A hybrid representation of architectural precedents

We present a hybrid representation of architectural precedents that separates precedent instances from the concepts they embody, where the concepts are defined in terms of multiple classification taxonomies. The representation allows us to combine the classical view of concept acquisition with aspects of the probabilistic and exemplar views and to organize the database into the equivalents of “episodic” and “semantic” memory. A first application and test context is provided by the Software Environment to Support Early Building Design (SEED), where precedents are used as prototypes and cases. But the representation is flexible enough to support the use of precedents in other application contexts.