Do Complex Adaptive Systems Do Real Work?

We're interested in theories that deal with the emergence of organization. More practically, we're interested in design heuristics for coordinating lightweight semi-independent processes without the use of centralized controllers. These types of systems are understood to be self-organizing and are said to exhibit "swarm intelligence".

We are pursuing an intuition that the origin, growth and adaptivity of organization can be understood through the application of Stuart Kauffman's Autonomous Agent theory.

"What must a physical system be such that it can act on its own behalf?" An autonomous agent, or a collection of them in an environment, is a nonequilibrium system that propagates some new union of matter, energy, constraint construction, measurement, record, information and work. It is a new organization of process and events."
Stuart Kauffman, Investigations, page 107

Alternatively, an Autonomous Agent is a collectively autocatalytic system that performs one or more thermodynamic work cycles. Autonomous Agents do the following:

  • measure useful displacements from equilibrium from which work can be extracted
  • discover devices to couple to those energy psources such that work can be extracted
  • apply work to develop constraints to extract work

To learn more about Autonomous Agents, view a web video lecture given by Stuart.

Now, what does this esoteric mumbo jumbo have to do with programming software agents? Well, one problem that immediately confronts developers when designing self-organizing software systems is how to design the agents to coordinate their actions to realize a useful emergent system. Some call this the inverse problem of agent design.

Papers

Thorp, J., Guerin, S., Wimberly, F., Rossbach, M., Densmore, O., Agar, M., Roberts, D. (2006) Santa Fe On Fire: agent-based modeling of wildfire evacuation dynamcis . in Proceedings of the Agent 2006 Conference on Social Agents: Results and Prospects, ANL/DIS-06-7, ISBN 0-9679168-7-9, Sallach, D.L., C.M. Macal, and M.J. North (editors), co-sponsored by Argonne National Laboratory and The University of Chicago, September 21-23.

Joyce, D., Kennison, J., Densmore, O., Guerin, S., Barr, S., Charles, E. and Thompson, N. (2006). 'My Way or the Highway: a More Naturalistic Model of Altruism Tested in an Iterative Prisoners' Dilemma'. Journal of Artificial Societies and Social Simulation 9(2) <http://jasss.soc.surrey.ac.uk/9/2/4.html>.

Agar, M., Guerin, S., Holmes, R., Kunkle, D., (2004). Epidemiology or Marketing?
The Paradigm-Busting Use of Complexity and Ethnography. In: Proceedings of Agent 2004:Challenges in Social Simulation

Guerin, S. (2004). Peeking into the black-box: Some art and science to visualizing agent-based models. Proceedings of the 2004 Winter Simulation Conference R .G. Ingalls, M. D. Rossetti, J. S. Smith, and B. A. Peters, eds.

Gambhir, M., Guerin, S., Kauffman, S., Kunkle, D. (2004) Steps toward a possible theory of organization. In: Proceedings of International Conference on Complex Systems 2004. Boston, MA.

Guerin, S. and Kunkle, D. (2004) Emergence of constraint in self-organizing systems. Journal of Nonlinear Dynamics, Psychology, and Life Sciences, Vol. 8, No. 2, April, 2004.

Gambhir, M., Guerin, S., Kunkle, D., and Harris, R. (2004) Measures Of Work in Artificial Life. submitted for publication.

Boyle, S., Guerin, S., Pratt, J., and Kunkle, D. (2003). Application of agent-based simulation to policy appraisal in the criminal justice system in England and Wales. In: Proceedings of Agent 2003:Challenges in Social Simulation

Boyle, S., Guerin, S., and Kunkle, D. (2006) . An Application of Multi-Agent Simulation to Policy Appraisal in the Criminal Justice System. In Chen, S. H., Jain, L., and Tai, C. C. (Eds.), Computational Economics: A Perspective from Computational Intelligence. Hershey, PA : Idea Group [book chapter]

Previous Applications/Demos

Below is a loose collection of proof-of-concepts and software experimentations. None of these 'scratchpad' applications are presented as functioning wholes. Please feel free to explore but don't be suprised by demos lacking instructions or obvious interaction paths.

