| dc.description.abstract | The aim of this research is to compare the effectiveness of several visual representations of statistical data uncertainty and to explore the influence of individual differences among users. A large body of work shows that providing uncertainty information is beneficial for decision-making. However, these advantages of showing uncertainty critically depend on how it is communicated. This large online user study (n=245) identifies how the quality of probability estimates compares across six visualizations and across users. Participants were presented six visualization types that each encode a probability distribution that represents a possible range of arrival times, predicted by a car navigation system. They were asked to report best estimate and two kinds of probability estimates (later than and range). Probability estimate accuracy and precision were compared across visualizations, question type and user types based on ten characteristics, among which both cognitive measures and personality traits. An ANOVA showed a main effect of visualization type, where discrete plots with few outcomes result in the most accurate and precise probability estimates and prove to perform best across all user classifications. Although task performance differs across users with different levels of cognitive abilities and personality traits, it can be concluded that visualization type has a much greater impact on performance than individual differences have. This suggests that, when designing an interface with an aim for high performance, it is more effective to focus on the graphic design of a chart than on personalization. The acquired knowledge contributes to the standardization of including uncertainty measures into information visualizations and to the development of user adaptive visualizations. | |
