Extracting actionable insight from large high-dimensional data sets, and its use for more effective decision-making, has become a pervasive problem across many application fields in both research and industry. The objective of our presentation is to report on some investigations of this problem covering both these areas. Taking as the problem domain the area of "unsupervised learning", we show that by tightly coupling statistical analysis technique with combinations of visualization components and techniques for interactivity, real-time analysis of multidimensional data can be efficiently made. We give particular attention to the ways in which dynamic visual representations can be used in these contexts to facilitate shared understanding. Our system is implemented and validated in the context of 3D medical imaging knowledge construction, knowledge management and geovisualisation.