contact information

Dr. Ted Carmichael

Department of Software and Information Systems

University of North Carolina at Charlotte

1612 Gordon Walters Dr., Charlotte, NC 28213

Email: tedsaid@gmail.com

links

Last modified

Complex Adaptive Systems

Complex Adaptive Systems (CAS) is the study of simple rules and agents that produce amazingly complex behavior. The agents can be almost anything: from ants or bees, to people, to the cells found inside plants and animals. Thousands of these actors interact, influencing each other and being influenced in turn.

The aggregate of all these interactions are the emergent, system-level structures and behaviors that we observe, both in society and in nature. Even intelligence and human consciousness are emergent features of the billions of neurons and their synaptic interactions.

We define a CAS as having a large number of self-similar agents that:
  1. Utilize one or more levels of feedback;
  2. Exhibit emergent properties and self-organization; and
  3. Produce non-linear dynamic behavior.
Advances in modeling and computing technology have led to a deeper understanding of complex systems in many areas, and have raised the possibility that similar fundamental principles may be at work across these systems, even though the underlying principles may manifest themselves differently.

My research focuses on the multidisciplinary nature of CAS, and all of my efforts in this regard are to advance two primary goals:
  • Goal #1: To improve our understanding of general CAS principles that transcend any one domain
  • Goal #2: To help build a community of CAS researchers that span across the natural, physical, and social sciences

Goal #1: CAS Research

We have undertaken a number of research projects that aim to improve our basic understanding of CAS principles and dynamics across all domains. The first of these is a general CAS model that is the result of my dissertation research.

The general model has two types of agents: the patches and the turtles. The patch main attribute is its current state, which is either the 0-state, the 1-state, or somewhere in between. Once a patch reaches the 1-state, it can affect neighboring patches towards the 1-state.

The turtles move randomly across the environment, and can affect patches towards the 0-state. Thus, the main dynamic of this system is between the 1-state patches affecting other patches, and the turtles affecting patches in the opposite direction.

The turtles have a limited lifetime, and can generate new turtles based on a positive function of how many patches each affects. This allows the system to produce a basin of attraction - a powerful negative feedback mechanism for sustainability of the system.

We applied this model to soft-tissue cancer and the immune system response. In this mapping, the patches are the cells that are either healthy (0-state) or cancerous (1-state). The immune cells are represented by the turtles. Each cancerous, 1-state cell can affect it's neighboring 8-cells (known as the Moore neighborhood), allowing the cancer to spread exponentially. The immune cells, if they are effecient enough, can surpress the cancer's growth.

Our next step was applying this same model to an entirely different domain: political dissent in a polity. The patches are citizens whose main attribute is their feelings toward the government. Dissenters can push other citizens towards dissent, while the government agents (turtles) influence these citizens in the opposite direction.

The full dissertation can be viewed here. A shorter, 10-page summery is available here.

Current and future work involves applying this general model to:
  • Predator/prey dynamics in a marine ecosystem
  • Terrorism and human rights
  • Resource allocation in an economic system
As we develop this model and its domain-specific applications, we will continue to advance our goal of a domain-agnostic general model. Thus, the general CAS model will also serve as a language of CAS. This will not only provide a common modeling toolset for multiple domains, but will also facilitate cross-fertilization of concepts and insights across domains.

Goal #2: CAS Community

Our general CAS tool advances the goal of providing a common language of CAS, regardless of the target domain. Such a general tool is only half a solution; also required is a community of researchers from across domains who are interested in collaborating on the toolset of CAS itself.

Therefore we have endevoured to build such a community. At UNC Charlotte we have established the Complex Systems Institute, a multi-disciplinary center that spans five colleges and ~10 departments.

We have also established, in conjunction with the AAAI (Association for the Advancement of Artificial Intelligence), an international symposia series on Complex Adaptive Systems. These symposia have been widely praised for their intellectual discussions, productive sessions, and diversity of participants. Each symposium has focused on a fundamental property of CAS. The first, in 2009, was on Threshold Effects across the Natural and Social Sciences. The second (2010) was on Resilience, Robustness, and Evolvability.

Our third symposium will be held this November in Arlington, VA. The theme of this year's symposium is Energy, Information, and Intelligence. The website main page gives more information on the significance of these aspects of complex systems, and the Call for Papers (due May 15th) provides examples and submission requirements.

We will also hold, in the summer of 2011, the first NEH Institute on Computer Simulations in the Humanities, a two-and-a-half week workshop on applying CAS and modeling techniques to research in the humanities.

These and other activities illustrate the important strides we have made towards building a multi-disciplinary community of CAS researchers, and we will continue to seek partnerships and opportunities that build on the results of these efforts.