Knowledge-Directed Mediation between Application Objects and Base Data Integration of database and artificial intelligence methods is critical for the next generation of information systems. In this paper, we offer a conceptual framework for such systems. We envision a partitioned architecture where multiple applications residing on distributed workstations exploit multiple remote data resources. Central to the model is an intermediate layer of mediators that perform the knowledge-based task of transforming data into information needed by the workstation applications. To illustrate this notion, we introduce a specific mediator, implemented in the PENGUIN system, that defines an object-based layer on top of relational database systems.