Abstract:
As research in a specific scientific domain advances, the concepts, methods, models and data all undergo changes and refinements. For some of these scientific artifacts we have developed sophisticated tools to manage versioning; for others, we have little or no record of how they are evolving because they are not represented explicitly or are represented as fixed or atemporal objects. Thus, current computational approaches do little to support the dynamic nature of knowledge and ignore the deeper knowledge and understanding lurking within this flux. Such a situation can result in misinterpretation and a shallow understanding of a scientific enquiry, causing conceptual gaps to emerge between knowledge producers and consumers and inhibiting the reusability of knowledge artifacts. In this thesis, we adopt an alternative view of ‘knowledge as process’ and propose a computational model of science where knowledge representation can be seen as an ongoing progression by connecting the products of science with the process of science. We develop a model of category evolution that incorporates representation of (i) the different facets of a category and changes in these facets, (ii) the exploratory and dynamic process of the construction and evolution of categories, and (iii) the contested meanings or multiple perspectives towards categories. The model further incorporates a change language to measure and analyze the changes in different facets, and to compare different categories from the same or different taxonomy. We then present AdvoCate (Adventures of Categories), an eScience tool, that incorporates the evolution model, supporting the process of modelling categories and their changes, while maintaining a versioning system that captures not only the different versions of categories, but also the exploration and evolution steps taken during their creation. The exploration process reveals helpful information regarding how and why changes were made, allowing a knowledge consumer to better understand the biases and decisions made during the process and the reasons behind them. We demonstrate the usefulness of this deeper representation using examples of category evolution from a land cover mapping exercise. The use cases demonstrate how a richer model of category, grounded in the process of science, supports a deeper understanding of scientific knowledge and the process of enquiry, and surfaces up the knowledge hidden within the process of science to support and improve the enquiry process.