On darwinist evolution in complex systems

Update 21 august 2020:

 

A fascinating 2020 paper by Prof. Irun Cohen and Assaf Marron on the units of evolution of cooperation and the physical forces behind it suggests that unit of evolution can be collections of related interactions at multiple scales. As I wrote below, the most inert system level is determined by interaction that creates bonds. I propose that the most inert level is the unit of evolution. What they add is the possibility that one huge system can have hierarchical systems each with their own time and geographical scale, and their own unit of evolution. This separates relatively behavioral universal patterns from patterns that are limited to local subsystems 

Notes to Human Evolution beyond Biology and Culture; Evolutionary Social, Environmental and Policy Sciences (Cambridge University Press; 2018), by Jeroen van den Bergh

My 2019 summer vacation read was Human Evolution beyond Biology and Culture, by Jeroen van den Bergh. Educated as evolutionary biologist and with a PhD in a social constructivist view on transition management, this book was a must-read to me.

 

I found it highly interesting and thought provoking. The book shows the explanatory power of (neo)Darwinist thinking, as it moves beyond proximate causes to ultimate causes of change, in practically any complex system. For those who haven’t read the book, at the bottom of this blog I summarize the book in my own words and examples.

 

The questions it provoked I elaborate first. Those points seem essential to me for a general theory on (neo)Darwinist, i.e. VSIR, evolution, in particular of social systems.

 

Easy classification of processes of change seems hard

It is tempting to look for generalities in the mix of theories Van den Bergh presents. These can be looked for in the relationship between subsystems. Obviously, it is often hard to draw boundaries between (social) subsystems, and their mutual relation will be even more difficult to define. It seems even harder to formulate simple, generalized criteria to classify any two given systems as either co-evolutionary, co-dynamical, one self-organizing from the other, or co-adaptational. Yet precisely these explanatory mechanisms produce the deepest, or most ultimate, understanding of any change. The following table make a first effort. It distinguishes three relationships between two subsystems:

-          Dependency: both systems are part of the selective environment of the other

-          Embeddedness: one system emerges from the other through self-organisation

-          Inertia: one system is more inert than the other if it is more robust to pressures from other, less robust, systems. A relatively inert system level codes for less inert “phenotype” levels; which again may enable emerging embedded system levels that are not coded for at the enabling level. These embedded levels can be equally inert as the levels that enable them.

 

Interaction between two systems (two CAS):

Equal inertia

Unequal inertia

Systems mutually dependent

2 embedded levels

Co-dynamics: both adapting to each other without evolution of their respective coding levels

Self-organization: one system adapts to the other and can be seen as emerging from it, but not the reverse

embedded level and its coding level

Co-evolution on top of self-organization: the embedded level is itself a coding level for even more embedded system, and both co-evolve (e.g. genes and memes)

Self-organization: the less inert system self-organizes from the relatively inert system

2 systems at the same coding level

Co-evolution: mutual VSIR influence (e.g. 2 genes or 2 memes in a population; or memes influencing genes they emerge from – downward causation)

Not applicable

Systems mutually independent

Co-adaptation: both co-evolving or co-dynamical with a joint selective environment

 

What makes a system more inert than another system?

Inertia seems a rather enigmatic characteristic, as 2 subsystems can be of completely different nature, like a company and a gene. Still it does not seem impossible to compare their inertia. Inertia is produced by the bonds between the agents in a system – the subsystems from which it emerges.  Bondage may be chemical, physical or social. A chemical bond is stronger than an H-bond or gravity. But if there is downward causality, and the theory is accurate, social systems apparently can be just as inert. In social systems, identity is a strong mechanism of social bondage. It is reinforced by rituals. Bridging social capital, connecting identities to create embedded meta-identities, is weaker (ref. Robert Putnam). In times of crisis, people tend to seek alliance first with whom they share their first identity. Genes may enable this behaviour, and identity memes may “be” the bonds.

 

However, looking at real life complex social systems, boundaries drawn around will always be arbitrary. It will also be difficult to know which one is embedded in the other, or if both are embedded in a third system. If there is coevolution, it is therefore hard to know if it is upward or downward or horizontal. Most memes therefore may have to remain a useful working hypothesis, like factors in factor analysis. But is the quest for ever more ultimate causes always necessary and feasible? 

 

What “makes up the DNA” of a social structure or a culture?

Interaction between two social subsystems is the outcome of both their individual behaviours. What is the ultimate explainer (or driver) of behaviour, sometimes metaphorically called their “DNA”? Van den Bergh gives a hint on page 448, where he asserts that transition policies have evolved as some kind of meta-governance: arrangements that accelerate the VSIR processes at a lower level, steering the direction of that evolution toward what is believed to be sustainable (or away of what is believed to be unsustainable – as a strange attractor). If such transition policies work, there can be different social system levels having a downward causation: the governance systems with their own logic emerge from the composing social systems, yet govern them and can influence their “DNA”.

