Bernd Fischer, Michael Lowry, and John Penix
Effective application of AI techniques requires an understanding of how the representation vs. reasoning tradeoffs impact the domain of interest. In this paper we evaluate these tradeoffs for the software engineering problem of automatically retrieving components from a library. We generalize our experience in building several automated component retrieval and automated program synthesis systems. We then develop a framework in which we can compare the tradeoffs taken in the various approaches and speculate as to how to effectively combine aspects of the approaches for future applications.