Songwen Hu¹, Alex Endert¹, Cindy Xiong¹; ¹Georgia Institute of Techonology

Forecast visualizations convey uncertain information intended to guide decision-making, yet the way this information is visually encoded can strongly influence how people interpret and trust the data. Some studies find that viewers place greater trust in simple, deterministic displays that emphasize one clear pattern (Elhamdadi et al., 2023), whereas others show that visualizing uncertainty explicitly can foster more calibrated trust and more appropriate decisions under uncertainty (Padilla, Ruginski, & Creem-Regehr, 2017; Witt & Clegg, 2022). Such variability suggests that perceptual and cognitive responses to uncertainty visualization are not uniform across observers but may depend on individual differences in how visual information is processed and integrated into judgment (Grounds & Joslyn, 2018). We drew on human-factors and visualization research to test whether interaction design could reconcile individual differences by allowing viewers to control how much uncertainty information they see. We designed interactive visualizations that either (1) communicated forecast uncertainty through confidence intervals (aggregated uncertainty) or (2) explicitly displayed multiple forecasts (individual model predictions). We compared these with non-interactive versions that fixed the display but contained the same information. Participants viewed humidity forecasts for two cities, made predictions about future trends, and rated their confidence and trust in the visualization. Results showed that participants trusted predictive visualizations and underlying data more when interactions were well designed than when they were static (MD=0.667, p=0.012). Interaction also altered participants’ predictions, leading them to perceive larger differences between the cities’ projected humidity values (MD=5.763, p<0.001), though confidence ratings were unaffected (MD=0.267, p=0.229). These findings demonstrate that interaction modulates how viewers perceive and mentally represent forecast uncertainty. Future work can further examine how interactive uncertainty visualizations shape attentional control and allocation during data foraging to better align interface design with the visual system’s capacities for probabilistic reasoning.