Effects of Forecast Number, Order, and Cost in Multiple Forecast Visualizations

Abstract

Although people frequently make decisions based on uncertain forecasts about future events, there is little guidance about how best to represent the uncertainty in forecasts. One common approach is to use multiple forecast visualizations, in which multiple forecasts are plotted on the same graph. This provides an implicit representation of the uncertainty in the data, but it is not clear how many forecasts to show, or how viewers might be influenced by seeing the more extreme forecasts rather than those closer to the mean. In this study, we showed participants forecasts of wind speed data and they made decisions based on their predictions about the future wind speed. We allowed participants to choose how many forecasts to view prior to making a decision, and we manipulated the ordering of the forecasts and the cost of each additional forecast. We found that participants viewed more forecasts when the outcome was more ambiguous. The order of the forecasts had little impact on their decisions when there was no cost for the additional information. However, when there was a cost for each forecast, the participants were much more likely to make a guess based on only the first forecast shown. In this case, showing one of the extreme forecasts first led to less optimal decisions.