| Abstract: | We subject a basic qualitative modelwhich appears throughout the qualitativereasoning literature (the ``bathtub' or liquidtank model) to a detailed theoretical analysisof its representation properties. We show thatthe standard model for this family of systemsdoes not cover the intuitive concept ofreal-world tanks, in that there are both simpletanks that do not obey the model, and thatthere are physically impossible shapes that doobey it and get considered by qualitativereasoners using the model. We demonstrate thatthese modeling problems may lead to a markeddecrease in the usefulness of employingqualitative reasoners in some cases. Weconclude that one should be careful aboutmaking both the assumptions required by themodel and the algorithm, and the family ofsystems that are actually reasoned about,explicit in the presentation of qualitativereasoners' output. |
