This study focuses on issues of knowledge representation and elicitation in Intelligent DSS (IDSS) environments. The types, characteristics, levels of logical view, and the levels of specificity and abstraction of "passive" and "active" knowledge in IDSS are discussed. A language for knowledge description, whose syntactical objects are entities, relationships, transformations, and constraints, and which allows four levels of specificity and abstraction is proposed. Then, a graphical, semantic model for the conceptual-schema representation of passive and active knowledge, called the extended ERA Model, is presented. Finally, it is argued that a multi-paradigm programming environment is required for the information-schema representation of the different types of knowledge in IDSS, and to support reasoning, inference, and inheritance. A LOOPS implementation of the knowledge representation and elicitation model is described in detail.