Jessica Witt¹ (), Amelisa Warden¹, Danielle Szafir²; ¹Colorado State University, ²University of Norh Carolina - Chapel Hill

Choropleth maps are a critical tool for communicating geospatial data, such as the spread of an infectious disease. How effectively a map communicates information depends on its design. Insights concerning map design principles can be gleaned from research on ensemble perception—the visual system’s ability to extract summary statistics from a scene, such as the mean number of disease cases. Ensemble perception suggests that showing more data points on a map and encoding data with colors that minimize variability may improve people’s ability to estimate statistical information from maps. We tested these recommendations and the effects of the size of geographic regions and data semantics to understand ensemble perception processes when viewing maps. People made yes/no judgements about whether the mean number of disease cases on a map exceeded a critical threshold. We found that people were better at estimating the mean of the map when presented with more data points (i.e., individual counties instead of regions) and lower variability across data points. That aggregating data into larger regions did not improve mean estimation suggests that while ensemble perception processes may bear some resemblance to statistical processes, they are incapable of performing the visual equivalent of a meta-analysis. We also found that the perception of ensembles in maps was influenced by the content of the map: people were more sensitive to the map’s mean when the map depicted the number of people with an infectious disease rather than when the map depicted the number of people with immunity to the disease. This reveals a top-down effect on ensemble processing. Our results have implications for theories of ensemble perception as well as for design recommendations for choropleth maps. Understanding how people interpret and derive meaning from geospatial data informs design guidelines for effectively communicating critical information to the public.

Acknowledgements: NSF BCS-1632222; NSF SES-2030059