 

In this light, one may refer to the work of Maslov, Graves and in particular Beck and Cowan (1995), who postulate a hierarchy (or levels, layers) of evolving values which each emerge from the previous layer if there is any spare energy at that level available. They coin these levels the VMemes, which they assert are universal in social systems, but not all levels come to expression in each individual, organization or each culture. More embedded (or higher) levels of VMemes may not always emerge, either as the potential carriers don’t have that capacity (i.e. “a latent phenotype”), or if the previous level is so consumed with survival in the short term that it has no energy left for the emergence of more embedded system levels. Like coevolution of memes and genes, there can in Beck & Cowan’s view thus be top-down causation from higher VMemes to lower VMemes. Beck and Cowan have in their 1995 book given mounts of circumstantial evidence (in a way similar to what Van den Bergh also uses to make his points). Like in Giddens’ duality of structure, VMemes emerge and evolve in a social environment or network, constituting their structure, but only supported by those nodes in the network prone to carry these VMemes. And like duality of structure, empirical testing is difficult as it amounts to reconstructing the emergence of the chicken and the egg.

 

How does leadership evolve?

There is increasing interest, also in the transition literature, for leadership. It is seen as a factor that either can break down systems or bond them together, different strategies that both may help larger systems to adapt to changing circumstances. From that angle, it would be interesting to link Beck & Cowan’s theory to recent theories on leadership in complex human systems (complexity leadership theory; e. g. Uhl-Bien and others), which theorizes about three kinds of co-evolving and symbiotic highly embedded sub-networks. These are termed administrative leadership (which controls human systems and selects), adaptive leadership (creating variety - innovation, meme mutation), and enabling leadership (that pre-select new memes before they are brought in the mainstream for final selection). The final selection takes place in the mainstream market, civil and electoral systems. From this point of view, memes coding for adaptive and enabling leadership can be seen as embedded in / emerging from administrative leadership, potentially downward causing a selection of technologies and favourable niches that help longer-term survival of larger systems, potentially even humanity.

 

As Van den Bergh writes somewhere I believe, liberal democratic, open societies with little repression, are better at adapting to changing circumstances than authoritarian societies, where administrative leadership leaves little room for enabling or adaptive leadership. However, open societies depend on bridging social capital which tends to emerge only after (not in) times of crisis. Ultimately, therefore, the balance of society depends on the availability of the natural resources that enable our economy, and whose lack can cause crisis. But it also depends on an immune system that can remove “toxic” memes that code for pointless self-destruction.

 

How does VSIR link to public decision-making?

Van Den Bergh describes the limits of public choice theory, but theories on network governance are also of interest. In these theories, the ultimate causes of decisions are difficult to observe, as they happen in development rounds where players constantly reproduce their positions and the underpinning narratives, but each time slightly different, coevolving with the social environment of the negotiators, until the social environment permits a breakthrough. This breakthrough is then accredited to the top decision-maker as “his or her decision”, as if the decision were part of his “memotype”. A top decision-maker unaware of his or her dependency on the underlying innovation / transition system may still want to control it: authoritarian. Reproducing one’s own memes without adapting them to what the environment permits may be based on strong hierarchical power. As Karl Deutsch wrote: power is the ability not to have to learn, i.e. it is being the most inert.

 

Highly embedded VMemes may emerge from the understanding that authoritarian control is unsustainable if circumstances change, and may entail slavery and other injustices, and that decision-making rules can be designed to balance the negotiating powers enabling the emergence of innovative forces (e.g. environmental assessment procedures can be such rules; see also Nooteboom in IAPA in press).

 

Obviously, an interesting question then would be: what would make such VMemes emerge? There would have to be some way to jointly observe and intervene at meta-level. This is where the matrix above may help: classifying subsystems according to the matrix above may help assess the link between inertia (durability) and change or plasticity that is needed for a sustainable development.

 

Science as evolutionary process – enlightenment as “managing the attractors”

Van den Bergh also refers to work that suggests science itself is an evolutionary process. It is interesting to note that the widely accepted scientific method proposed by Karl Popper is based on falsification. This implies that the tree of knowledge is driven by a strange attractor, like the idea of sustainable development itself is a strange attractor: one can only observe what it is not. Van den Bergh does not mention attractors, but they are defining emergent characteristics of complex systems. It is interesting to speculate that less embedded VMemes are mainly point attractors, and more embedded VMemes are mainly strange attractors. This would have lessons for dealing with uncertainty: people carrying such highly embedded VMemes would deal with uncertainty, coping with complexity, in fundamentally different ways than people who have not developed such VMemes. The enlightenment movement, in particular Popper’s falsification theory, implicitly has defined a strange attractor its leading principle, thereby consciously managing attractors.

 

The epistemology of meme research

Finally, memetic research will inevitably encounter epistemological problems. Constructivists will contend that positivist science (falsification based on trials) is largely impossible. Action research is needed. Scientific interpretations can only be second hand; the first hand being the interpretation of the subjects themselves. Whereas more uncertainty remains, plausibility and social learning accelerated by social scientists is still useful. This topic is discussed in Nooteboom & Teisman (in prep. Chapter in Victor Galaz (ed.) Global Challenges, Governance and Complexity" for Edward Elgar).

