Why do we have legislated cliffs instead of gradients? Because approximately nobody understands lerp. And linear interpolation is the simplest (nontrivial) gradient scheme. Consequently, we get cliffs or, if we're lucky, lookup tables that approximate gradients with stair-step successions of small cliffs.

This isn’t a very satisfying explanation to me. I mean, the idea of a cliff causing bad incentives is extremely widely understood (to the point where people under progressive taxation policies avoid taking overtime because they misunderstand the policy and believe the cliff to be there).

I wonder why we don’t define these things in terms of logistic functions or something. That’s high school or at least first semester college math, right?