Telling the Truth about the Future
We live in a world of fallible prophets. The weather man might tell you that 10 days from now, at 12:00pm, the weather in Boston will be precisely 70 degrees Fahrenheit. Your trusty GPS tells you that despite the heavy snow and heavier traffic, you will reach your dentist's office in precisely 8 minutes and 42 seconds. And your omniscient mother tells you that if you run around in the rain then you will wake up with a cold precisely tomorrow.
Inherently, we know that weather men, GPS systems, and on ocassion even mothers, can be wrong. The beach day in Boston gets rained on once in a while, the dentist appointment gets missed, and that cold is yet to be caught.
The truth about the future is that it can never be perfectly foretold. So how do we communicate uncertainty to those who are relying on predictions to plan the future in an informed way?
This summer, I will be working with Jessica Hullman, Matt Kay, and Sean Munson at the University of Washington to develop techniques for visualizing uncertainty. We will be designing, implementing, and testing ways of visualizing the uncertainty of bus arrival predictions in the app OneBusAway.
OneBusAway gives realtime arrival predictions for the erratic buses of Seattle. This is a critical tool in a city where buses rarely arrive on time, schedules are changed suddenly, and many routes are traversed infrequently. Though these realtime predictions are a huge improvement, they can sometimes be wrong. Downtown, these delays are usually just a matter of 1 to 2 minutes. But as we move away from the Space Needle towards the surrounding suburbs, buses become more infrequent and more delayed, and OneBusAway predictions become more important and more inaccurate. Even in the heart of Seattle, OneBusAway can sometimes be wrong.
The dwellers of Seattle already know both the value and the capriciousness of OneBusAway, and each individual user has developed an internal, intuitive algorithm for how to deal with the uncertainty of predictions. We want to standardize how uncertainty is measured and represented, so that people can have something more concrete to rely on than their own experiences when trying to decide how wrong the bus could be.