Issues

An architecture for intelligence may be classified accoring to the physical structure and mechanisms incorporated into its architecture", what capabilities it claims to support, what properties it exhibits in relation to the implementation of the capabilities and architecture, and even what environments it is meant to operate in. However, this analysis ignors issues such as how efficient it supports its capabilities and how general, scalable, or rational the resultant system actually is. There is no clear set of benchmarks available to analize systems with respect to each of the following critera, therefore what is included with each archtecture is an analysis along the dimension which seem to pertain most to that particular architecture. Note that the following dimensions were modified from J. Laird's Prefacefor the Special Section on Integrated Cognitive Architectures,SIGART Bulletin 2, 1991.

Rationality

Are the agents actions always consistent with its knowledge and goals?

Taskability

Can the system respond to external commands? How?

Scalability

Is the system limited to a small range of problems, or does it sacle up to larger problems with more knowledge?

Efficiency

What are potential bottlenecks in system. Will it preform in real-time?

Psychological Validity

Does the architecture make any attempt to model aspcets of human behavior?