Incompleteness, error, approximation, and uncertainty: An ontological approach to data quality

Ontology for geographic information is assumed to contribute to the design of GIS and to improve usability. Most contributions consider an ideal world where information is complete and without error. This article investigates the effects of incompleteness, error, approximation, and uncertainty in geographic information on the design of a GIS restricted to description of physical reality. The discussion is organized around ontological commitments, first listing the standard assumptions for a realist approach to the design of an information system and then investigating the effects of the limitations in observation methods and the necessary incompleteness of information. The major contribution of the article is to replace the not-testable definition of data quality as "corresponding to reality" by an operational definition of data quality with respect to a decision. I argue that error, uncertainty, and incompleteness are necessary and important aspects of how humans organized and use their knowledge; it is recommended to take them into account when designing and using GIS.