The Dynamical Basis of Social
Emergence
John Collier
Philosophy, University of KwaZulu-Natal
collierj@ukzn.ac.za
https://web.ncf.ca/collier
Bertalanffy Center for the Study of Systems Science
November 25, 2014
Outline
• Three problem areas for social emergence
• Some misconceptions about emergence
• The problems in more depth and the nature of
emergence
• Summary on emergence
• Summary on social emergence
• Some topics for discussion
The problems: 1. Weak and strong
emergence
• Emergence is recognized as coming in two forms:
weak and strong (Collier, Bedau, Chalmers)
• They have very distinct properties, and should
not be confused.
• Weak emergence is reducible and fully analyzable
(all computational models so far, flocks, game of
life, etc.)
• Strong emergence is not reducible, and is not
computable (so far). This is traditional emergence
in 19th Century and 20th Century philosophy (J.S.
Mill, C.D. Broad).
The problems: 2. Qualitative
characterization
• Emergence is typically described in qualitative (or in
logical) terms: novelty, irreducibility, whole greater
than sum of parts, unpredictability, self-organized, selfinteracting, etc.
• However, we interact with things through forces and
flows, i.e., dynamically. We often ignore this in
everyday life (naïve realism), but in science we need to
pay attention to how we detect things in order to
understand what we detect.
• So we need a dynamical characterization of emergence
to be able to test for it by interactions.
The problems 3: Boundary condition inseparability
from system laws is diagnostic of emergence
• In classical mechanics and quantum mechanics system laws and
boundary conditions are typically assumed to be independent of
each other. This is always true if the system is a conservative
system, typically called a Hamiltonian system. All QM systems are
Hamiltonian systems. (Technically, Hamiltonian systems are a
broader class.) None of these systems can show strong emergence.
• They are also called holonomic, meaning their variables depend
only on location and perhaps time, but not velocity.
• Near Hamiltonian systems can be approximated with perturbation
theory at one end of the scale, or step functions at the other end of
the scale, but in the middle predictability fails, computability (even
approximate) fails, self-organization is possible, novel properties
appear, and , of course, the system is strongly self interacting, as
system laws and boundary conditions are not separable.
The problems 3 continued: Diagnosis
• All known such systems are dissipative.
• But we typically don’t have a detailed
mathematical description of the dynamics of
biological and psychological systems, let alone
social systems, so how can we determined if this
diagnostic dynamical condition occurs?
• Typically, we fall back on qualitative properties,
but this leaves us open to errors in judgement.
• What is dissipated?
Some mistaken ideas about emergence 1
There are some mistaken ideas of emergence that are
relevant to the three problems for understanding and
testing social emergence
• Mystericism (separate substances, causal independence, and
teleology) ̶ two critics once objected that my account of
emergence makes it too simple, comprehensible, which they
thought begged the question
• Emergence is purely epistemological, possibly resulting from
limitations on our minds. Only nominal. Eliminable in principle
if we think differently.
• Emergence is a property of our models of the world, not of
the world itself. Suggests that with the right models,
emergence would disappear. If only it were that easy!
Some mistaken ideas about emergence 2
•
•
•
Emergence results from complexity alone, (Bar-Yam 2011): “In conventional views
the observer considers either the trees or the forest. Those who consider the trees
consider the details to be essential and do not see the patterns that arise when
considering trees in the context of the forest. Those who consider the forest do
not see the details. When one can shift back and forth between seeing the trees
and the forest one also sees which aspects of the trees are relevant to the
description of the forest. Understanding this relationship in general is the study of
emergence.”
He goes on to give an example: “Consider a key. A description of a key's structure
is not enough to show us that it can open a door. To know whether the key can
open a door, we need descriptions of both the structure of the key and the
structure of the lock. However, we can tell someone that the function of the key is
to unlock the door without providing a detailed description of either.”
Emergent computation is more than merely a simulation of emergence.
Computation that terminates is by definition computable and hence reducible. As
a corollary, numerical models of strongly emergent systems are necessarily not
emergent themselves, and thus are misleading.
Problem 1: Weak versus strong emergence
• The distinction is useful to make between properties (or systems) that
must be understood as a whole, including their large scale interactions but
are nonetheless reducible (weak emergence) from those that are not
reducible (strong emergence). The NECSI version of emergence is the
weak sort.
• The distinction has been made by (among others) myself and Scott Muller
(1998), Mark Bedau (1997), David Chalmers (2006). Note that Chalmers’
usage is divergent and applies only to mental qualia.
