9-FEB-75 20:02:22,3149
Date:  9 FEB 1975 2002-PST
From: SRIDHARAN
Subject: Methodological issues by which the various programs can be vivisected.
- - - -
KNOWLEDGE BASED SYSTEMS

1.  Representation of knowledge:  
     Procedures
     Semantic nets
     Specific data structures such as weighted transition graphs etc.
     Models of various sorts
     Causal rules in the SITUATION-ACTION form

2.  Knowledge Acquisition:
     Acquiring knowledge from an expert
       Provide convenient form for input by human expert
       Provide internal consistency check of what is being fed in
     Acquiring new knowledge from program's own experience
     Adjusting the data base of knowledge in the light of experience

3.  Inference mechanisms:
     Problem solving strategy
       Exhaustive search followed by evaluation
       Plan formation followed by constrained search and evaluation
       Heuristic guided search
       Rule based reasoning
     Deductive mechanism
       Natural deduction guided by rules
       Special purpose Theorem Proving
       General Theorem proving capability
       Elements of fuzzy or uncertain reasoning
     Inductive or Generalizing capability
       Induction viewed as Heuristic Search
       Grammatical Inference as a paradigm for induction
       Pattern Recognition approaches

4.  Where does the emphasis lay in actual system development?
     On the rules guiding the process vs. On the Processes that use the
       rules effectively.
     On the representation of information to the system vs. On the
       control structure that utilizes relevant information at the right
       time.
     Settle on a reasonable knowledge structure and tune up the inference
       mechanisms vs. Settle on a simple inference process and constantly
       alter and tune up the actual information that goes into the program.
     Does the program have any MEMORY of past inputs and events, 
       whether storing everything or having some kind of summarizing
       ability?  Does the program behaviour change over time?

5.  Miscellaneous issues on the interface to the user:
      Is the program Error-checking? Error-tolerant? Error-correcting?
      Use graphics? Interactive?
      Allows Natural Language (sic, English) to input from user;
        Prints in english; Can knowledge base be queried in English?

      Is there any extent of Mixed initiative (i.e. does the user 
        have the option of not answering a question being  asked of
        him and respond with a question wishing to get more information
        from the program)?
      Does the program explain its own reasoning steps and goals?
       Can it answer the questions like, HOW will you use this
       information?  WHY do you ask to know?

6.  Details on Systems features necessary for the implementation;
     The technology that will enhance the system.

7.  Some information on "modes of collaboration" between the experts
    in the application disciplines and computer science (A.I.?)
    experts.!
     Special problems that might be of interest to others in AIM.

-------