The paper presents a cognitive analysis of design as a problem-solving activity. Design is characterized through results from empirical studies of various design tasks. Based on this characterization, it will be argued why and how knowledge modeling can be considered as a design task. Knowledge modeling has been qualified as the "construction of a set of successive representations". Processing different types of representations at different levels and from different viewpoints plays an important role in design problem solving. It will be argued that other characteristics of design may also apply to knowledge modeling. As there are not yet data available on the cognitive activities involved in knowledge modeling, hypotheses for studying these activities may be inspired by what is known about design: design problems have no pre-existing, complete problem definition, and no unique, correct solution; the design activity does not follow a pre-existing plan, it is opportunistically organized.