Monday, June 24, 2019

Cultural Multi-Level Selection - "Just So" or "Good to Go"

There's an interesting conversation up on Letter Wiki between David Sloan Wilson (from multi-level selection fame) and Massimo Pigluicci (from the Extend Evolutionary Synthesis fame).  Massimo's challenge to David is to explain how a multi-level perspective on historical events is informative and falsifiable.  From the jist of the conversation, I'd wager that Massimo leans toward a "post-modern" complexity science frame.


For those not up on complexity theory in the social sciences, people tend to split into two main camps:
  1. There are post modernists who tend to view complexity/chaos as a lens or metaphor for analysis. They basically see small factor effects or emergence as necessarily obviating causal or deterministic reasoning. Cillers is the major reference here. Each discipline, however, has their own context specific experts. In education it is Davis & Sumara, Keith Morrison, and increasingly Mark Mason.
  2. There are also (probabilistic) determinists. This group is probably most characterized by a more sciency crowd who takes strange attractor topology fairly literally. This view sees human cultural tendencies in terms of knowable-enough attractor basins. For some, like D.S. Wilson, this comes from gene-culture mechanisms. Others take a more sociological view (like social psychologists) and just accept the existence of cultural basins, which while not rigidly determistic do tend to produce analogous structures over and over again despite a decent amount of surface feature chaos. In education, Koopmans and Reigeluth are probably the best examples. Peter Turchin is a good general reference optimized for historical thinking.


Massimo asked David to produce some concrete example of what (cultural) multi-level selection can say about specific historical events, and how it can be falsified in order to prove it is not some nice sounding "just-so" pseudo science.

The debate settled down to a challenge to meet Tinbergen's four questions for an evolutionary process
  1. function
  2. history
  3. mechanisms
  4. development
Let me see how I'd answer this for cultural multi-level selection approach to education.

Tensions between adjacent levels of selection mean that educational systems are torn between focussing on behaviours optimized for large group focus and smaller group focus. This may involve a focus on large group equity wherein individuals give up some degree of localized optimizations in favour of broader coherence and efficiency of scale effects. This may involve the creation of a large group styled morality for education. This may involve limits on how far one can advantage or fail to advantage any particular sub-system including identity based groups. The small group orientation focusses more on localized optimizations at the expense of large group defined norms. This may involve setting up a gifted charter school or focussing one's teaching and teaching philosophy on academically oriented or rank based endeavors.

Educational histories like Tyak & Cuban's Tinkering To Utopia strongly suggest quasi-periodic oscillations between things I interpret as large group orientations and things I interpret as smaller group orientations. Over time the system expresses deterministic chaos toward a higher level of selection.

Emergence largely happened in the West in the early to mid 1800's. This is when the public system emerged as a competitor to religious based community "schools". These systems gradually took over from private tutor arrangements and formal boarding schools for older, affluent and academically proficient students.

Education is adaptive. It provides real benefits for those who pursue it. This may occur directly via skills, or directly via "ritual-like" acceptance into a caste. Today, accreditation exemplifies part of this process. But social grease still abounds (what's the main purpose for getting into Harvard....). Thus real fitness, including long-term probabalistic fitness is part of the mechanism. Variation between people's orientation to education and adherence to its norms is another mechanism. Heritability of these orientations as per the various educational attainment and IQ inheritance, is the last mechanism.

The unit of selection if culturegens (Richerson & Boyd's gene-culture elements). These have varying and unspecified levels of gene-culture mixes in them. Selection is happening first at the cultural level, but over long periods of time, it is resonating with other sources of selection at the gene level. The interface between gene's and culture is under-specified, and I suspect the main point of contention with Massimo. I would just point to the clear historical trend for larger polity sizes over human history and point out that cultures with long histories at large polity sizes tend to fare much better in large civilizations than do individuals with sparse genetic history at a large polity size.

The default explanation for this is, of course, drift. But it could also happen by cutting off the low z tails in a pro-social distribution (via war, famine or other periodic event), or by selecting for high z tails via increases reproductive success (the Genghis Khan progeny idea) or by increased survivability (the rich people tended to have more kids that survived idea).

I think I've already mentioned enough to give a rough idea on how this may have developed.

Sunday, June 23, 2019

Reframing the Fabric of Reality for Complex Systems Modelling

Joe Norman had a tweet about a rather ambitious project in complex systems theory

 I see three big muses with respect to this endeavor
  1. Charles Peirce's pragmatics as interpreted from a modern (systems) complexity frame
  2. Stuart Kauffman's Reinventing the sacred
  3. Reformulation of "reality's" fabric, like Einstein did moving 3-space into 4-space, or Peirce (and others) did for the concept of infiinity.

Peircean Pragmatics

I'm no expert on Peirce so I won't go into any detail here. Philosophers are always a good muse for scientists. To often we get caught up in details and start forgetting about bigger metaphysical questions.

Peirce's main contribution here might be on the functional irreducibility of precision when time/effort/tools/community are considered. At some point knowledge (hence system framing) progress yields increasingly marginal benefits. Thus full system framing of complex systems faces non-linear effort/informational bounds. A 1% increase in frame details may require 1000x as much work as the previous 1% change. The degree of complexity/chaos (Lyapunov exponent) characterizes this difficulty (up to a point at least). The problem is this fractality is theoretical and doesn't account for interpretational divergence. This is where things really get crazy (as post-modernists tend to attest)

Intuition & Forward Prediction

Kauffman did a bang up job of things in Reinventing the Sacred. However, it was not well received. I just don't think people were ready to think about the implications of moral based forward intuition. Nor could most people get over the baggage associated with anything that came across, explicitly or implicitly, religious. Acadamia's recent acceptance of the science of religion and the role "irrationality" plays in social systems should increase acceptance of this work.

Kauffman's main point as I remember it was that rationality and science is fine for backward looking explanations, but is ill suited for forward looking things at anything other than micro-time scales and things with low to no complexity.  Once you start getting into the edge of chaos, intuition becomes much more accurate. While I don't think he ever states this, part of the reason is transformation and the Hawthorne effect (things being studied tend to improve). Just having something in mind can alter small factors enough to produce significant outcome differences. Morality (or its equivalent) becomes an excellent frame from which to analyze futures deep in the edge of chaos.

This has lots of implications for complex systems framing. It implies that you have to endogenize human morality or irrationality into your fabric of "system reality".  But, you quickly hit complexity's wall of small factor effect indeterminancy - you never know what will and won't matter.

Reforming the Fabric of Reality

I think that leaves complex system folk, like Joe Norman, having to think about some very significant meta-physical questions. How do you quantify and model process meta-physics, especially when emergence is a significant reality?

I'm not sure this can be done without thinking about a new frame/fabric upon which to place social dynamics.

What would this fabric need to have or to endogenize at the fabric level?
  • gene-culture co-evolution
  • human quasi-irrationality
  • morality's (or equivalent's) small effects influence (as per Kauffman)
  • Peircean pragmatics, particularly the interative triangulation of precision with non-linear effort costs
  • some sense of indivisibility. Peirce's (re)conception of infinity as a non-Cantor set seems to be an example to model. Process meta-physics is one example of how this was operationalized. I suspect something equally as radical is needed for complex systems framing....