Driving problems: using data validation debugging confirming hypotheses monitoring exploration exposition few/no a priori ideas illustrate known things for others "overview, zoom and filter, details on demand" infovis mantra, Shneiderman 1996 6