share price was followed by a collapse. Newton commented: “I can calculate the motions of heavenly bodies, but I cannot understand the madness of men.”

The share price of the Ford Motor Company fell to $13 per share by January 2023. Hence pattern matching followed by assuming that history will repeat itself, would have led to investments and a subsequent loss. A more proper problem definition would involve prior information of importance for the fundamental value of the company, uncertainty and risk. A model that explains why market bubbles occur and how they develop, would help improve your risk perception. However, I would lie if I told you that after this course, you will be able to build dynamic models of share prices that will make you rich. This is a problem far more complex than the geography of Europe.

However, I can warn you about methods that promise too much. In general, predicting is very difficult, even though all future events are not totally, unpredictable “black swans”. Prediction is difficult because there are so many unknows as well as unknown unknows. So similar to Kepler, those who make predictions rely on the assumption that historical patterns will repeat themselves. In support of this claim, John Sterman (1988) found that expert predictions of inflation and of future energy demand were both well explained as extrapolations of historical trends. Other studies also find trend extrapolation. By being aware of and understanding the causes of bubbles and cycles, you are warned about the danger of trend extrapolation.

To summarize, humans are eager to observe and are very good at recognizing patterns in observations. Since observations of changes over time are difficult to understand, a simple strategy is to assume that history will repeat itself. This requires a wait-and-see strategy to observe trends, before trend extrapolations can be made.

The complexity of ecologic-social-economic systems makes it difficult to make reliable predictions. Does this meand that there is little to gain from making and using simulation models. However, the main purpose of systems thinking and modelling is not to make predictions. Recall from the P’HAPI recipe that the purpose is to understand problems and to test policies. This is easier and more reliable than making predictions. For example, Sterman (1988) did not need a very complicated model to demonstrate the biases that follow from using the method of trend extrapolations to make forecasts; to show that trend extrapolation is used and does not work well in many systems.

Reference: John Sterman: “Modeling the Formation of Expectations: The History of Energy Demand Forecasts”, International Journal of Forecasting, 1988, Vol 4, 243-259.

This article is from the free online

Reaching UN Sustainable Development Goals (SDGs) through Systems Thinking

Created by
FutureLearn - Learning For Life

Reach your personal and professional goals

Unlock access to hundreds of expert online courses and degrees from top universities and educators to gain accredited qualifications and professional CV-building certificates.

Join over 18 million learners to launch, switch or build upon your career, all at your own pace, across a wide range of topic areas.

Start Learning now