VOTE, written in T [Slade, 1987], comprises over 14,000 lines of code, and over 9,000 lines of data. The VOTE program relies on an object-oriented database system for managing the underlying knowledge used in decision making. In this appendix, we discuss that database system, and each of the basic knowledge structures used by VOTE: issues, groups, members of Congress, bills, and strategies. In the previous appendix we described the natural language generation mechanism used by VOTE.
One of the technical challenges facing artificial intelligence is coping with the complexity of large AI programs. As AI models and theories get more detailed, the corresponding programs get larger. As programs get bigger, they become harder ...