Object-oriented database system has become increasingly popular and influential in the development of new generation database systems. This motivates the investigation of mechanisms for data mining in object-oriented databases. In this paper, we propose our first step towards knowledge discovery in object-oriented databases by extension of the attribute-oriented induction technique from relational databases to object-oriented databases. By the development of sophisticated generalization operators and generalization control mechanisms, attribute-oriented induction method can be successfully extended to knowledge discovery in object-oriented databases. Furthermore, we show that knowledge discovery will substantially enhance the power and flexibility of querying data and knowledge in object-oriented databases.