|
Empirical Methods
The design of AI systems is typically justified informally. For example, one might say, "Planner X is designed to be reactive because the environment changes rapidly in unexpected ways." We believe this style of justification is too informal to support (a) demonstrations of the necessity of an agent's design, (b) evaluation of the design, (c) generalization of the design to other tasks and environments, (d) communication of the design to other researchers, and (e) comparisons ...
|
