A critiquing expert system to assist in operations of the San Francisco water supply system.

摘要

A critiquing expert system was developed to assist in operating San Francisco's water supply network. The critiquing system goes beyond a traditional expert system by refining (rather than supplying) the user's proposed operating plan through a critique. A traditional expert system requests problem-specific data, then provides the operator with a plan. In the critiquing approach, the operator submits not only relevant information to the system, but also a proposed plan. The system evaluates the plan and provides feedback, which includes suggestions for improvement, warnings, and alternatives.; The SFWD was motivated to have the critiquing system developed because of the perceived benefits in formalizing operating expertise in a critiquing system. Operating decisions are based on heuristic knowledge, not mathematical models. When personnel leave, the SFWD loses key information about how to operate the water supply network. A critiquing system can improve operators' decisions by providing expert feedback on their proposed plans, and can aid in training novice operators.; Building the critiquing expert system provided several technical challenges. A new paradigm was designed to implement critiquing in an expert system development tool. Also, developing a critiquing system is more complex than developing a traditional expert system. The critiquing paradigm and system development techniques designed for this research can be used to build critiquing systems in a variety of domains.; The research included experiments to test the postulated advantages of the critiquing approach over the traditional approach to expert systems. The results were unambiguous; the critiquing system was preferred to the traditional expert system for each of several measures of system performance and acceptability.; The research makes three main contributions. First, the research establishes the feasibility of implementing a critiquing system for decision support in a civil engineering problem domain. Second, the research demonstrates, both theoretically and empirically, the substantial benefits of the critiquing approach to expert systems. Third, the research reveals ways the organization influences the system's development, and how system development profoundly influences the organization. Not only the system itself, but also the development process that creates the system, fosters organizational acceptance and use of the system.