The Evolution of Complexity

The Evolution of Complexity:
an introduction
Francis Heylighen
Evolution, Complexity and Cognition
group (ECCO)
Vrije Universiteit Brussel
A Transdisciplinary Perspective
Conceptual scheme applicable to all complex,
evolving systems
•
Particles, molecules, cells, organisms, societies,
galaxies…
Unifying models in all classical disciplines
•
Physics, chemistry, biology, psychology, sociology,
economics, etc.
Requires some simple concepts and
assumptions that are generally valid
Classical science
Characterized by
• analysis
•
reductionism
Focuses on separate
components
Complexity
complexus = entwined,
embracing
•
•
•
distinguishable parts
that are connected
so that they are difficult to
separate
differentiation + integration
in between order and disorder
•
the "edge of chaos"
What is a System?
Distinguishable parts (differentiation)
Connected into a whole (integration)
Distinct from the environment
•
Separated by boundary
Yet, open
•
= interacting with the environment
•
Exchanges across boundary
Emergence
Whole = more than sum of the parts
connections create properties that are not
inherent in the parts
•
emergent properties
examples
•
car: max. speed = emergent, weight = sum
•
music: melody, rhythm, harmony = emergent
•
salt (NaCl): taste, color, shape, ... = emergent
Evolution
Emergence and change of systems over time
Produced by BVSR
•
Blind Variation and
•
Selective Retention
•
of the “fittest” configurations
Fitness = ability to maintain and multiply
•
in a given environment
Evolutionary Progress
“Survival of the fittest” is a tautology
•
what is fit = what survives = what is selected
Logically necessary principle →
•
automatic mechanism, no explanation needed
Assume variation
•
•
Some configurations fitter, some less fit
Fitter ones are preferentially retained →
Fitness tends to increase
3 ways to achieve fitness
1. Intrinsic robustness/stability
•
E.g. a diamond
2. Adaptedness
•
“fitting” in to a specific environment
•
E.g. koala in eucalyptus forest
3. Adaptivity
• Flexibility, ability to adapt to a variety of environments
•
E.g. humans
Each leads to different types of complexity
Co-evolution
System + Environment is too simple
•
The environment is much too complex to be reduced to
a single influence
Better: interacting agents
•
Agent= (relatively) autonomous system
• E.g. molecule, cell, organism, person, firm
Agents undergo variation and selection in an
environment of other agents
• Change in one agent requires adaptation in the agents it
interacts with
•
→ On-going, mutual adaptation
Emergence of Networks
Two Agents interact
•
Mutual variation and selection
•
Until they reach a fit configuration
•
•
Reciprocal adaptation
→ creation of bond, link or coupling
Many agents developing many links → network
System as Network of Agents
E
f
k
l
c
e
j
h
I
O
a
i
d
g
b
S
Formation of Bonds
Two systems encountering each other may
develop a stable connection or bond
e.g. Two atoms forming a molecule
Two people forming a couple
Formation of Bonds
•Many agents may get linked together,
forming a system or “superagent”
•Superagents in turn get linked together
forming a “super-super-system”
•This produces structural complexity
Differentiation and Integration
linked components are integrated into new whole
non-linked components are more strongly
differentiated
Selforganization
of Hierarchies
Growth of structural
complexity
Evolution of adaptivity
Individual agents too tend to become more
complex
•
By increasing their adaptivity
Adaptivity achieved by control or regulation
•
Compensating “perturbations” (changes in
environmental conditions)
•
by appropriate actions
E.g. chameleon compensates changes in
background color by changes in skin color
Law of requisite variety
The larger the variety of perturbations, the
larger the variety of actions the agent should
be able to perform (W.R. Ashby)
•
A complex, variable environment demands a
large repertoire of actions
However, the agents must choose the right
action for the right condition
→ law of requisite knowledge
agent must “know” appropriate rules
of the form:
condition → action
Functional complexity
Control laws → selective pressure for:
•
More variety of action (functional differentiation)
•
More knowledge rules to connect conditions and
actions (functional integration)
→ growth in functional complexity
Growth in ability to deal with complex problems
→ growth in agent “intelligence”
Combining structural and
functional complexity
Agents develop links → structural complexity
But become more adaptive in their actions →
functional complexity
Becoming collectively more adaptive requires
not bonds (“hard” connections), but
coordinated actions
Actions that together achieve more than alone:
synergy, cooperation
Example: office organization
Coordination mechanisms
Alignment of targets
Avoiding conflict or friction
Division of labor
Differentiation or specialization of agents
Workflow
Actions performed in right sequence
Aggregation of results
Regulation
Correcting errors via feedback
Self-organization
spontaneous appearance of
order or organization
not imposed by an outside
system or inside
components
organization distributed over
all the components
•
collective
•
Robust
QuickTime™ and a
TIFF (Uncompressed) decompressor
are needed to see this picture.
Self-organization of
coordination
Stigmergy
•
Trace left by action stimulates performance of
subsequent action
•
Examples
•
Ant pheromone trail laying
•
Wikipedia
Hebbian learning
•
•
Successful sequences of actions are reinforced
Unsuccessful ones are weakened
Conclusion
Variation and selection automatically increase fitness
• which indirectly increases complexity
Fitness can be achieved via
•
Stable bonds → structural complexity
• → Hierarchies of supersystems
•
More adaptive agents → functional complexity
• → Evolvability and individual intelligence
•
More coordinated actions → organizational complexity
• → Collective intelligence, “social” systems