Network visualization has been a lively topic for a half century, but the intense challenges from many facets of this problem demand diverse solutions. While the popular force-directed approaches produce appealing presentations for websites and print, their benefits are limited to showing macro features such as clusters. Interactive approaches that give users control of node and link visibility enable them to make more fine-grained analyses that lead to important insights about relationships among nodes or the presence of exceptional nodes and links. Another important task is to spot the absence of expected nodes and links. One strategy is coordinating network visualizations with statistical measures from graph theory and social network analysis to give users interactive control of ranking, filtering and clustering (http://www.cs.umd.edu/hcil/socialaction). A second strategy involves a novel layout technique to arrange node positions according to their attributes in stable yet comprehensible semantic substrates (http://www.cs.umd.edu/hcil/nvss).
Both strategies were implemented, then evaluated and refined by case study qualitative methods with domain experts (political analysts, healthcare consultants, counter-terrorism experts and bibliometricians) who worked on their own problems during 1 to 6 weeks of observation.