The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

PhD Thesis #9305

Title:A Method for Understanding Experimental Computer Programs in Artificial Intelligence Research
Authors:Bernaras Iturrioz,A
Date: 1993
Abstract:Tradition has it that work in Software Engineering has little or nothing to offer Artificial Intelligence (AI) research. The argument is that building computer programs in AI is very different from building computer programs for application. Although I accept this, I believe that some Software Engineering concepts and constructs can be used to improve the research done in AI, in particular in the Symbolic Paradigm. This paradigm is based upon Newell and Simon's Physical Symbol System Hypothesis, which states that a physical symbol system has the necessary and sufficient means for general intelligent action. Eperiments in this paradigm involve building experimental computer programs to investigate whether a physical symbol system can be further organised to exhibit particular kinds of intelligent action. As part of the experimental procedure, it is necessary to identify and to understand the components and structure of a program that cause its observed behaviour. This involves a static analysis of the program (in contrast to a dynamic analysis which is concerned with testing the scope of its behaviour). To understand a program's structure directly from an inspection of the actual code can be very difficult. It is likely to be easier to understand a form of a program in which the significant features of the program are described in a way abstracted from implementational details.This thesis is concerned with the use of Software Engineering abstraction constructs to help in the process of understanding computer programs that are built as part of experiments in the Symbolic Paradigm. It is also concerned with developing and testing a method to analyse these programs in an organised and structured way.

[Search These Pages] [DAI Home Page] [Comment]