Declarative vs. Procedural Knowledge Representation:

Architectures with declarative representations have knowledge in a format that may be manipulated decomposed and analyzed by its reasoners. A classic example of a declarative representation is logic. Advantages of declarative knowledge are:

  • The ability to use knowledge in ways that the system designer did not forsee
  • Architectures with procedural representations encode how to achieve a particular result. Advantages of procedural knowledge are:

  • Possibly faster usage
  • Often times, whether knowledge is viewed as declarative or procedural is not an intrinsc property of the knowledge base, but is a function of what is allowed to read from it. Production systems, for example, are declarative if productions may view themselves, and are procedural it they can. (cf. glass-box/black-box control knowledge)

    A particular architecture may use both declarative and procedural knowledge at different times, taking advantage of their different advantages. An example of this is Atlantis, which has a low-level. The distinction between declarative and procedural representations is somewhat artifical in that they may easily be interconverted, depending on the try of processing that is done on them.

    Examples of declarative architectures are:

  • Dynamic Control Architecture by B. Hayes-Roth. All
  • ERE by Drummond et al. knowledge is in uniform conceptual graphs.
  • Homer by Vere & Bickmore. Knowledge is explicit in frames.
  • Icarus by Langley. Knowledge is explicit in Labyrinth.
  • MAX by Kuokka. Knowledge is stored lframes that are readable by all.
  • Prodigy by Carbonell et al. Knowledge is store in FOPC-like Prodigy Description Language (PDL).
  • Theo by T. Mitchell et al. Knowledge is explicit in frames.

  • Examples of procedural architectures are:

  • SOAR by A. Newell et al.. Productions may not examine each other.
  • Subsumption Architecture by R. Brooks

  • Examples of declarative & procedural architectures are:

  • Atlantis by E. Gat. The control layer uses procedural knowledge while the deliberative layer uses declarative knowledge.
  • Behavior-Based Architecture by R. Brooks. This is Subsumption Architecture augmented with global state in the form of Hormones.
  • RALPH by Ogasawara and Russell. Different Execution Architectures (EAs) use different types of knowledge.
  • Teton by VanLehn & Ball. Knowledge is in productions and on a scratch-pad.

  • Other Properties. Back to the Title Page.