Generating Predictions to Aid the Scientific Discovery Process

Proceedings of the AAAI Conference on Artificial Intelligence, 5

Volume

Science

Learning

NGLAUBER is a system which models the scientific discovery of qualitative empirical laws. As such, it falls into the category of scientific discovery systems. However, NGLAUBER can also be viewed as a conceptual clustering system since it forms classes of objects and characterizes these classes. NGLAUBER differs from existing scientific discovery and conceptual clustering systems in a number of ways. I. It uses an incremental method to group objects into classes. 2. These classes are formed based on the relationships between objects rather than just the attributes of objects. 3 The system describes the relationships between classes rather than simply describing the classes. 4. Most importantly, NGLAUBER proposes experiments by predicting future data. The experiments help the system guide itself through the search for regularities in the data.