Purchase this article with an account.
Alexis V. Soto, Melissa A. Schoenlein, Karen B. Schloss; More of what: does the dark-is-more bias for colormap data visualizations operate on numeric or conceptual magnitude?. Journal of Vision 2021;21(9):1960. doi: https://doi.org/10.1167/jov.21.9.1960.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When interpreting colormap data visualizations, people infer that darker colors map to larger quantities (dark-is-more bias). In the colomaps tested in previous studies, numeric magnitude always corresponded to concept magnitude (i.e., # of animals corresponded to “greater” animal sightings) (Schloss et. al., 2019; Sibrel et al., 2020). However, conflicts can arise in which larger numeric magnitude corresponds to smaller concept magnitude. For example, if number of seconds is used to represent the concept of speed, then larger numbers correspond to less speed. Under such conflicts, does the dark-is-more bias operate at the numeric or conceptual level? In Experiment 1, participants interpreted colormaps representing fictitious data on the time alien animals took to notice a scientist observing them. The “congruent group” (n=30) was told the colormaps represented duration, with larger numbers on the legend labeled “longer” (more duration) and smaller numbers labeled “shorter” (less duration). The “incongruent group” (n=31) was told the colormaps represented speed, with larger numbers labeled “slower” (less speed) and smaller numbers labeled “faster” (more speed). Participants indicated whether the time was longer/faster for animals represented on the left or right side of the colormap. The design included 2 numeric encoded mappings (dark-more/light-more) x 2 color scales (Hot/Colorbrewer-blue) x 2 legend label positions (more-high/more-low) x 20 repetitions (320 trials). The congruent group showed a dark-is-more bias, with faster response times for dark-more numeric encoding, but the incongruent group showed the opposite effect (interaction: p<.001). This suggests that the dark-is-more bias operates at the conceptual level. Experiment 2 was like Experiment 1, except without concept labels on the legend. Here, both groups were faster for dark-more numeric encoding (p<.001; no interaction p=.093). Thus, interference from the conceptual level may be confined to cases where concepts are explicit in visualizations.
This PDF is available to Subscribers Only