Influence of Presenting Uncertainty Information on the Evaluation of Watershed Plans by Users of an Automation-Assisted Environmental Decision Support System

Abstract

An important aspect of human-computer interfaces used to support automation-aided decision-making is the communication of uncertainty. However, people may misinterpret or ignore uncertainty information. We investigated how evaluations of suggested conservation plans in an environmental decision support system were influenced by visualizations displayed with and without uncertainty. We also examined whether test-users can be utilized to assess decision support systems (e.g., when end-user testing is not feasible), and whether users’ statistical backgrounds influenced decision making. Levels of uncertainty were manipulated via a novel Asymmetric Uncertainty Metric. Seventy-seven participants (53 test-users, 24 end-users) evaluated conservation plans in two counterbalanced online sessions, one presented with uncertainty information and one presented without. Linear regression results indicated that evaluations made with uncertainty differed from those without (Λ = 0.663, p < .001), suggesting that the presentation of uncertainty information in decision support systems can modify decisions about conservation plans. End-users were more influenced by the inclusion of uncertainty information than test-users, (b _end-users = 0.22–0.28, b _test-users = 0.08–0.12) (p < .001), and individuals who used (p = .016) or had been trained to use (p < .001) statistical data were also more sensitive to the inclusion of uncertainty information (b _wsb = 0.17–0.18) than those without (b _wnsb = 0.07–0.11).