It sounds like one of the core issues is dimensionality reduction. I would suggest reducing everything to text. Text is one-dimensional. In this case: >> Like

It sounds like one of the core issues is dimensionality reduction. I would suggest reducing everything to text. Text is one-dimensional. In this case: >> Like any other map, we can add textual labels — street signs if you will. This allows us to navigate the design space verbally with commands like: “take me to a maple leaf.” Once there, we could say something like: “take me a bit closer to an oak leaf.” The change here would be subtle. Rather than a two-way conversation between tool and user, instead it's more like the user is telling a story and the tool is making suggestions along the way: start: "Create a leaf" tool inserts a suggestion: "Create a [maple,oak,cherry,...] leaf" user updates the story with the clarification: "Create a maple leaf" tool suggests: "Create a [red,green,...] maple leaf" user: "Create a green maple leaf" user starts typing more: "Create a green maple leaf on a" tool inserts suggestions: "Create a maple leaf on a [branch,wreath,...]" Hope this is helpful.