Beyond the classical monoscopic 3D in graph analytics: An experimental study of the impact of stereoscopy

Abstract

Popular in the nineties, 3D visualization has since garnered much criticisms. While historically the vast majority of 3D visualization evaluations have been based on the classical Brunelleschi's perspective rendering, today new questions are raised with the recent access to high quality stereoscopic representations. Do these new interfaces offer improvements over the traditional 3D counterparts? This paper describes a comparative evaluation of monoscopic 3D and stereoscopic 3D for an important problem in graph visualization: the detection of communities. Our results show that stereoscopic 3D outperforms both 2D and monoscopic 3D in the task resolution. The 2D condition always yields the lower response times. Mono-scopic and stereoscopic 3D obtain similar response times. A complementary study based on an original methodology reveals that these apparent similarities actually correspond to different interaction strategies - rotations around the graph. Under the monoscopic condition, participants tend to explore the display space more thoroughly while under the stereoscopic condition, they seem to favor the adoption of a smaller set of viewpoints on the graph.