Visualization of Uncertainty Aids Spatial Judgments but Fails to Improve Metacognitive Efficiency

Abstract

Uncertainty is an element of many decision-making tasks and inherently compromises performance. Research has found only equivocal evidence that uncertainty representations—displays that explicitly denote data quality—offset the performance costs of uncertainty. As yet, though, no work has examined the potential benefits of uncertainty displays to metacognition, display readers’ ability to assess the quality of their own decision-making processes. The current study examined the benefits of uncertainty visualization to first-order (Type 1) and metacognitive (Type 2) sensitivity in a spatial judgment task. Data revealed only small improvements in Type 1 and Type 2 sensitivity with visualized uncertainty displays, and gave no evidence of disproportionate gains to metacognition.

A preliminary classification of different tasks that make use of uncertainty visualizations

“…Decision-making is the process of identifying and choosing among competing courses of action (Padilla et al, 2018). Recent studies have focused on the use of uncertainty information to support decision making, and several papers have shown that decision-making accuracy can be improved when individuals are provided with uncertainty information compared to when they are not (Burton et al, 2017;Fernandes et al, 2018;Nadav-Greenberg & Joslyn, 2009). Furthermore, Joslyn and LeClerc (2012) found that providing explicit advice about which outcome to choose does not necessarily result in better decision-making performance than providing uncertainty information.…”

“…Metacognition enables individuals to assess the quality of their own cognition, regulate it effectively, and learn the correct approach for their future performances (Burton et al, 2017). A review of literature showed three categories of metacognitive tasks that can be applied to uncertainty visualizations: Self-Monitoring, Metacognitive Control, and Self-Regulated Learning (Schraw & Moshman, 1995;Winne & Azevedo, 2014).…”

“…One method to mitigate the effects of uncertainty is by providing graphical representations of probabilistic information about an uncertain variable (i.e., uncertainty visualizations). Uncertainty visualizations are effective across many different domains and tasks, especially when compared to not having any uncertainty information (Burton et al, 2017;Fernandes et al, 2018;Nadav-Greenberg & Joslyn, 2009;Pugh et al, 2018). Yet, the literature has shown that uncertainty visualizations differ in the extent they improve individuals' task performance, and there is no agreement on what factors cause these differences (Nadav-Greenberg et al, 2008;Pugh et al, 2018).…”

Visualizing Data to Support Judgement, Inference, and Decision Making in Learning Analytics: Insights from Cognitive Psychology and Visualization Science

“…More recently, Matejka and Fitzmaurice (2017a) illustrated this same point compellingly with a more sophisticated simulated annealing technique. 2 They further demonstrated that depending on the type of visualization, the representation of data dispersion and uncertainty may be inherently occluded. As illustrated in Figure 2, visualizations with similar statistical properties can look very different when visualizing the underlying data distributions.…”

“…Visualize uncertainty to support inference-making 5. Visualizing uncertainty is a useful aid for researchers to make statistical and inferential decisions 2,7 Visualize data in ways that enable you to see and understand the nature of the data dispersion and to test assumptions of the data.…”