Interesting and important work. Communicating the uncertainty contained in data is challenging and we really could use some better tools.

If I understand the rationale behind this research, the use of probability distributions as the default visual representation is justified by the fact that actual measured points shouldn’t be shown when observed data used to derive a distribution may not be representative of the underlying population.

By showing points from a hypothetical population generated from a distribution derived from the small sample set, aren’t you still violating the spirit of the recommendation not to show individual data points?