[FRIAM] Posts from the Scotts

[Marcus Daniels]

Glen writes:

< But then he goes toward monism-by-unification with "agreeing on what matters" and whole-equilibrium-implies-part-equilibrium. >

The context was Steve's query about whether I think that multi-objective optimization is better than single-objective optimization.   That's not monism, it's a high dimensional Pareto curve where some variables are weighted as less important than others.   Lack of internal consistency or lack of evidence could be  reasons to weight them down.  That's not how it works in the idiocracy, of course.

< I really like the idea that we need such things because we're resource and inference-style constrained computational devices ... as I imagine someone like David Wolpert might argue. But why can't the types over which one infers and one's inference style be dynamically context-sensitive such that practically, methodologically, deep pluralism is a more accurate way to model what it is _we_ are? Why is that sort of thing so repellent? >

In a recent Handmaid's episode June asks Eleanor (who has severe mental health problems and a husband that is a war criminal) if she’s ever considered leaving Gilead.  Eleanor responds “You mean somewhere where I can get mood stabilizers instead of herbal tea?”   Besides the implications of pluralism on practical governance, there's the issue of overfitting.   If every situation can be explained by an unlimited number of parameters, it is never really modeled but just memorized.   What insane thing do I have to do to survive another day in Gilead?   Many handmaids lose their mind and do just that -- not just obey, but internalize the coercion as a good thing.   It occurs to me that a valuable ability of adulthood (at least in the US) is be at ease with insincerity.  In my mind it is the same thing as turning off the unification engine.   

Marcus