Interactive forecasts: exploring very uncertain projections of quantitative variables

Predictive systems use a variety of Big Data and Artificial Intelligence technologies to ‘predict’ the future course of selected quantitative variables (inflation, house prices, passenger numbers etc.) with various degrees of precision and certainty. The challenge for user interface design is to display the output of these technologies, particularly information about uncertainty, in a way that supports exploration and decision-making by everyday users. This Work in Progress paper reports the design and formative testing of an ‘Interactive Forecast’ - a mock-up of a predictive system, in which the very uncertain future path of a quantitative variable was not represented explicitly by chart elements, but rather implied by the whitespace between elements (axes, labels) and other data (lines on the chart). A video of the mock-up is online. In the formative test, participants’ ‘guess-timates’ of the future value of the quantitative variable were consistent with reading the display in the intended way. However, the test also identified confusions and ambiguities with the mock-up, which hampered participants’ interpretation. Future work will continue to iterate, to complete the design of the Interactive Forecast, and to refine the user testing process.