 

SwarmEffects
SwarmEffects

RedfishGroup Project

Explores flocking behavior in the presence of environmental effectors.
PeerPhoto
PeerPhoto

RedfishGroup Project

Explore some artifacts from research. (Currently no instructions or obvious interactivity)

Application to facilitate the sharing and archiving of consumer digital photos. All files are stored and duplicated on a peer network. There are no centralized servers. The developed algorithms employ distributed software agents that use only local information. Through the agents' interactions, coordinated network behavior emerges without a centralized controller. The peer network is adaptive to patterns in user demand through the evolution of the network's topology and edge weights. Data files have autonomous agency and follow gradients in the network established by concentrations of user load. Data movement toward the gradient's source reduces perceived data latency. The algorithms provide robustness against node failure. Ant algorithms are used in the search for files.

demonstrates a self-organizing peer2peer network that adaptively allocates resources of disk space, cpu and bandwidth.

 
MOTH - Iterated Prisoner's Dilemma

MOTH

Is it more strategic to know when to leave a relationship than it is to know how to behave within one?

This applet extends the standard iterated prisoner's dilemma to include the choice to leave a relationship. The strategy MOTH (My way Or The Highway) beats Tit-for-Tat in early tests with a limited sampling of the strategy spaces of partners.

Interestingly, MOTH is dominant over Tit-for-Tat while both strategies are of equal complexity. MOTH unconditionally cooperates and conditionally leaves when defected against. Tit-for-Tat, conversely, conditionally defects when defected against and unconditionally stays.

Explore this Netlogo Applet by Owen Densmore extending the earlier work by David Joyce, John Kennison, and Nicholas Thompson.

WebEconomy
?
 WebEconomy

under development. Web P2P version of Jim Herriot and Bruce Sawhill's NumEconomy. Check out Partecon.org

I believe such object-based economic web tools will come into existence in the near future. And I believe that such tools will be very powerful means of coordinating activities within supply chains and within the larger economy when linked by automated markets.
-Stuart Kauffman, Investigations, page 228

 
ShapeShifter

ShapeShifter

requires password.

Created for BiosGroup: Proof-of-concept for Honda R&D. Includes user-directed evolution of 3D shapes on consumer-level machines. Demonstrates crossover and mutation of shapes. Enables consumer-designers to share potential car shapes with other users via p2p network. Prototype app.

demonstrates GA's to search high dimensional design space, Peer2Peer and the deployment of identical codebases in screensaver and web browser context. The Shapeshifter project was mentioned in Clippinger's Biology of Business
NetworkModel
NetworkModel requires password.

Created for BiosGroup: 3D visualization of a network flow algorithm. Indicates capacities, transactions and optimization. The challenge was to maintain discrete transaction granularity which traditional network flow algorithms tend to aggregate into homogenous, continuous flows. Transparency of the algorithms mechanism was an important tool for a large energy company to communicate rerouting and reductions of pipleline transactions to customers and regulating bodies. Prototype

may be expanded as core visualization tool. Demonstrates distributed model-view-controller software pattern.

Additional areas of research are detailed on Daniel Kunkle's page.

Below are some demos that may serve as future components to an AutonomousAgent software implementation. Or, they may just be eye-candy. Either way, enjoy.
Visualization components
PartyBox
PartyBox

Emergent behavior from interacting agents with low-level rules. Demonstrates contribution of noise to reduce frustration and the establishment of a dynamic equilibrium.

experiment to generate 3D agent based model
GraphLayout
GraphLayout

General 3D graph layout with springs and mass nodes. Demonstrates self-organization to minimum energy states.

Distributed model-view-controller
ShapeGA
ShapeGA

Feasibility study of real-time 3D deformation, evolution and rendering in a web browser for ShapeShifter.

Genomes specifying 3D shape properties are randomly generated. Offspring are created with mutation and crossover.

