Modelling to Predict or to Learn?

Many in regional scenario planning are aware, on some level, that scenario planning made its way to urban planners through the business and strategic planning literature. In an article from that business world called Modelling to Predict or to Learn? A.P. de Geus addresses the issue of prediction within scenarios. Specifically, he asks whether scenarios might be more useful as a tool for learning about the connections between issues and how to adapt in the face of change.

The article is a quick read, and worth it for the anecdotes about the famous Royal Dutch Shell scenario-building process alone. But I think anyone involved in scenarios, a world in which we’re often asked to predict “what will happen if we do X” with increasing precision, will find value in it. Below are a few of them, but I really do recommend the entire article:

On what scenarios were often intended to do vs. what they actually did:

Scenarios aim at perceptions inside the heads of decision makers  obliging them to question their assumptions about how their business world works. The scenarios should lead the managers to change and reorganize their inner models of reality (dixitPierre Wack in the Harvard Business ReviewSept.-Oct. 1985). As a result, these decision makers will take decisions different from the ones which they would have taken, had they not been exposed to the scenarios. But, how to achieve this ‘aha’ experience in the minds of managers was not quite clear. It was seen as the real challenge of  scenario analysis and clearly not solely dependent on the eloquence of presentation and the beauty of the charts…there was at best a shaky, indirect connection with the decisions taken.

On what a model would need to make reliable predictions:

For a model to produce reliable predictions of living systems such as companies,markets, national economies, etc., it is necessary that such a model 1) is indeed complete; 2) is a precise representation of reality. Then for a manager to act on such predictions,it is necessary that the model is recognized by him/her as a complete and precise representation of reality.

On the difference between predictive models vs learning models (scenarios):

“[W]e are talking about a quite different category of models. We are no longer talking about the model, the understanding of this world as it has been acquired by a modeller or planner…We are talking about the understanding of [the manager’s] world as it has been acquired by this manager or this management group – however incomplete or deficient their model may be. By computer modeling their world, we give them a’toy’ (a representation of their real world as they understand it) with which they can ‘play’, i.e. with which they can experiment without having to fear the consequences.




Comments are closed.