Human representation and reasoning about complex causal systems

Abstract

We characterize what is known about how people represent, reason about, and predict the behavior of complex systems. People tend to simplify complex systems in three ways: First, people resort to heuristics that are selective in the information they consider. These heuristics often yield satisfactory results though they can lead to systematic error. Second, when people do try to take more information into account, they often use a model that has a simple linear form that ignores most of the interactions and sources of unpredictability in the system. Finally, when going beyond heuristics and simple linear combinations, people tend to build a mental causal model that reflects the causal structure of the system by representing qualitative structure relating the mechanisms that lead from causes to effects. The bulk of the chapter concerns the nature of causal models. Although people excel at representing how individual mechanisms work and how they are linked to each other, they tend to neglect cycles of causation, often fail to reason quantitatively, and sometimes ignore relevant variables.