Testing the Limits of the Spatial Approach: Comparing Retrieval and Revisitation Performance of Spatial and Paged Data Organizations for Large Item Sets

Carl Gutwin, Michael van der Kamp, Jeremy Storring, Andy Cockburn, Cody Phillips

Abstract: Finding and revisiting objects in visual content collections is common in many analytics tasks. For large collections, filters are often used to reduce the number of items shown, but many systems generate a new ordering of the items for every filter update – and these changes make it difficult for users to remember the locations of important items. An alternative is to show the entire dataset in a spatially-stable layout, and show filter results with highlighting. The spatial approach has been shown to work well with small datasets, but little is known about how spatial memory scales to tasks with hundreds of items. To investigate the scalability of spatial presentations, we carried out a study comparing finding and re-finding performance with two data organizations: pages of items that re-generate item ordering with each filter change, and a spatially-stable organization that presents all 700 items at once. We found that although overall times were similar, the spatial interface was faster for revisitation, and participants used fewer filters than in the paged interface as they gained familiarity with the data. Our results add to previous work by showing that spatial interfaces can work well with datasets of hundreds of items, and that they better support a transition to fast revisitation using spatial memory.

TL;DR: We compare retrieval and revisitation performance between spatially-stable organizations of items and traditional paged interfaces, for datasets that are much larger than what has been tested before.

Keywords: Filtering, spatial memory, revisitation