Perceptual + Cognitive Biases in Data Visualizations

Your visual system can crunch vast arrays of numbers at a glance, providing the rest of your brain with critical values, statistics, and patterns needed to make decisions about your data. But that process can be derailed by biases at both the perceptual and cognitive levels. I'll demonstrate 3 instances of these biases that obstruct effective data communication. First, in the most frequently used graphs - lines and bars - reproductions of average values are systematically underestimated for lines, and overestimated for bars. Second, when people see correlational data, they often mistakenly also see a causal relationship. I'll show how this error can be even worse for some types of graphs. Third, we've all experienced being overwhelmed by a confusing visualization. This may happen because the designer—an expert in the topic—thinks that you'd see what they see. I'll describe a replication of this real-world phenomenon in the lab, showing that, when communicating patterns in data to others, it is tough for people to see a visualization from a naive perspective. I'll discuss why these biases happen in our brains, and prescribe ways to design visualizations to mitigate them.

Cindy Xiong
Speaker Title
PhD Candidate
Speaker Institution
Northwestern University
Speaker Biography

Cindy Xiong is a PhD Candidate at the Department of Psychology at Northwestern University, researching perception, cognition, and decision-making in data visualizations. She focuses on biases across these levels: illusions, misunderstandings, and irrational choices. Her research has been funded by a Northwestern Cognitive Science Fellowship and a Design Research Fellowship. She is currently completing a UXR internship at the MathWorks Image Processing and Graphics. She is one of the founding leaders of VISxVISION (, a partnership dedicated to increasing collaboration among visualization researchers and perceptual + cognitive psychologists.