Agent-Based Simulations vs. Traditional AI

While a valid and useful answer to the first question can often be found, there is at least one significant drawback to this approach so many simplifying assumptions must usually be made about the real system, in order to render the top-level problem a soluble one, that other natural, follow-up questions such as Why do specific behaviors arise or How would the behavior change if the system were defined a bit differently cannot be meaningfully addressed without first altering the set of assumptions. An analytical, closed-form solution may describe a behavior, however, it does not necessarily provide an explanation for that behavior. Indeed, subsequent questions about the behavior of the system must usually be treated as separate problems. 

Any system whose top-level behavior is a consequence of the aggregate behavior of lower-level entities - biological systems, neural systems, social systems, economic systems, among many others - can, in principle, be simulated by an agent-based model. 

Target Open Systems Traditional AI has focused on closed systems in which the interaction between the problem tlomain and the external environment is kept to a minimum in contrast, agent-based systems are open systems, and agents arc directly coupled with their environment. 

Target Multi-Objective Goals Most traditional AI systems deal with problems in a piecemeal fashion, one at a time in contrast, the individual agents in an agent-based system must deal with many conflicting goals simultaneously. 

---