Mental saccades in control of cognitive process

This paper proposes a cognitive architecture that uses mental saccades to perform cognitive search in support of motivated behavior and learning. It is intended to control the behavior of robots in real environments and avatars that learn how to operate in virtual worlds. A mental saccade is a parallel concept to the visual saccade and yields a sequential cognitive search for the most likely solution to a problem. This model uses an attention switching mechanism that combines the effect of observations, internal motivations and abstract cognitive planning. Thus, a system that uses this model, will not only follow its internal motivations but will also take advantage of opportunities that present themselves in the environment. This model is intended for development of computational cognition, learning and intelligence in a machine.