Declarative Representation of Knowledge (Glass Box Hypothesis)
When the format of all knowledge within an agent is determined by the
architecture, it is called declarative. Declarative representation of
knowledge provides capabilities and, according to current analysis,
precludes capabilities.
Prodigy argues that declarative knowledge allows the human
supervisors to learn what the agent knows and that this is an important
feature. This is contrasted with
Soar
in which knowledge is coagulated and compiled such that humans outside
cannot examine what the agent knows.
- Humans can determine what the agent knows.
- Knowledge manipulation tools can be shared within the agent
- Facilitates open knowledge.
Capabilities constrained or precluded by the declarative structure of the
agent
include:
- The agent cannot develop
episodic memory.
- Complex operations will be slow due to the explicit appearance of all
knowledge involved.
- Expressiveness of the agent is constrained by the representation.
Agents that employ declarative representations of knowledge:
Press
UP to go the property list.
Press
HOME to return to the Table of Contents.