Describes the preliminary design of a Rule-Oriented System for Implementing Expertise (ROSIE). This system will serve as a tool for model builders seeking to apply expert knowledge to the analysis of problems and the evaluation of solutions in complex domains, especially domains for which useful analytic models are unavailable. Models built within the ROSIE system constitute simulations in the application domain. Control in these models is data-directed. That is, actions are specified by sets of rules having the basic form: "IF (conditions) THEN (actions)." The actions are executed when data matching the conditions appear in the model's data base. Major features of ROSIE include hierarchic structures on data elements and rules, user selection of rule-invocation strategies, user control of rule iteration, user control of rule and data activation, and sophisticated user support tools.