“Meaningful Modeling: What’s the Semantics of ‘Semantics’?” by Harel. On the importance of clear semantics for visual and modeling programming languages:

In general, people tend to take diagrams too lightly, finding it difficult to consider a collection of graphics serious enough to be a language and profound enough to be the real thing. Perhaps the blame lies with the early failure of visual programming techniques to replace conventional programming languages.

As a result, we often see the doodling phenomenon—a mindset that says diagrams are what an engineer scribbles on the back of a napkin, but the real work is done with textual languages.

Sadly, too many language designers and methodologists share this view. Some find it difficult to understand why we can’t simply add more graphical notations to a visual formalism without spoiling an easy to understand semantics by introducing special cases or concept combinations that contradict each other. For example, in private communication, people have proposed all kinds of extensions to statecharts, such as (actual quotes) a new kind of arrow that “means synchronization” and a new kind of box that “means separate-thread concurrency.”

https://researchgate.net/profile/Bernhard-Rumpe/publication/2956210_Meaningful_modeli[…]000/Meaningful-modeling-Whats-the-semantics-of-semantics.pdf

Good old Harel. A lot of math notation likewise has a weird tension with doodling.