• In 1986 I introduced the notion of cohesion to cover dynamical unity in
general. Cohesion is the property that makes a dynamical unity of parts of
something. The cohesion in strong emergence is not reducible.
• Cohesion applies to properties as well as systems (Every Thing Must Go).
All systems are individuated by their cohesion. If the cohesion of the
system is an emergent property (weak or strong, respectively), then the
system itself is emergent (weak or strong, respectively). So property
emergence is more fundamental than system emergence.
Paul Humphreys and “fusion”
• Many theorists use an obsolete reductionist ontology of particles and
collisions, which we call (Every Thing Must Go) “microbangings” of myriad
“little things”. This ontology is not compatible with current science.
• Paul Humphreys pointed out that when parts act under each others’
influence many of the properties of the parts no longer exist, the system
forming a “fusion” with properties of its own. For example, a planet
orbiting the Sun has a dynamics that are the result of the interaction of
the planet's properties and those of the Sun, forming a fused dynamics.
These are the real dynamics of the system, not some construct out of the
independent dynamics of the parts of the system and their relations.
• Humphreys recognized that the fused dynamics are possibly reducible, but
the whole is cohesive and has at least the property of weak emergence.
• This idea helps to resist the tendency that there must be some still existing
components with their own individual properties. The system is not the
sum of its components because the components are fused. A
nonreducible fusion will be strongly emergent.
The historical ground of the emergence concept
• Origin of the concept – vitalism debate versus naturalism and the unity of
the sciences
– G. H. Lewes in 1875: something that is incommensurate with its components and not
reducible to their sum or their difference
– J. S. Mill (A System of Logic, Bk.III, Ch.6, §1, 1843) does not use the word: “All organised
bodies are composed of parts, similar to those composing inorganic nature, and which
have even themselves existed in an inorganic state; but the phenomena of life, which
result from the juxtaposition of those parts in a certain manner, bear no analogy to any
of the effects which would be produced by the action of the component substances
considered as mere physical agents. To whatever degree we might imagine our
knowledge of the properties of the several ingredients of a living body to be extended
and perfected, it is certain that no mere summing up of the separate actions of those
elements will ever amount to the action of the living body itself.”
– C. D. Broad (The Mind and Its Place in Nature, 1925): While Emergentists, too, are
physical substance monists (“there is only fundamentally one kind of stuff”), they
recognize “aggregates [of matter] of various orders” — a stratification of kinds of
substances, with different kinds belonging to different orders, or levels. Each level is
characterized by certain fundamental, irreducible properties that emerge from lowerlevel properties. A primary concern of Broad was an ontological deflation and ultimate
unity of both the sciences and of the world.
• Clearly these people were talking about strong emergence.
Problem 2: Logical conditions for (strong) emergence
1.
2.
3.
4.
Non-reducibility of emergent properties
Non-predictability of emergent properties
Holism
Novelty of emergent properties
Problems with this approach: Conditions 1-4 are hard to detect experimentally and difficult to distinguish from
epiphenomenal properties. We need dynamical (interactive) conditions for emergence.
•
•
Humphreys’ fusion is necessary but not sufficient, though dynamical).
Robert Rosen (1991): profound criterion: analytic but not synthetic models
–
–
–
•
Synthetic models are a sum of their parts and pairwise relations, or can be analyzed fully in terms of
their inputs and outputs
Analytic models need not have any corresponding synthetic models. If so, they cannot be reduced to
their parts or input-output relations.
However this is a logical condition of models, not systems themselves.
Wimsatt (2007): non-aggregativity – closer to a dynamical view, but not much
different from Lewes and Mill.
Complex organization and emergence
•
•
•
(Collier and Hooker 1999) The basic
idea of an organized system is that it
is interconnected in complex ways, so
that there are both local and nonlocal effects.
Complex organization
– involves neither summation nor
top down control, but shows an
interaction of bottom-up effects
and top-down effects
– Complexly organized systems
cannot be decomposed
Thus they show the characteristics of
emergent systems
Problem 3: Computational notions connected to
dynamics
• Non-reducibility: A property is nonreducible if it cannot be computed from
additive (or subtractive) combinations of the dynamics of its parts.
• Non-predictability: the trajectory of a system is predictable if and only if
there is a region η constraining the initial conditions at t0 such that the
equations of motion will ensure that the trajectory of the system will pass
within some region ε at some time t1, where the region η is chosen to
satisfy ε. For indeterministic systems, the values are determined to the
extent determined by the probabilistic factors in the laws. Otherwise it is
non-predictable. Failure implies chaotic regions and possible bifurcations.
