[FRIAM] ockham's razor losing its edge?

[steve smith]

glen wrote:
> Maybe it's just confirmation bias. But parsimony is either obsolete or 
> a red herring. 

I have always felt that at best it was a first order heuristic. My own 
experience both with highly structured technical projects and fairly 
ad-hoc DIY projects ranging from building things to repairing things to 
growing things is that Einstein's version (as quoted by Pieter) of "As 
simple as possible but no simpler"  is a wonderfully loaded challenge.

Virtually every advanced trick I know of in the aforementioned domains I 
learned from someone who had "been there and done that" enough to know 
all the places the best "short cut" is a "long cut".    Bring everything 
back into spec as best you can before "tweaking" the things you know how 
to.    I'm bad at this, but it really does help to calibrate a tool 
before using it (e.g. miter saws work much better if the indicator of 
90deg REALLY IS 90deg, easy to check easyish to adjust, worth it).   But 
to support the idea of parsimony, realizing that rather than getting out 
a fancy square, one can simply cut a piece of scrap and flip it over and 
see if it aligns with the blade!

In the spirit of (simulated) annealing, it is worthwhile to go explore 
some of the *less parsimonious* paths from A to B even if later you come 
back and optimize for parsimony in the context of a (more) global 
optimum found by allowing yourself discursive searching.


> Perhaps it used to be useful when inference to the best explanation 
> was our only choice. We didn't have the resources (or impetus) for 
> multiverse analysis. Modern model selection executes a wide array of 
> models and chooses the ones that best fit whatever task is at hand. 
> Sure, there are still some resource limitations. But the primary 
> bottleneck is type of observation bias I call "schema bias". Schema 
> bias is the paradigm or framework of most familiarity to the modeler 
> (or modeling team or entire community) that prevents them from 
> including *other* types of models with which they're not familiar. 
> I.e. there are parts of the schema that are accidentally fixed that 
> could vary, were the scheme different. Proverbially, to a man with a 
> hammer, every thing looks like a nail. Anything outside or fixed 
> within one's schema is invisible.
>
> But all this is a normal part of pluralism, a concept we've beaten to 
> death on this list. But I've found a new advocate that I hadn't seen 
> before, which is why I'm taking one more whack at the dead body of 
> monism in this post:
>
> https://chrisstroop.wordpress.com/2020/05/26/a-personal-update-and-some-thoughts-on-pluralism/ 
>
>
> Granted, she's talking about religion. But it's only a hop, skip, and 
> jump from there to logics, modeling methods, or no-go theorems about 
> the universe.
>
>
> On 1/30/25 9:23 AM, Roger Critchlow wrote:
>> This was in the Complexity Digest feed this morning, it looks like fun.
>>
>> https://www.pnas.org/doi/10.1073/pnas.2401230121 
>> <https://www.pnas.org/doi/10.1073/pnas.2401230121>
>>
>> What makes a model good or bad or useful or risible?
>
>