[FRIAM] Murdoch and Trump

[uǝlƃ ☣]

Did Epstein ever respond to your (& Derr's) criticism?

"Opaque" isn't a perfect substitute for "obtuse", but it's OK. By "obtuse", I really mean "low interpretability", where interpretability is the extent to which one can *read* and *understand* the structure of a model. It's mostly used in the analysis of neural network solutions to various problems. "Opaque" is synonymous with "incomprehensible" ... zero interpretability. "Obtuse" means somewhere in the middle, but probably toward the opaque end of the spectrum. [†]

"Expressibility" means "what a model can do", the behaviors it can express. For example a "flying squirrel" can't fly. But it can glide. So, the squirrel cannot express flying. Thus, a flying squirrel is an OK model for some types of airplane like gliders, but not others like jet planes.

"Parallax" is the more general concept for "triangulation". In triangulation, 2 perspectives are used to locate a 3rd thing. As far as I know, these two are fairly standard English words. So, all you need to do is look in the dictionary.


[†] The terms "black", "white", and "gray" box are sometimes used to indicate this "readability" property. A black box would then be opaque. A gray box would be obtuse. And a white box would be transparent. I don't like that lexicon. But I suppose it's fine for most people.

On 1/22/20 10:04 AM, thompnickson2 at gmail.com wrote:
> I understand what an obtuse model, here, although I understand it because I substitute the word “opaque” for obtuse, so perhaps I don’t understand it.  In the extreme, it’s a model that we don’t understand any better than the process it models.  It seems to go back to my argument with Epstein <http://jasss.soc.surrey.ac.uk/12/1/9.html> who asserted that models are unconnected with explanations.  No, not THAT Epstein.
> [...]
> Can somebody explain to me in Defrocked-english-major-language about parallax and expressibility?   Thanks,


-- 
☣ uǝlƃ