The BOOLE2 system described here presents a computational model of Boole’s discovery of logic as a part of mathematics. The authors also provide a historical account of this discovery. BOOLE2 tries to reproduce both the process of discovering logic as a branch of abstract algebra and Boole’s account of the process that led to it. The system’s discovery methods are also tested in three other cases: two versions of a subset of Gregory’s geometry, and the first principles of differential calculus.

BOOLE2 models the given science using a frame-based representation for concepts and a production system for actions. The system starts with the knowledge and goals that were known before researchers discovered whether a science is symbolizable, and it decides whether the discovered laws make it possible to express that science in symbolic terms. Unlike other discovery systems, BOOLE2 is exclusively guided by theory instead of experimentation.

The proposed approach is a step toward a theory of scientific discovery. The presentation is clear and concise. The authors include an appendix showing system traces. The paper should interest those working in knowledge representation, knowledge discovery, and the history of science.