The Search for the Laws of Self-Organization and Complexity
By: Stuart Kaufman
This book describes the discovery of (self-)organization in nature and its role in the emergence of (the evolution of) life.
How does order arise in nature? According to Darwin, natural selection driven by random mutation is the only source of natural order. Such order may seem accidental: if we start the universe again, different forms of order could emerge.
The central argument of this book is that order is not random or accidental. Out of all possible forms, only very few forms of emerging order have the proper balance of stability and adaptability.
New forms of order emerge spontaneously as parts of a system self-organize into new entities. Natural selection then picks the fittest of these newly self-organized entities.
This book describes the general properties of self-organization and under what conditions it emerges.
The book is hugely ambitious, pioneering, old, but not dated; an interesting mix of spacey philosophy and very dense math and science. A lot of ground is covered. Among the many examples in the book, it would have been great to see it tackle the brain as a self-organizing system as well, which would have introduced a system’s capacity to learn (change its connections and connection strengths). Also, it would have been interesting to explore the concept of entropy in the context of self-organization (as per “The Vital Question“).
- You can understand, but not predict self-organizing systems.
- Quantum uncertainty, chaos theory, theory of computation.
- Chaos theory:
- Any small change in a chaotic system can and typically will have large and amplifying effects.
- Initial conditions will have to be known to infinite precision to predict any outcome.
- Unlikely we can be sufficiently precise, so we can’t accurately predict long-term behavior.
- Power law systems versus normal distribution systems.
- Normal distribution:
- Average has a lot of information value.
- Average gets more accurate with more data.
- Power law:
- Average has very little information value.
- Each new event is unpredictable.
- There are no peaks.
- Average “just” gets higher with more data (see also “Ubiquity“).
- Normal distribution:
- Only systems on the border between stability and chaos are able to survive and evolve.
- Too rigid = no adaptation to random mutations in the environment.
- Sufficiently rigid = withstand small changes and survive.
- Too chaotic = no stability, any random mutation may trigger a reaction.
- Sufficiently chaotic = adapt to small changes and evolve.
The spontaneous emergence of order in complex systems is investigated and explained in four major areas :
- The origin (and subsequent evolution) of life.
- As a collective emergent property of complex systems of chemicals.
- The development of fertilized eggs into an adult.
- As the emergent property of complex networks of genes.
- The behavior of co-evolving species in eco-systems.
- Resulting in avalanches of speciation and extinction.
- Non-natural complex systems.
- The evolution of technological innovation and the creation/destruction of clusters of industries.
The principles of self-organization are:
- A system forms.
- Potentially connected parts.
- A system grows.
- The number of parts grows.
- The number of connections between them grows exponentially.
- Catalysts emerge.
- If a system is big enough, “catalyst” parts may emerge.
- Catalysts increase the likelihood of other parts connecting and reacting.
- Slow reactions become fast reactions.
- Catalytic sets emerge.
- If there are enough catalysts, strings of connected catalytic reactions emerge.
- Auto-catalytic sets emerge.
- If there are enough catalytic sets, an auto-catalytic set may emerge.
- Parts that are formed in one of the reactions become catalysts to their own formation.
- End state: a stable, closed loop system.
- Able to maintain itself and reproduce itself.
- Given a steady supply of input parts.
Through the use of mathematical concepts, network theory and the application of simulation games, the book demonstrates that there are good and bad end states for a complex system to end up in.
Some end states are too rigid and some are too chaotic. If the end state is too rigid, systems are not able to adapt to changes in the environment or random mutations. If the end state is too chaotic, the smallest external or internal change has the potential to destabilize the complex system.
Only systems that are on the border between stability and chaos are sufficiently stable to withstand small changes, but also has the potential to evolve. These are the self-organizing systems that selection acts upon. Evolution is not random, but selects from a small pool of sufficiently stable orders. (See also “The Vital Question“, where the evolutionary landscape is similarly restrained by the structural energy demands of (living) systems).
Using the theory of “fitness landscapes” and the ability of complex systems to explore such landscapes looking for the highest fitness peaks, the book explains (mathematically) why initially, evolution moves very rapidly and changes are very drastic and wide and why it slows down and becomes more narrow over time. This applies to evolution in natural systems, as well as non-natural systems (such as technological innovation: initially fast and wide-ranging, then slow and narrow.)
The theory is also applied to the next level of evolution, how species interact within an eco-system, showing that eco-systems have a tendency to self-organize for a variety of species and connections between them that allows for the highest fitness and lowest rates of extinction. It also shows that co-evolving systems are subject to the power law, where the size of any speciation or extinction event is unpredictable and any small change can trigger a large avalanche of events. Similarly, in non-natural systems, one innovation can trigger the creation and destruction of a small or large number of related industries.