Understanding Complexity and Emergence
After reading â€œCritical Massâ€? by Philip Ball and delving into the mechanics of phase transitions, I wanted to explore the current academic discourse around the study of complex systems and emergence. To that end, I read several articles from the journal â€œComplexityâ€? dating from 2002 through the most current 2007 issue.
Corning, Peter A. â€œThe Re-emergence of â€˜Emergenceâ€™: A Venerable Concept in Search of a Theory.â€? Complexity 7.6 (2002): 18-30.
Chu, Dominique, Roger Strand and Ragnar Fjelland. â€œTheories of Complexity: Common Denominators of Complex Systems.â€? Complexity 8.3 (2003) 19-30.
KlÃ¼ver, JÃ¼rgen. â€œThe Evolution of Social Geometry: Some Considerations about General Principles of the Evolution of Complex Systems.â€? Complexity 9.1 (2004) 13-22.
HÃ¼bler, Alfred W. â€œUnderstanding Complex Systems: Defining an Abstract Concept.â€? Complexity 12.5 (2007) 9-11.
Schuster, Peter. â€œA Beginning of the End of the Holism versus Reductionism Debate?â€? Complexity 13.1 (2007) 10-14.
Background on the authors:
Peter Schuster: is current editor of the â€œComplexityâ€? journal and has been affiliated with the Santa Fe Institute. He is at a university in Austria in the field of theoretical chemistry.
Peter Corning: has served as the director of the non-profit Institute for the Study of Complex Systems and as a founding partner of a private consulting firm in Palo Alto California. His field is behavioral genetics.
Alfred Hubler: is at the Center for Complex Systems Research, Department of Physics, University of Illinois at Urbana-Champaign.
Jurgen Kluver: is a professor of Information Technologies and Educational Processes at the University Duisburg-Essen. His fields of research include: mathematical and computational sociology, theoretical sociology, sociology of science, theory of science.
Dominique Chu: is an academic fellow at the Computing Laboratory at the University of Kent in Canterbury. His interests include bio-inspired computing, computational biology/computational modeling of biological systems, molecular computation.
A recurring theme is that a combination of â€œreductionistâ€? and â€œholisticâ€? approaches is seen as necessary to understanding the phenomenon of emergence and the evolution of complex systems in nature. The term â€œreductionismâ€? once meant an understanding of the â€œpartsâ€? of a system, while â€œholismâ€? implied something almost mystical that could not truly be understood. These terms have taken on different meanings in the study of emergence. Peter Schuster makes these points succinctly in the most current issue, while Peter Corningâ€™s 2002 article also makes a more detailed distinction between the two approaches:
Reductionist approach to understanding complex systems (also described as â€œsystems scienceâ€?):
1. Understood by studying interactions between particles
2. Systems are deterministic yet unpredictable
3. Particles are seen as having inherent tendency to self-organize
4. There is a search for fundamental laws that explain behavior (but underlying causal agency is not specified)
5. Treats emergence as an â€œepiphenomenaâ€? (resulting from interactions, but having no causal effect)
6. Explains â€œhowâ€? systems work
Among others, Corning labels the following â€œreductionistâ€?: BarabÃ¡si, Kauffmann, Holland, Buchanan
Presumably, the study of phase transitions and the modeling applications described in â€œCritical Massâ€? by Philip Ball would also be labeled reductionist in the sense that Corning describes it.
- System understood as multi-leveled. Causation is upward, downward, and horizontal.
- Effects may be co-determined by the context and the interactions between the whole and its environment
- Causation is iterative â€“ synergistic effects of interactions are also causes of other effects
- New emergent properties arise at higher levels of organization
- Properties of the parts are modified, transformed, reshaped by their participation in the whole
- Organized, purposeful activity: instruction-driven as opposed to law-driven (e.g. genetic code)
- Attempts to explain â€œwhyâ€? evolution occurs
Corning labels the following â€œholisticâ€?: Casti, Corningâ€™s own â€œSynergism Hypothesisâ€?
Kluver is looking for general principles that determine the laws of evolution, particularly in complex social systems (a reductionist approach, per Corning). He makes a helpful distinction between different types of system dynamics. This relates to my blog posting earlier in the semester about the different classes of computer models: (System Dynamics, Agent-based)
First-order dynamics: the kind of dynamics that a system exerts by changing its states but not changing its rules of interaction. (Example: boids)
Second-order dynamics: An adaptive system is characterized by second-order dynamics, i.e., a dynamics, that a system generates by changing its rules of interaction according to environmental demands. (Example: Forresterâ€™s â€œSystem Dynamicsâ€?)
Third-order dynamics: combines both features of ï¬?rst- and second-order dynamics: like ï¬?rst-order dynamics it unfolds by its own logic; like second-order dynamics it takes environmental demands into account and changes its own rules of interaction; yet in addition it is also able to vary its own structural initial conditions, which started the whole process. A system capable of third-order dynamics can deï¬?ne its own criteria of success and thereby change its environment. (Example: Agent-based modeling)
Chu and co-authors Strand and Fjelland discuss the intrinsic limitations of computer and mathematical modeling techniques. This seems to echo Corningâ€™s discussion of reductionist and holistic approaches:
Real systems in nature have the property of â€œradical opennessâ€? resulting from the rich connections between real systems and their environment. In contrast, computer models are closed. In order to construct a workable model, scientists select a relatively small number of system elements deemed relevant and formalizes these elements into mathematical equations or computer code.
Thus, there is a family of overlapping possible and actually realized models. This is the source for what is called â€œcontextualityâ€?. A system is â€œcontextualâ€? if it includes one or more elements that also occur in a different system, or if it is itself a shared element between more than one system. In other system(s) the shared elements take part in causal processes different from those included in the original system. The property of contextuality is a consequence of the partitioning of the world into system and environment that precedes any modeling enterprise. Likewise, global models will not be contextual.
Hubler talks about the production of new knowledge in the study of complex systems. He is critical of current research, in that â€œbecause of the traditional preference for abstract work, abstract research results with very little experimental grounding are being published at an ever accelerating rate, whereas experimental work receives comparatively little attention and funding. This raises the question how much knowledge is being created by current complex systems research.â€? He is concerned that current research is not developing the network of concepts that is necessary to understand complex systems holistically. He concludes that a â€œpracticalâ€? understanding of complex systems and practical applications of research are not likely to be developed in the near future.