Reasoning from imperfect knowledge

One reason why there is a lack of cross-references between articles on knowledge representation in the Cognitive and the Information Sciences is that cognitive scientists are interested in descriptive models of how people reason whereas information scientists are interested in prescriptive models to help people reason. Formal ontologies such as the Suggested Upper Merged Ontology aid human reasoning by providing (1) an accurate knowledge base, (2) a formalization of the knowledge base as axioms, and (3) a logic to derive new information through deductive reasoning. However, all systems confront obstacles when reasoning from imperfect knowledge consisting of ambiguous, conditional, contradictory, fragmented, inert, misclassified, or uncertain knowledge. We use work from the Cognitive and the Information Sciences to analyze obstacles for both computers and people when confronted with ambiguous, contradictory, misclassified, and uncertain knowledge. A concluding section considers both practical and theoretical applications of ontologies in the Cognitive Sciences.