They help with interpretability, which is a step brother. See for example "From Shapley Values to Generalized Additive Models and back" https://arxiv.org/abs/2209.04012
Indeed, Thank you!
They definitely do, I'm planning to release a new package to do that in a couple weeks