Monday, June 3, 2019

Visualizing uncertainty

The theme of this year's VizUM, the University of Miami's visualization symposium, is the role of uncertainty in visualization. The date is Tuesday, November 12, and registration is free. I'll present about How Charts Lie and the comprehension gap between designer and reader:
Scientists, statisticians, designers, and journalists are often taught that “a picture is worth a thousand words”, that we should “show, don’t tell”, and that charts are “intuitive” and useful to “simplify” information. This talk explains why these myths, if taken at face value, are wrong and dangerous, and what we can do to help the public understand charts, graphs, maps, and infographics better. Contrary to what we commonly hear, there is always an uncertain gap between what we intend to communicate and what our readers end up seeing; bridging that gap should become a priority to any visualization designer.

Our star speaker, Northwestern University's Jessica Hullman, will talk about “Supporting Reasoning with Uncertainty Using Data Visualization”:
Charts, graphs, and other information visualizations amplify cognition by enabling users to visually perceive trends and differences in quantitative data. While guidelines dictate how to choose visual encodings and metaphors to support accurate perception, it is less obvious how to design visualizations that encourage rational decisions and inference. Jessica will motivate several challenges that must be overcome to support effective reasoning with visualizations. 
First, people’s intuitions about uncertainty often conflict with statistical definitions. Jessica will describe how visualization techniques for conveying uncertainty through discrete samples can improve non-experts’ ability to understand and make decisions from distributional information. 
Second, people often bring prior beliefs and expectations about data-driven phenomena to their interactions with data (e.g., I suspect support for candidate A is higher than reported), which influence their interpretations. Most design and evaluation techniques do not account for these influences. Jessica will describe what’s been learned by developing visualization interfaces that encourage users to reflect on their expectations and use them to predict and improve belief updating.

Join us in November, when Miami is sunny and warm (as usual).