Optimized to Death: The Hypernetic Law of Experience

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The author brings up a delicate and very important topic in this manuscript that can contribute to broadening the horizon of understanding cybernetic systems. The author argues that all modern adaptive systems no longer operate on the basis of deterministic transitions, but under stochastic gradients, where recursive feedback smooths out variations and compresses diversity. Over time, this repeated optimization makes the system appear deterministic, even if at the local level there is randomness. This phenomenon is generalized under the name Hypernetic Law of Experience (HLE), according to the author. The central idea of ​​the paper is very interesting, but I recommend the following improvements:

In the Introduction section I recommend:

• add an explicit description of the concept of complex adaptive system (CAS). After all, the systems analyzed and exemplified in the paper are complex adaptive systems, not simple optimized systems. Specify that CAS is characterized by nonlinear interactions, continuous adaptation, exploration and exploitation, but also the maintenance of internal variety because HLE occurs inside CAS, not outside them;
• clarify that HLE does not contradict CAS properties, but describes a degenerative trajectory of CASs so that it is not understood that HLE denies adaptation, but that it is a structural limit of adaptation;
• complete the section with a clear connection to classical cybernetics, namely that systems are governed by feedback loops (recall here Norbert Wiener's definition and Stefan Odobleja's contributions). I also recommend adding a clear reference that any cybernetic system has a regulator operator that pursues stability or performance, but the regulator becomes problematic if it maximizes performance only in the short term, without variety preservation mechanisms.

 

In the Background section I recommend:

• clarify that the phenomenon of manifold collapse under recursive optimization appears only conditionally, not inevitably, and to explicitly position HLE in relation to the classical properties of CAS;
• clarify that HLE should not be understood as a deterministic law of adaptive systems failure, but as a diagnostic tool for the risk of fragility induced by excessive optimization.

In section 4, I recommend clearly mentioning that it is not the properties of CAS that are described, but a particular regime of their operation, when recursive optimization dominates the control mechanisms. Also, it might be good to formally link the parameters in the Rebis equation to the feedback mechanisms specific to cybernetics systems (e.g., in real cybernetics systems λ is not exogenous, but is regulated by a control operator, and the collapse occurs when the controller itself becomes subordinate to optimization).

Author Response

Please see the attachment. Thank you again to the reviewer.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This article highlights a very interesting and underrated phenomenon that is otherwise essential in the development or the modeling of human or artificial agency namely, that of recursive degradation of variety and control capabilities is a potential vulnerability in all recursive systems.

This is an original and challenging perspective that deserves spreading through publication, and discussed further by the scientific community.

Moreover it resumes to the public the less known Law of Experience of Ross Ashby (please complete surname is Ross Ashby).

At the end of pag. 1 the sentence or paragraph seems truncated or missing "The Hypernetic Law of Experience (HLE) generalizes this tendency:", however not explained.

I've felt some kind of repetition in the sections, particularly 3 and 4. I ask the author to double check it or render it more readable.

We suggest to use numbering for the equations. The first one seems wrong (subscript issue? Or not understood).

and λₜ ∈ [0, 1]? What is lambda? (Although repeated many times afterwards)

We suggest to double check all the mathematics and its rigorousness throughout the paper.

Please also use some references to add to equations, in order for reader to find more extended and rigorous explanations of the maths.

When author cites “narrow behaviors that result in distinct disadvantages” I feel the overfitting issues in learning have the very same nature, to be discussed.

Figure 1, would deserve more explanation to be fully understood and appreciated.

 

 

 

Author Response

Please see the attachment. Thank you again to the reviewer.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors The work can be published, my recommendations have been addressed.