RSVP Please complete this form by EOD Feb 4th if you’d like food.
Monday, Feb 6th
12:30-1:30 PM
Speaker: Andrew McNutt
Room: JCL 298
Abstract
Data-driven tasks, such as data analysis or making visualizations, are a growing part of everyday life for countless people. Despite decades of consideration, these tasks remain challenging. Navigating the complex and interleaved steps to manipulate, comprehend, and present data, while considering dozens of usage guidelines, often yields unsafe error-prone results. Failures include charts that yield misinformation or even create analytical errors that shake the world (such as by worsening the great recession). Common solutions tend to emphasize automated recommendations (which can give way to automation biases) or push for data literacy (placing the burden of work on the viewer).
In this talk, I propose a middle ground that keeps the human in the loop, in a manner that minimizes biases and makes it easier and safer to work with data interfaces. I will describe a suite of methods for automatically validating visualizations that synthesizes approaches from software engineering (like metamorphic testing and linting) in an easy-to-interpret manner. To support these analyses, I develop a formalism that allows best practices for unknown chart forms to be derived, so that such validation processes can be generated automatically. The pioneering premise of this work is that validation is an essential part of the visual analytics process, and is crucial for a future in which data tasks are accessible to anyone of any skill level.
Bio
Andrew McNutt is a Ph.D. candidate in Computer Science at the University of Chicago. He builds systems and theories for visualization, data analysis, and programming more generally. His work operationalizes domain knowledge (such as for making charts) into software systems and interfaces that are approachable, trustworthy, and respectful of the user’s agency. Andrew’s work regularly appears at top venues in Visualization and Human-Computer Interaction venues, such as IEEE VIS and ACM CHI, where he has won multiple honorable mention awards. He once accidentally caused a giant snowball to rip through the side of an apartment building, an event which ended up on the national news.