Architectures with Uniform Representation of Knowledge

In order to operate effectively in any world, an agent must be furnished with some knowledge about that world. In some instances, this knowledge is encoded on a very low level, describing little more than a direct relationship between sensor inputs and effector actions. Most intelligent agents, however, operate with a higher level understanding of the world, wherein knowledge is encoded symbolically.

At this level, world knowledge typically addresses a number of characteristics of the world: objects and their attributes; actions, their preconditions and results; episodic knowledge about past events. Some agents include additional meta-level knowledge about the agent itself, such as its plans, goals, and processes.

By employing a uniform representation of knowledge, agents have the potential to reason about ALL aspects of its world. Not only can it reason about objects and actions; it can also reason about its own actions, enhancing and improving them, increasing its efficiency, heightening its understanding of the world, and facilitating effective learning.

The most common methods for achieving a uniform representation involve variations on highly general data structures such as frames or productions. The following is a list of architectures that do use a uniform representation

Back to List of Other Properties