Previous: Embodied representation Up: The GLAIR Architecture Next: Consciousness

Alignment

When a GLAIR-agent notices something in its environment, it registers that it has come to know of an object. Regardless of whether the agent recognizes the type of the object, we want it to explicitly represent the existence of the object in the Knowledge level while processing sensory information about the object in the Perceptuo-Motor level. Similar to sensing objects, when properties of objects or relationships among objects are sensed by the GLAIR-agent, we want it to explicitly represent these properties and relationships, even if no more is known about them than the fact that they exist. We use unnamed intensional concepts for this purpose [Shapiro \& Rapaport 1987].

Having sensed an object, an assertion is made about the object being sensed in the GLAIR Knowledge level. Once the object is no longer in the ``field of perception'', the assertion about its being sensed is removed. This is tantamount to disconnecting the relationship between the symbolic representation and the world. If at the Perceptuo-Motor level a previously sensed object is again being sensed, we reassert the fact that the object, the same one represented before in the Knowledge level, is being sensed. An example of this type of (unconscious) perception is when we look at an object, look away, and then look back at the same object. The unconscious level can provide a short term sensory memory in which memories of objects are stored, and when we see them from time to time, the conscious layer is alerted to that fact. We can think of this phenomenon as a type of continuity in perception at the unconscious level. We believe that if we assume this continuity, we should re-use previously constructed representations to represent again-sensed objects. In order for a GLAIR-agent to re-use its previously established representations about objects for again-sensed objects, we either have to assume that the agent has a continuity of perception at the unconscious layer or that a conscious matching of existing representations to sensed objects is performed.

lammens@cs.buffalo.edu