[FRIAM] Formalizing the concept of design

[uǝlƃ ☣]

It seems to me you're still directly on topic.  Nick's emphasis on hierarchy leads directly to (forgive me, here) the *flatness* or flattenability of dynamical systems equations versus whatever units multi-level selection might operate over.

It's probably just another fit of apophenia.  But I just finished my incompetent reading of McShea and Brandon's "Biology's First Law", wherein they criticize the Hardy-Weinberg "law" (at least as a universal biological law) and, later in the book, assert that their ZFEL is strictly hierarchically applicable.  They go on to reference Bouchard 2008 (http://www.fredericbouchard.org/textes/BOUCHARDcausal_persistence_fitnessPHILSCI08.pdf) and say:

McShea and Brandon, § A Generalized ZFEL for Physical Systems:
> "In this more general understanding, reproduction would be just one route to persistence, the route biology employs in a world  of mortal organisms.  It is a mechanism that increases the probability that a given phenotype in existence at some time will also be present at some later time.  The organisms die but the lineage persists."
This hearkens back to Nick's attempt to paraphrase Rosen with:

On 10/25/18 2:55 PM, Nick Thompson wrote:
> We need a science of biology that is materialistic but NOT mechanistic. 
> [...]
> But if life is an organization of things from another organization, the question becomes, “What kind of an organization could scaffold the organization we call life.

Rosen specifically targets operational closure and material openness.  But McShea and Brandon seem to argue obliquely against that earlier in the book:

McShea & Brandon § Forces and Null Expectations:
> "... Newman and Müller (2000) have argued that accurate inheritance (...) is an evolutionary achievement, the result of natural selection, and is not evolutionarily primitive (...).  We agree.  But heritability, in the evolutionarily relevant sense, does not require anything like what Newman and Müller have in mind.  As Griesemer (2000) has emphasized, biological reproduction involves material transfer; that is, the parent transfers not simply information, not jut a 'blueprint,' but an actual bit of matter that used to be parent and that now becomes offspring. ... And this material transfer ensures some degree, even if low, of fidelity of reproduction."

Anyway, there are 2 questions that this apophenic fit lead me to: 1) Are the "higher order" constructs (including your interfering distributions, M&B's "lineages", units of selection, etc.) reducible to "lower order" constructs -- i.e. what I infer as John's implicit assertion that a sequential machine should be able to reproduce *any* chronicle? And 2) Are material and organization *actually* separable or do they always remain at least a tiny bit dependent?

Regardless of any of that, though, I'd also appreciate any and all opinions about M&B's ZFEL.  In my googling, I failed to find criticisms of it.  My skeptical homunculus refuses to remove the handcuffs from my gullible homunculus *until* I find a scathing criticism.




On 11/6/18 11:58 AM, Eric Smith wrote:
> On 10/29/18 12:25 AM, Stephen Guerin wrote:
>> As we've discussed over the last few years, The Action Principle (energy * time) and least (stationary) action may provide a more fundamental selection principle in biology than natural selection and could be a mathematical formulation you're asking for. Many applied problems in complexity like ant algorithms using dual pheromone fields, level-set methods, and route search on a road network using simultaneous floodflill from both origins and destinations might be considered least action path selection. I make the claim on intuition - I expect Eric Smith would reject or accept this based on more formal understanding.
> 
> I don’t want to just drop this, but I don’t know how to respond to it usefully.  I think of the two (principle of least action (PoLA) and natural selection (NS)) in completely decoupled thoughts.  For me, PoLA in the classical form is equivalent in content to dynamical equations, but because it formulates them as an extremization principle it more readily exposes consequences of symmetry.  In quantum mechanics, I can find the same thing as a stationary-path consequence of interference of phase advances over many paths.  In statistical mechanics I can find a “stochastic effective action” that captures stationarity through a similar kind of interference, but no longer among quantum phases, rather in some interaction of distributions with the shadows of late-time questions we might ask about them.  (Sorry that formulation is so cryptic; for those who prefer that one just show what one means by calculating, there is this:
> https://arxiv.org/abs/1102.3938
> )
> 
> For me, NS comes up in response to a completely different collection of questions (which may or may not be about the same phenomena).  I think of NS as being about whatever it is that makes time different from just another dimension of space, so that there is always something falling apart that can only be maintained by being passed through a filter.  I would prefer to use NS (or maybe, better, “Darwinian selection”) as a subset of the previous general sentence, to refer to phenomena that are organized in architectures of individuals and populations, as distinct from simple kinetic phenomena in general.  Of course one does not have to draw the boundary there, but I find it a good way to use a new word to distinguish individual/population-based phenomena from general kinetic organization, for which we have other terms already.  Also NS is about information in the same sense (exactly) as Bayesian filtering is about information.  Sometimes effects of any of these, as they act in populations, can be expressed in terms of actions, but I don’t think of the service that action gives in displaying the nature of a calculation as being the same thing as NS does in declaring what kinds of phenomena we are talking about.
> 
> Sorry I could not offer better, or more likely I am not understanding where the conversation is.