THE ROLE OF EXPERT SYSTEMS AS MODELS FOR PLANT GROWTH, ENVIRONMENTAL CONTROL AND FARM MANAGEMENT
The early part of the discussion focused on what an expert system is and how it differs from the other type of models. Although there was differing opinions as to these questions, most seemed to except the definition proposed by Scown:
"An expert system applies artificial intelligence reasoning and problem solving techniques to kowledge encoded about a specific problem domain in order to simulate the application of human expertise."
That means we are using computer technology to emulate the reasoning process of an expert as applied to a specific problem. Most could accept this as a workable definition of an expert systems.
The discussion then progressed to identifying problems that appear to be appropriate for the application of expert systems. This discussion was aided by reviewing the concepts contained in two expert systems papers that were presented at the conference. The first of these papers discussed an expert system called "CUCUMBER" which was develop by Anne Gohler. This expert system -- still in its first stage of development -- makes nutrient recommendations for cucumbers. It takes into considerations the type of soil, the stage of growth, and numerous other factors. It handles both quantitative and qualitative information in making the recommendation.
The other paper that played a role in directing his discussion was the one by Thomas Schmidt-Paulsen. This paper contrasted expert systems with the use of traditional models. Mr. Schmidt-Paulsen's paper argued that the ideal method for knowledge representation might be the integration of traditional models with expert systems, the so-called hybrid approach.
There was no strong opinions expressed regarding what type of problems that might be most appropriate for expert systems. However, there did seem to be consenus that if an algorithm was appropriate for the problem at hand, it should first be evaluated as to its effectiveness before trying to develop an expert system. Others also expressed the opinion that expert systems might be a very effective tool for interpreting of the results generated by a traditional model such as a simulation model. Many felt one of the reasons why existing simulation models are not being used to their full capacity, relates to the difficulty decision makers have in understanding the results. By careful use of an expert system to help interrupt the results of a traditional model the decision maker should be better able to utilize his/her heuristic skills to arrive at better solutions to the problem..