 

MY SUMMARY OF VAN DEN BERGH’S BOOK (IN 16 POINTS)

  1. The defining VSIR-characteristics of (neo-)Darwinist systems (page 63-) are most clearly found in biological systems where they operate on genes. VSIR stands for variation, selection, innovation (mutation), and replication of some (relatively inert) system level.
  2. Self-replicating systems (like genes) to some extent structure their (relatively less inert) environment (self-organization). Genes thus self-organize into “phenotypes”: organisms, that again may again self-organize their close environment to some extent for better survival (“extended phenotype” proposed by Richard Dawkins; ref page 116).
  3. Evolution of genes has led to intelligent phenotypes. These have the ability to observe and interpret their environment, to plan and to communicate – the latter in order to plan and act together in groups, like many animals do (individual and joint reflection).
  4. Genes can evolve to code for intelligent phenotypes which themselves can adapt to changing circumstances (without the genes having to adapt). Several phenotypes, being in each other’s environment, may then adapt to each other without gene evolution. This is defined as co-dynamics (ref. page 116). For example, tree shapes in a forest can adapt to their neighbours. Or, China and the US may mutually adapt their trade negotiation strategy without changing their ultimate (i.e. inert) drivers (memes, see hereafter).
  5. Co-dynamical communicative interactions can produce evolving system levels, that is: defined by their own self-standing VSIR-mechanism. These embedded levels are enabled by genes, but the genes do not code for the internal interactions at the new system level. E.g., genes enable to learn languages, but languages are not coded for by genes. Languages have their own relatively inert codes (words, expressions, to some extent grammar) that can evolve, in human brains self-organise into sentences, and that survive differentially according to selection by their social environment, in analogy of genes.
  6. Hypothesising an analogy with biological VSIR, the communicative gene-analogues are defined as memes, and the phenotype-analogues are not just words and sentences, but also social structures such as organizations and cultures.
  7. Although the boundaries around memes are more difficult to draw than the boundaries around genes, communicative systems do have degrees of inertia and embeddedness, implying that there also will be replicating levels and self-organized levels. These are messages, ideas, that circulate in a population.
  8. Communicated messages move around in populations, especially in human populations, driving the individual and collective behaviour. Moving around, they are constantly reproduced (copied) by humans, and therefore mutations are possible. The social environment selects by either reproducing (resonating) them or not.
  9. Whereas the meme level is embedded in the gene level, the former does not emerge from the latter by self-organisation. In terms of inertia, they can be equal: memes can enable a downward causation (memes influencing genes) (page 96) in addition to the existing upward causation (genes enabling memes): in this case they are in each other’s selective environment  (page 199).
  10. Which memes resonate well depends on the properties of the existing social system (i.e. the existing “memotypes” and emerging social structures). In changing circumstances successfully resonating memes need not be the best for survival of their hosts; some memes can even destroy their host populations. (Note: e.g. Luhmann and Leydesdorff prefer to describe memes as waves, social agents as having resonance frequencies, in analogy of light which can be described as electromagnetic waves but also as particles (photons)).
  11. Such two-way influence between different VSIR-systems is co-evolution. It is caused by interdependency of the two systems of VSIR nature. Co-evolution can also happen between different genes within a population, or for genes that code for interdependent species, like the bee and the flowering plant. The same goes for memes, e.g. memes that code for democracy may co-evolve with memes that code for education systems.
  12. Two systems living in the same environment may both simultaneously change and still not be co-dynamical or co-evolutionary: both systems may also adapt to a common environmental change rather than having direct interdependency. This is termed co-adaptation (page 200).
  13. A social structure may be relatively inert: memes create order in social chaos; but a structure still self-organizes from something even more inert. Social structures can adapt to each other and to changing circumstances. If two social structures both change, it can be either co-dynamics, co-adaptation, co-evolution or co-incidence. In social systems, it can be difficult to distinguish these mechanisms, looking for ultimate causes. The question is, of course, is it useful to keep on digging deeper for ultimate causes, if the underlying system levels are robust enough to survive and continue enabling the embedded levels.
  14. Closely linked is the idea of autopoiesis, the theory of living systems (originally, Maturana and Varela, referred to on page 284). Living systems, or autopoietic systems, are in fact individual “organisms” (biological or cultural) that may be part of a population that has VSIR characteristics. Their replication mechanism is enabled by feeding on their environment (dissipating some energy).
  15. (Social) Complex Adaptive Systems (CAS) (page 262; 288) include autopoietic systems but are more general. CAS emerge by self-organization and have their internal metabolism but do not necessarily reproduce on their own level, like living organisms do. They are the living vehicles reproducing memes and genes. Social CAS, however, have much vaguer species distinctions, as there is only asexual reproduction; expansion of power. Many highly embedded CAS are unique or similar only in an abstract sense.
  16. Wilson’s distinction between CAS1 and CAS2 (page 288), the latter being composed of agents themselves CAS, seems useful. The earth is clearly a CAS2 under the Gaia hypothesis (page 107). Formulated in this way, any two large enough subsystems of the earth are CAS2. Social subsystems and social CAS are therefore practically synonyms.