Abstract

This project develops a view of data models and the modeling of data in science. It begins by examining the history of the concept of a data model. It then discusses an important property of data models, that they are not exact. They are dissimilar to their target in certain relevant respects. The examination then turns to the question of why scientists accept certain data models but reject others. One conclusion is that human interests must be taken into account before the relevant similarities and dissimilarities between the model and its target can be determined. This project focuses on the extent to which theoretical interest guide the scientists decisions about what counts as an acceptable or unacceptable data model. ;Next the project examines artifacts and data models. By specifying a data model scientists have identified certain similarities and dissimilarities between the data model and its target. Some of these dissimilarities are termed artifacts. Scientists will expend time and effort to identify and remove artifacts in their data models. A categorization of artifacts based on theoretical interests is presented. ;Finally, confocal microscopy is presented as a detailed illustration of data models and their use. It is shown that the treatment of artifacts varies from assay to assay. Although each of the assays described produces a confocal micrograph, artifacts that are considered problematic in one assay are considered unproblematic in the next. Second, there are multiple levels of data models in cell biology. The levels are not always clearly defined, as it is often hard to tell when data model processing results in a new data model. A major point here is that confocal microscopes do not produce "raw", or uninterpreted data. Finally, it is shown that data model processing often depends on theoretical assumptions and theoretical goals of the scientists