On the roles of function and selection in evolving systems ru en

"Evolution of the Grey Aliens" — a famous lost painting by Vrubel. (c) ChatGPT

Recently, I encountered the scientific paper On the roles of function and selection in evolving systems (DOI: 10.1073/pnas.2310223120).

Michael L. Wong with co-authors concisely describes the phenomenon of evolution as a universal process inherent to our universe (not just the biosphere of planet Earth), and they do so within the context of a systems approach.

Their ideas resonate with my worldview, so I decided to summarize the article for myself briefly and, of course, for you.

And suppose you think that understanding evolution is unnecessary. In that case, I recommend you reconsider, as evolution affects not only hedgehogs and monkeys but also software, work teams, countries, ideologies, and even the thoughts in your head [ru] — understanding how all this works is essential.

The authors define an "evolving system" as a collective emergent phenomenon of many interacting components (subsystems of this system) that manifests as an increase in their diversity, distribution, and patterned behavior as time progresses.

Read the original article for details

This essay presents my interpretation of the article.

It should be close to the original, but a literal retelling was not my goal. On the contrary, I pushed to adapt the ideas of the original to my worldview (which I present in this blog).

Thus, some definitions and terms may differ slightly. In places where this is particularly important, I have left remarks. However, if you value precision, I recommend reading the original article.

Such a definition enables us to extend the concept of evolution to a plethora of systems we can observe in our universe. Perhaps even to all of them.

Examples

The evolution of stars begins with the nuclear fusion of hydrogen and helium, resulting in the formation of new elements and isotopes. These new elements initiate further reactions, creating even more new elements and isotopes. This process culminates in a supernova, which significantly increases the diversity of elements in the local region of the universe. Thus, the system evolves from a small set of elements and isotopes to a large variety through "patterned" nuclear synthesis processes.
After forming into planetoids from cosmic dust, chemical elements initiate the "evolution of minerals," which transforms a small initial diversity of elements, under the influence of planetary dynamics, into thousands of stable combinations. For example, Earth has over 5,900 known minerals, with an expectation of discovering about 3,500 more new types.
Biological evolution began with simple molecules but has led to an explosive growth in their diversity and complexity, which in turn initiated the evolution of unicellular organisms and later multicellular organisms.

The authors postulate that a system of many interacting agents demonstrates an increase in diversity, distribution, and/or patterned behavior when numerous configurations of the system are subjected to selective pressure.

The paper identifies three levels of selection pressure that lead to evolution.

First-order selection

Configurations of matter tend to persist unless kinetically favorable avenues exist for their incorporation into more stable configurations.

In other words, systems tend to settle into some energetic optimum, escaping from which (into another) requires a lot of external energy.

Stars, stellar systems, planets, minerals, etc., may serve as examples here.

First-order selection tends to create "energy pockets" that serve as sources of free energy for the systems within them.

Second-order selection

If the function of a system contributes to its persistence, then that function will be subject to selection — systems will evolve towards reinforcing that function.

By "function" in this context, the authors mean a process that causally influences the internal state of the system or its external environment.

The authors note that a key property of such functions is the control of dissipated energy. They highlight several main types of such functions:

Stars achieve homeostasis by balancing gravitational collapse with the kinetic energy generated during nuclear fusion, allowing fusion to continue.
Fire achieves autocatalysis by heating surrounding materials to combustion temperatures, prolonging the burning process.
Life closes a feedback loop through various learning mechanisms, including biological evolution and neurocognitive processes, which in turn support the continuity of information transfer, promote survival, reproduction, and continuation of metabolic activity.

Feedback loop

The original article uses the term information processing and extends it to biological evolution.

In my opinion:

The term information processing is broader than necessary for this case, and thus more vague. Information processing, as a phenomenon, can be directed towards anything, not necessarily towards maintaining stability and/or evolution of the system.
Including biological evolution as one of the types of information processing seems strange to me. I agree that there is a certain logic in this, but for me, it is too unstable.

"Feedback loop", on the other hand, implies active regulation of the system's state and excludes meta-processes, thus, more accurately reflects the essence of the phenomenon.

Evolving systems can be nested (with no restrictions on nesting), wherein components support the existence of a supersystem while simultaneously evolving within its context. In other words, the supersystem acts as an environment in which subsystems evolve.

It is essential to note that discussing the functions of such subsystems is only meaningful in the context of the supersystem in which they exist and evolve. To understand the function of a system, one must examine both the system itself and its context or environment. Or the system and its supersystem, if we stick to the terminology of systems.

Ancillary functions

The authors call the functions of nested systems ancillary functions (in relation to the supersystem). In my opinion, such terminology unnecessarily disrupts levels of abstraction and confuses, at least me. Therefore, I will not explicitly introduce a new term here and will simply refer to the functions of systems, subsystems, and supersystems.

Also, it may be the place where I missed something essential in the original article, so if you find it, please let me know.

For example, from the perspective of an organism, there may be selection pressure from its community, and the community may experience pressure from higher ecological units of selection, and so on.

Another feature of such processes is their ability to change over time. For instance, there is a theory that insect wings and bird feathers initially served a thermoregulatory function rather than flight, and only later were adapted for flight, changing their function, and, over time, the functions of the supersystem and neighboring subsystems.

Such changes in functions lead to a change (expansion) of the system's area of applicability (solution-search space), which alters the kinetic barrier of the system, allowing it to access new/alternative sources of free energy and evolve further.

Third-order selection

Selection pressure favors systems that can open-endedly produce (invent) new functions, a.k.a. selection pressure for novelty.

The more subsystems are interwoven, the more diverse functions emerge, and the more challenging it becomes to identify causal relationships between them.

Selection piramid

One can observe that the first-order selection creates conditions to start the second-order selection, and the second-order selection creates conditions to start the third-order selection.

We may assume that this pyramid is not limited to three levels and that higher levels of selection may exist, for example, in symbolic or social systems.

Authors do not elaborate on this question.

The authors hypothesize, and I agree with them, that the paradigm shift is possible in biology in the future, similar to the transition from classical mechanics to quantum mechanics.

Just as we replaced localized individual particles and discrete electron orbitals with wave functions and electron clouds, we may one day replace biological individuals with a "fuzzy/cloud" representation of living systems. This will not negate the existence of individual biological units but may significantly adjust the ontology of biological processes.

Generally, viewing the systems from the vantage point of statistics can be interesting even in the context of systems engineering. There are some movements in this direction, for example, Process Mining.

Quantitative measurement of evolution

The authors introduce the quantitative law of increasing functional information based on entropy and show that the functional information of a system will increase (the system will evolve) if the set of configurations of the system is subjected to selection based on at least one function.

I will not delve into the details of the formulas, as I have no deep understanding of entropy, and they are not crucial for the essence of the paper.

Consequences

In the discussion part of the article, the authors summarize their findings and highlight several interesting implications of their theory. From them, I want to highlight one of the most interesting.

The rate of evolution of systems can be artificially regulated. The formula for calculating functional information suggests that the rate of evolution of a system can be increased in at least three ways:

By increasing the number and/or diversity of interacting agents.

By increasing the number of different configurations of the system.

By increasing the selective pressure on the system. For example, in chemical systems, this can be achieved through more frequent cycles of heating/cooling or wetting/drying.