I'm curious if anything was modeled being undervalued for predictive value.

Human sight is overvalued in the context of identifying optical illusions, but we don't have a lot of effective alternatives for making decisions about how we move about the world.