Data visualization is limited by the curse of dimensionality: it becomes quickly hard to display a multivariate dataset as the number of variables increases. Pair plots become hard to read, parallel coordinate displays are suboptimal for various reasons, dimensionality reduction does not always work, data tours are an interesting concept but become impractical in dimension > 5, Chernoff faces are fun but never found serious use beyond their novelty factor.

Enter causality: if we have a directed acyclic graph that we trust to represent the actual causal structure of the data and we are lucky that it is sufficiently sparse, we can focus our visualization efforts on subsets of variables that are causally related to each other, as these will have a much smaller dimension. Further discussion here: https://mariopasquato.substack.com/p/causal-visualizations-for-multivariate

By the end of 2026, will I be impressed by a new type of data visualization (new to me, though consider that I am quite knowledgeable about dataviz) that leverages causality? It can be either some new type of plot I find in the wild, on dedicated dataviz journals, sites, or competitions, or something I will develop.

Subjective judgement, I won’t bet.