Why You Don't Want to Ask Too Many Questions to Clusters?

The idea is that for a specific lustre you don't want to ask too many questions to the data. Decision tre has that appeal for all reasons. We see it elsewhere in explainability. But thinking to that hypothetical executive i mentioned, if you bring me a tree that takes 20 pieces of paper to no glued together to visualize, cause it's so deep and so wide, i'm probably not going to be supersatisfiedu what else can we constrain here? Well, yes. And that's precisely the topic of our newest paper, ri.