• Connecting to dynamics: All holonomic systems (constraints are
expressible as functions as spatial coordinates and perhaps time, but not
in general velocity) are reducible and predictable. All traditionally studied
Hamiltonian systems (the subject of most physics) are holonomic.
• So non-holonomicity is required for emergent systems with emergent
properties. The state taken by such systems depends on the path taken.
Separation of boundaries
and system laws
• Conrad and Matsuno (1990)
– Differential equations provide the major means of describing
the dynamics of physical systems in both quantum and classical
mechanics. The indubitable success of this scheme suggests, on
the surface, that in principle it could be extended to a universal
program covering all of nature. The problem is that the essence
of a differential equation description is a separation of itself
from the boundary conditions, which are regarded as arbitrary.
• We cannot do this, in general, for non-holonomic systems.
• The problem is not that we cannot separate laws and
boundary conditions, but rather that the system itself does
not allow this sort of solution. We can still understand the
system in terms of laws and boundary conditions, but this
does not allow us to give full solutions for its behavior.
Necessary conditions for emergence
• Non-Holonomic systems
– Basically, energy (or some central system property) is not
conserved, e.g., dissipative systems.
– Boundary conditions and system laws cannot be separated
in principle in the system dynamics. In particular, the
system can (under some conditions) do work on its
boundary conditions, constructing new properties.
– Near holonomic we can approximate at one end by step
functions, and at the other end by perturbation theory.
– Radically non-holonomic systems, however, are a problem.
I call these radically non-Hamiltonian if the rate of energy
dissipation is of the same order as some system property.
Likely strongly emergent systems
• Mercury and the Sun (3:2 rotation to orbit period)
• Bénard Cells
• A wheel (region of stability, but outside that
unpredictability emerges).
• Cell differentiation (not fully under DNA control)
• Ant behaviour of some sorts
• Non-selective sympatric speciation
• Living systems (the original prototype for strong
emergence)
• The mind and/or its contents (?), Social properties (?)
Problem 3: applying the dynamical conditions to
general emergent systems
• My characterization of emergent systems uses
concepts from physics that are clear in that
realm.
• However, as some of the examples show,
emergent systems are often not easy (or perhaps
possible) to be characterized in terms of physical
equations.
• We need other ways to characterize emergence
dynamically for systems that are not obviously
physical, or physical chemical.
Some possible approaches
1.
2.
3.
4.
We can look at most systems from the perspective of forces and
flows in networks. This is a dynamical representation. If the laws
governing the forces and flows interact then we have a candidate
for emergence. In particular, there are certain types of system
equations that are mathematically impossible to solve.
If system networks of causal processes show closure, such that
inputs depend on outputs, we have a good candidate for an
emergent system or subsystem (Rosen – “closure to efficient
causation”).
Many higher level systems (biological, mental) can be
characterized in terms of information. If information is dissipated
in the system at a similar rate to the rates of one of the properties
on which it depends, then the property is likely to be emergent.
Is there a similar social property? Is it information? Might there be
others? These are all open questions to the best of my knowledge.
Ordinary
Partial
Differential
Differential
Trivial
Easy
Difficult
Easy
Difficult
Intractable
Many Parameters Intractable
Intractable
Impossible
One Parameter
Very Difficult
Very Difficult
Impossible
Very Difficult
Impossible
Impossible
Equation:
Algebraic
One Parameter
Linear
Several
Equations
Parameters
Nonlinear
Several
Equations
Parameters
Summary on emergence in general
• Weak and strong emergence often confused. This
requires clear criteria for reducibility that can be
tested, since the properties of strongly and weakly
emergent systems are significantly different.
• Only dynamical criteria can be tested through
interactions. Logical and qualitative criteria are difficult
to test directly in a reliable way.
• Emergence requires dissipation. Energy is obvious in
physical systems, but information dissipation might
also work. Other properties might also be appropriate
(word usage, meaning dissipation, corporate
dissipation).
Summary on social emergence
• Information dissipation might apply, but how
is not obvious.
• It is especially difficult to distinguish strong
and weak emergence, so care is advised.
• Other candidates for the dynamic basis of
social emergence are not obvious. They may
be multiple and apply in different
circumstances.
Some topics for discussion
• Cultural emergence
– New and integrated trends
– Through interactions of pre-existing cultures
• Emergence of a public sphere that is quasiindependent and hence (interactively)
autonomous from government and private
spheres
• Interaction with environment and other social
systems while maintaining a degree of autonomy.