Using Theories of Decision-Making Under Uncertainty to Improve Data Visualization

Explore theories of decision-making under uncertainty to enhance data visualization techniques in this Law & Society Fellow Talk by Jessica Hullman from Northwestern University. Delve into the challenges of designing robust visualizations for data-driven inference and the limitations of current research methodologies. Examine recent work at the intersection of visualization and theory, addressing issues in visual data analysis, data communication, privacy budget setting, and responsive design. Learn about innovative approaches to measuring visualization value in exploratory data analysis and communication. Discover how theorizing reasoning under uncertainty, mediated by data representations, could transform research and practice in the field. Gain insights into topics such as perceptual accuracy, pattern finding, effect size judgment, and the challenges of learning from experiments. Understand the importance of defining decision problems in visualization design and explore applications in aggregation choices and interface evaluation.

Intro
data summaries for inductive inference
objective: perceptual accuracy?
good perception = rational judgment?
objective: pattern finding
optimizing for pattern finding encourages NHST?
minimize error in effect size judgment/decisions
non-robust strategies → illusion of predictability
The distance heuristic
when optimizing for PoS isn't enough...
challenges in learning from experiments
when does a better visualization matter?
defining a decision problem (from Kale et al. 2020)
dead in the water (Gelman and Weakliem 2009)
post-experiment: rank behavioral agents with vis
post-experiment: rank heuristics
design applications: aggregation choices
what characterizes a good interfaces problem?