SphereCA
SphereCA

A two-dimensional cellular automata on the surface of a sphere.

experiment to getNeighbors on 3D surface
Organizational Design

Research relevant to Organizational Design:

  • Can we model a firm or supply web as an open, far-from-equilibrium organization? What are the relevant flows? What are the conserved quantities?
  • Can we address the coupling of reversible processes and irreversible processes in adaptive organizations?
  • How can we measure or describe symmetry breaks in causal relationships between agents?
  • What does it mean for individuals to lose degrees of freedom as the system self-organizes?
  • Can we identify a calculus that allows for the emergence of constraints? Pi? Fusion?
  • How can we scientifically approach describing a firm's or supply web's identity - that which maintains and adapts itself by changing its structure?
  • In what contexts is it appropriate for these organizations to be self-organizing and in what contexts are command and control organizations more desirable?
  • How can distributed software tools provide mechanisms of entrainment between agents within and between organizations.
  • If an autonomous agent is a self-organizing system that can use information to adapt, what might a corporate information system look like to support these processes?

We suspect that firms and supply webs can be understood as autonomous agents that self-organize toward emergent levels of intention beyond the aggregate intentions of its lower level agents. This intention can arise via emergent dynamics coupling with new contextual environments at higher spatial and temporal scales. This coupling can provide top-down constraints that selects among the possible bottom-up emergent states (Juarrero, 1999). Within this framework, we try to understand what it means for a firm or supply web to act on its own behalf.

We believe that economic organizations are, to varying degrees, self-organizing social systems. Traditional organization management science is grounded in the paradigm of the organization as machine. As a result, organizations tend to be externally designed and mechanistically regulated via command and control hierarchies. While efficient, these organizations have difficulty adapting in environments of great change. Our goal is to develop networks of software agents that can act as generative tools to transform the management of firms and supply webs from command and control organizations to self-organizing adaptive systems. Increasingly, these agents will be used to regulate resource assignment both within and between firms.

Reading List/Links (no particular order)

Jain and Krishna. Graph Theory and the Evolution of Autocatalytic Networks (pdf)

Walter Fontana, and Leo W. Buss. Barriers to Objects: From Dynamical Systems to Bounded Organizations In: J. Casti and A. Karlqvist eds. Boundaries and Barriers, pp.: 56-116, Addison-Wesley, 1996.

Bonabeau, Eric. Agent-based modeling: Methods and techniques for simulating human systems PNAS Vol 99 Sup 3 (pdf)

Smith, Eric and Foley, Duncan. Is Utility Theory So Different From Thermodynamics? Santa Fe Institute Working Paper, 2002

Kauffman, Stuart. Investigations, Chapter 3 - Autonomous Agents and Chapter 4 - Propagating Organization
(a web preprint of Investigations)
(video lecture of Autonomous Agents)

Kugler, Peter and Turvey, Michael. (1987). Information, Natural Law, and the Self-Assembly of Rhythmic Movement Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. -- Out of Print

Schneider, Eric D. and Kay, James J. Order from Disorder: The Thermodynamics of Complexity in Biology

Schneider, Eric D. and Kay, James J. Life as a Manifestation of the Second Law of Thermodynamics

Economics and Social Interactions: State and Market Formations. SFI Research Focus Area

Network Dynamics. SFI Research Focus Area

Guerin, S. and Kunkle, D. (2004) Emergence of constraint in self-organizing systems. Nonlinear Dynamics, Psychology, and Life Sciences, Vol. 8, No. 2, April, 2004.

Parunak, H. Van Dyke. Go to the Ant: Engineering Principles from Natural Multi-Agent Systems (pdf)

Heylighen, Francis. The Science of Self-Organization and Adaptivity (pdf)

Juarrero, Alicia. Dynamics in Action - Intentional Behavior as a Complex System Chapter 9 - Constraints as Causes

Parunak, H. Van Dyke and Brueckner, Sven. Entropy and Self-Organization in Multi-Agent Systems (pdf)

Cefn Hoile's research site at BTexact