For starters I would caution against reinventing the wheel. Would be improvers need to be aware of and acknowledge the tools that already exist, and build on them if they can.
For example, Vensim can already accomplish the following:
“How can we use systems maps to capture the effects of our proposed interventions? The ripple effects from our interventions should be easy to trace through the system. We should immediately be able to see both the intended and unintended consequences of our actions. And it should be clear what the dominant factors and feedback loops are within our intervention.”
Even its PLE free version, Vensim already offers automated, live capture of dynamic effects in all variables of changes in some variables using sliders on desired variables, which are built into a model. One can even see the immediate consequences of structural change required by some policy options by including different substructures that can be switched in and out.
Secondly, I would agree with other commenters who said that telling the story is the big challenge to face now. Reducing the story to language risks the linearization that grammar and its isolated subject-verb sentences implies. The challenge is to present the story simply by drawing the important conclusions about feedback effects and changing feedback dominance that are suggested in a model. Again, would be improvers need to acknowledge progress made on this problem by existing tools like Netsim/Forio.