Issues addressed

Goal-directed, intentional learning

Agents confronted with multiple conflicting or exclusive goals must choose between the goals. Once commited to a goal, an agent must decide when to learn, when to react and what to learn.

Integration of reasoning, execution and learning

Deliberations can often be a more time-consuming process than actions and the time-scale of deliberations can often exceed those of important characteristics of the environment. This introduces the architecural property of graceful interruption and resumption of tasks, distributed attention and the determination of when to quit. There is little discussion of these properties except that, in some case, selective attention has been achieved by MAX.

Learning, a time and resource intensive process, must be applied judiciously before commitments are made whenever possible.

Return to the Top of this architecture.