Investigating Representation Alternatives for Communicating Uncertainty to Non-experts

Abstract

Abstract. Non-experts are confronted with uncertainty of predictions everyday when, e.g., using a navigation device or looking at the weather forecast. However, there are no standards for representing uncertain information and representations could be easily misleading. Thus, we selected twelve representations that provide different levels of uncertainty information. We compared the representations in an online survey with 90 participants where we asked participants to judge their support in decision-making, familiarity, easiness to understand, and visual appeal. We further evaluated the four most promising representations in a turn-based online game. Players had to make decisions in a farming scenario based on a displayed weather forecast. The results of the survey and the game indicate that a function graph of a probability distribution function is the best way to communicate uncertain information. Nevertheless, our results also show that presenting more uncertainty information does not necessarily lead to better decisions.

Input Controls for Entering Uncertain Data

“…They suggest displaying a normal probability distribution function and a cumulative probability function on top of each other. This is in line with findings of Greis et al [7], who found that laymen performed best in a game when supported by a probability distribution function plot. Color gradients also proved suitable for visualizing probability distributions [9], and have even been suggested as an alternative to bar charts with error bars [3].…”

Tailoring Infographics on Water Resources Through Iterative, User‐Centered Design: A Case Study in the Peruvian Andes

“…If a statistically significant relationship (p = <0.05) was found, we then used the Wilcoxon signed-rank test as a post hoc analysis to see which two poster responses, and in which criterion, were significantly different. Similar statistical approaches have been used in comparable analyses to this one (Greis et al, 2015;McMahon et al, 2016;Skau et al, 2015).…”

Communicating Hurricane Risks

“…Such visualizations of uncertainty, risk, and probability that are consumed by the public are notoriously difficult to both render and interpret [19,25]. However, uncertainty in visualizations can help people to make better estimates [18], suggesting that uncertainty should be made more obvious in hurricane risk images.…”

“…Findings contribute to multiple concerns in the expanding field of human-computer interaction, including that of information visualization and challenges with representing concepts like uncertainty, risk, and probability in consumable ways for a broad public [19,25]; the relationship between formal and informal risk communication by computer-mediated means in disaster response [20,44]; understanding the nature of information diffusion on widely adopted social platforms, and devising methods for that study [17,48]; and supporting needs of the weather scientific and practitioner communities to understand socio-behavioral phenomena around their information products [42].…”