CHS: Medium: Collaborative Research: Empirically Validated Perceptual Tasks for Data Visualization

Project: Research project

Project Details


Knowledge occupations require discovering, understanding, comparing, and communicating patterns in data, maps, and diagrams. Those patterns are ubiquitously transmitted from technology to humans via visual displays. Dynamic displays are increasingly used to effectively leverage the power of the human visual system to show a process unfold, translate among multiple views, or respond to interactive queries. While such displays should, in theory, be the most efficient way to transfer information from technological systems to human minds, they instead tend to overwhelm, confuse, or misinform. We propose to refine a perceptual model of why and when visual processing works or fails in dynamic displays, and to create guidelines that allow designers to construct efficient displays for users and trainees.

Intellectual Merit
This use-inspired basic research will extend existing knowledge in perceptual psychology and cognitive science, through close collaboration with a diverse set of domain specialists, across data visualization, geospatial visualization, education, data journalism, business analytics, and security analytics. The novel empirical work is organized into five sets of experiments.
Set 1: What kinds of abstract information do designers seek to communicate with dynamic displays, and what types of displays do they choose? What works, and what fails, when empirically tested?
Set 2: Our visual system computes image statistics, and this ability can be leveraged to compute statistics in static data displays – we will explore what statistics people can compute in dynamic displays.
Set 3: Dynamic displays can facilitate comparisons between datasets, encoding their different directly as a motion signal. We will compare performance in dynamic displays to static, across a varied of tasks.
Set 4: Dynamic changes to displays can transition to new contexts for data, such as a schematic diagram to a 3D model. Because these changes tend to overwhelm the viewer, these experiments will produce guidelines for the power and limits of visual understanding of these transformations.
Set 5: Selective attention is critical for dynamic displays, radically changing what information is encoded. We will extend current salience models to predict attention in dynamic displays.

Broader Impacts
The proposed work will generate a new understanding of how we have adapted our visual system to extract information from artificially constructed displays, and would produce guidelines that would impact designers of data visualizations, data journalism stories, instructional interventions, systems for business analysts, and analytics displays for security. Our laboratory has a strong record of cross disciplinary data visualization research, and the PI serves as a a Papers Co-chair for IEEE Information Visualization. The PI teaches dozens of courses and workshops each year on data visualization and science communication to research and business communities. Guidelines will be transmitted to the US workforce via social media accounts of data visualization authors, authors of workforce-oriented data visualization textbooks, and a lay-friendly book by the PI. Our laboratory is committed to recruitment of under-represented groups, from college admissions to graduate student levels.
Effective start/end date10/1/199/30/23


  • National Science Foundation (IIS-1901485)

Fingerprint Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.