I've started to play with directed graphs kind of like Bayesian networks to visualize my belief structures. So a node is a belief (with some indication of confidence), while connections between graphs indicate how beliefs influence (my confidence in) other beliefs.

This seems useful for summarizing complex arguments, easy to memorize, and (when looking at a belief structure that's bigger than my working memory) for organizing and revising thought.

However, there are a few decisions in how to design the visual language of such graphs that I can't see obvious solutions to. If I include necessity and sufficiency, which seems really useful, how does that square with the confidence calculations? How should I represent negation (the other logical connectives are fairly obvious)? Should I have different types/shapes of nodes, and if so, which?

So I'd like to see the work of others who have done similar diagrammatic depictions of belief networks, to play with them and see what works for me. I've seen influence diagrams, but I'm not convinced the choices made there are obviously the best ones. Does anyone have pointers to other existing Bayesian diagram schemes I should look at?