System Modeling

Systems are often difficult to run experiments on.  Therefore, it is beneficial to model the systems to learn more about how they behave and so you can experiment with them.  The System Modeling Institute uses computer models to learn about systems.  Models do not necessarily have to be good at predicting to be useful.  One is able to learn about assumptions during model development.  In addition, model behavior can demonstrate parameter sensitivity, key phase transitions, and chaotic attractors.

Types of modeling used at the Institute

• Agent based modeling
  • Independent agents interacting with their own rules
  • Good for modeling interactions, especially if not all agents behave the same
  • Often see emergent behavior that isn’t obvious from the rules that the agents use
• Discrete Event Modeling
  • Uses Transactions, service times, queues. Time is discrete and the system is modeled in events
  • Good for performance modeling, finding bottlenecks and transaction based systems
  • Efficient modeling when there are long periods of time with out events followed by a large amount of events
• Dynamic System Modeling
  • Uses Stocks, Flows and Feedbacks and time is continuous
  • Good at modeling large systems at a high level
• Virtual Worlds
  • People interacting in virtual worlds
  • Good for when agent (people) behavior is difficult to predict or quantify
  • You can create a virtual economy or political system and do things to the system that you wouldn’t want to do in the real world.