Ah, hierarchical Bayesian / hierarchical gaussian — I shall not abandon ye as very useful but the hype of complete I must set aside. There’s several reasons why Bayesian and Gaussian each are not enough for a “top to bottom/bottom to top” view of complex systems that I’ve dug into in the past, so I knew enough to have a skeptical placeholder as I read the new material, glad to find some good, coherent texts on autism finally. This does not take away from their use in that way. The explanatory power of hierarchical Bayesian is huge and covers a good emount of territory, and hierarchical gaussian, which reminds me of octree is great at capturing vague-to-specific / specific-to-vague within a particular class of concept, but does it map cleanly to sensory? That’s where it gets tricky. Embodied cognition models which map straight to senses and hierarchical belief revision showing extreme precision in sensory for autism combined with a vague (or leads to a vague) belief that is imprecise and/or uncertain and/or weak are similarly simplifying just HOW abstract the human brain can get. A better model will allow for multi-model composition and also destruction / merging of distinctions. Distinctions _can be_ lost, particularly if they are never remembered in the first place. I don’t think that this is an issue as much in autism, as the hierarchical bayesian _does_ very well capture the strong sensory ties present in the experience of autism as well as the mono-channel or preferred multi-channel (hand-eye, taste-visual, etc). But I do think it’s a problem to characterize strong belief generalizers as their generalizations can be so powerful they can override what sensory data says. Unless… unless that’s the point of hierarchical Bayesian? I never liked Bayesian because the notion of belief-revision being dramatic implies a very strong sense of permanence that must be shaken (revised). But what if all feels impermanent around and you work hard at maintaining integrity of self-and-world because of this constant “dark energy” pulling-away-from-itselfness of it all? Then how much dramatic revision is needed when one is constantly revising? So full circle. These are first thoughts upon waking. It jumps into the middle-of-the-story of something I just learned but attempting to connect it to what I know. Further research will be to find out what specifically about Gaussian and Bayesian I had objected to as generalizers previously to confirm. I think it was due to their single-variable natures.

Ah, hierarchical Bayesian / hierarchical gaussian — I shall not abandon ye as very useful but the hype of complete I must set aside. There’s several reasons why Bayesian and Gaussian each are not enough for a “top to bottom/bottom to top” view of complex systems that I’ve dug into in the past, so I knew enough to have a skeptical placeholder as I read the new material, glad to find some good, coherent texts on autism finally.
 
This does not take away from their use in that way. The explanatory power of hierarchical Bayesian is huge and covers a good emount of territory, and hierarchical gaussian, which reminds me of octree is great at capturing vague-to-specific / specific-to-vague within a particular class of concept, but does it map cleanly to sensory?
 
That’s where it gets tricky. Embodied cognition models which map straight to senses and hierarchical belief revision showing extreme precision in sensory for autism combined with a vague (or leads to a vague) belief that is imprecise and/or uncertain and/or weak are similarly simplifying just HOW abstract the human brain can get.
 
A better model will allow for multi-model composition and also destruction / merging of distinctions. Distinctions _can be_ lost, particularly if they are never remembered in the first place.
 
I don’t think that this is an issue as much in autism, as the hierarchical bayesian _does_ very well capture the strong sensory ties present in the experience of autism as well as the mono-channel or preferred multi-channel (hand-eye, taste-visual, etc).
 
But I do think it’s a problem to characterize strong belief generalizers as their generalizations can be so powerful they can override what sensory data says.
 
Unless… unless that’s the point of hierarchical Bayesian? I never liked Bayesian because the notion of belief-revision being dramatic implies a very strong sense of permanence that must be shaken (revised). But what if all feels impermanent around and you work hard at maintaining integrity of self-and-world because of this constant “dark energy” pulling-away-from-itselfness of it all? Then how much dramatic revision is needed when one is constantly revising?
 
So full circle. These are first thoughts upon waking. It jumps into the middle-of-the-story of something I just learned but attempting to connect it to what I know.
 
Further research will be to find out what specifically about Gaussian and Bayesian I had objected to as generalizers previously to confirm. I think it was due to their single-variable natures.
===
https://www.repository.cam.ac.uk/handle/1810/275943 paper that gave me a little focus to my skepticism.
—–
 I’ll still look for it though in research as it is very useful for autism due to its strong sensory focus. But I’m skeptical of its power to explain neurotypical delusion, unless one can perhaps disconnect the sensory in place of … wordplay. Semantic -> belief type of thing. Hm.
—————

Leave a comment

Your email address will not be published. Required fields are marked *


3 − one =

Leave a Reply