Geovisual analytics and the science of interaction: an empirical interaction study

Abstract

Among the most pressing research and development challenges facing geovisual analytics is the establishment of a science of interaction to inform the design of visual interfaces to computational methods. The most promising work on interaction to date has attempted to identify and articulate the fundamental interaction primitives that define the complete design space for the user experience. In this paper, we take the logical next step beyond this prior research, reporting on a controlled interaction study to learn how variation in interaction primitive combinations impacts broader interaction strategies (i.e., to learn how interaction primitives relate for both design and use). GeoVISTA CrimeViz – a geovisual analytics application developed in partnership with the Harrisburg (Pennsylvania, USA) Bureau of Police – was leveraged as a living laboratory for examining the nature of interaction strategies as they are built from interaction primitives. Ten law enforcement officers with the Harrisburg Bureau of Police completed a set of 15 benchmark tasks based on a three-stage interaction primitive taxonomy while their interactions were logged. Experimental results revealed several noteworthy characteristics of the relationship between interaction primitives and interaction strategies, including an increased reliance on the interface as the objective increases in sophistication and the effectiveness of, although at times over-reliance upon, Shneiderman’s visual information-seeking mantra as an analytical strategy. Further, consistently successful and suboptimal interaction strategies were characterized in terms of their constituent interaction primitives and articulated as user personas, allowing for the establishment of interface design and use recommendations for circumventing negative personas.