Interesting and important work.
Andrew Staroscik

Hi Andrew,

Thanks for your question! If the observed data points are not representative of the population, doesn’t that imply that the inferred distribution will also not be representative? In which case it isn’t clear what we gain by showing the distribution instead.

More generally, the set of hypothetical outcomes that are shown in using HOPs should be representative of the inferred distribution. Of course, as draws, they will never offer a complete depiction. But I’m not convinced that there is a good reason to avoid showing individual draws if doing so helps people more accurately infer probabilities from the distribution (which is what we’ve been observing). A critical piece of our work on HOPs is to understand the trade-offs involved in showing a set of draws rather than depicting a distribution, so that we can choose the right approach for the right data sets and tasks.