Understanding and Supporting Trade-offs in the Design of Visualizations for Communication.

Information Visualization;Data Storytelling;Uncertainty Visualization;Computer Science;Engineering;Information

Hullman, Jessica RuthFinholt, Thomas A. ;

University of Michigan

关键词: Information Visualization;    Data Storytelling;    Uncertainty Visualization;    Computer Science;    Engineering;    Information;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/107170/jhullman_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

A shift in the availability of usable tools and public data has prompted mass manufacturing of information visualizations to communicate data insights to broad audiences. Despite available software, professional and novice creators of visualizations that are intended to communicate data insights to broad audiences may struggle to balance conflicting considerations in design. Studying professional practice suggests that expert visualization designers and analysts negotiate difficult design trade-offs in creating customized visualizations, many of which involve deciding how and how much data to present given a priori design goals. This dissertation presents three studies that demonstrate how studying expert visual design and data modeling practice can advance visualization design tools. Insights from these formative studies inform the development of specific frameworks and algorithms.The first study addresses the often ignored, persuasive dimension of narrative visualizations. The framework I propose characterizes the persuasive dimension of visualization design by providing empirical evidence of several classes of rhetorical design strategies that trade-off comprehensive, unbiased data presentation goals with intentions to persuade users toward intended interpretations. The rhetorical visualization framework highlights a second trade-off: the act of dividing and sequencing information from a multivariate data set in separate visualizations for ordered presentation. I contribute initial evidence of ordering principles that designers apply to ease comprehension and support storytelling goals with a visualization presentation. The principles are used in developing a novel algorithmic approach to supporting designers of visualizations in making decisions related to visualization presentation order and structuring, highlighting the importance of optimizing for both local or ;;single visualization” design in tandem with global ;;sequence” design.The final design trade-off concerns how to convey uncertainty to end-users in order to support accurate conclusions despite diverse educational backgrounds. I demonstrate how non-statistician end-users can produce more cautious and at times more accurate estimates of the reliability of data patterns through the use of a comparative sample plots method motivated by statistical resampling approaches to modeling uncertainty.Taken together, my results deepen understanding of the act of designing visualizations for potentially diverse online audiences, and provide tools to support more effective design.