Why uncertainty visualization is hard
- Efficient encodings for uncertainty can be hard to find
- try putting mean, variance, and interval estimation in one plot + doing this when useful channels are already used up
- Even if you get the encoding right have to contend with making sure people understand it
- explain a CDF to someone (N.B.: sometimes this works! task has an effect here)
- Even if people understand have to deal with how people perceive uncertainty
- linear-in-log-odds and other perceptual models of probability
- Even if people correctly perceive the uncertainty have to deal with how people make decisions under uncertainty
- Even if you solve all these problems in one context a new context may cause your solution to break
- see e.g. deterministic construal errors
- Plus, you still have to build the damn thing
- But some strategies for building uncertainty vis using the grammar of graphics helps a lot here (the how of this book)