AbstractShifting the emphasis from a model to a modeling task, which involves both a computer model and a modeler, we ask what makes a problem complex. We propose that a modeling task can be seen as a set of questions‐and‐answers, nested at multiple levels. The role of the modeler then lies in posing the questions and choosing the best procedure to answer them, while the role of the model lies in answering the questions, via algorithmic, thus logically simple, procedures. Within this framework, complexity is broadly related to the number of question‐answer levels involved in the process and the nature of the questions posed. Addressing this complexity depends crucially on the ingenuity and creativity of the modeler. This may lead to a view of complexity, which is no more observer‐independent, but rather accounts for both historical and cultural development, that is the context of the problem at hand. © 2008 Wiley Periodicals, Inc. Complexity, 2008.