Instructor: John H. Holland
Many of our most difficult contemporary problems depend upon an understanding of systems consisting of agents that adapt and learn – ecosystems, markets, political systems, the Internet, nervous systems, immune systems, reaction networks in biological cells, and so on. These systems, called complex adaptive systems (cas), exhibit properties such as “emergent” structures, “complex” conditional interactions, perpetual novelty in behavior, and diversity in agents (there is no “best” agent). Because of these properties, cas require novel techniques for analysis and understanding. This class will introduce and explore techniques, such as agent-based modeling, that have been most effective in helping us to explore and understand the behavior of cas.
This is a highly interactive class with students from all over campus. You will be expected to contribute to the class discussion and will be graded accordingly. There will be a final paper which you will present to the class.
1) Holland, John H. Hidden Order: How Adaptation Builds Complexity. Cambridge, MA: Perseus, 1996.
2) Holland, John H. Emergence: from Chaos to Order. Reading, MA: Perseus, 1999.
The class aims to develop a range of ideas, examples, models, and intuitions that provide a deeper understanding of cas. All techniques will be fully developed in class, starting from elementary principles. The order of topics will depend partly upon particular interests of the class, but the following topics, at least, will be covered:
• Performance systems [sets of condition/action rules].
• Signal-passing systems - their pervasiveness from cell biology to language.
• Parallelism - systems with many rules active simultaneously.
• Agent-based models (models with multiple interacting agents).
• Credit assignment - strengthening stage-setting and predictive rules.
• Rule discovery - genetic algorithms.
• Building blocks - their role in everything from perception to invention.