Source: artificial intelligence
For many engineering organizations, the most difficult skill requirement to comprehend and to acquire may be that of the knowledge engineer. Why cannot the domain expert and knowledge engineer be the same person? This approach has been satisfactory in some situations, particularly if an appropriate software tool was available. More typically, however, different skills are required for these two positions. Also, it is difficult for experts to effectively interview themselves. As discussed in my previous posts, the rational, retrospective self-analysis of expertise is frequently incorrect.
A good knowledge engineer must have communication skills, patience, adaptability to new situations, an understanding of Al implementation techniques and an engaging personality. A good sense of humour is also helpful. It is unreasonable to expect a knowledge engineer to become a domain expert in a new field within a short time. However, the knowledge engineer must try to both value and understand the domain expert’s basic concepts, buzz words, operating environment, and what information is important. This understanding will make the sessions with the domain expert run smoothly and efficiently.
In spite of developing a limited understanding of the domain, the knowledge engineer is still not the expert. For example, the knowledge engineer should not be the one responsible for selecting a particular alternative from among several proposed by different experts. Rather, he or she should attempt to find an approach that can accommodate all points of view. The knowledge engineer forms the bridge between the domain expert and the software specialist who will design the knowledge system.
Several AI projects in a large organization benefited from the contributions of an excellent knowledge engineer. This knowledge engineer had art educational background in the humanities rather than in computer science or engineering. This lack of theoretical technical knowledge was more than compensated for by outstanding communications skills. The knowledge engineer had a knack for interpreting how various experts performed their function and what knowledge was utilized in solving their problems. Another valuable capability was art adaptability to the different ways in which the experts expressed themselves. Some experts preferred written expositions; others used pictorial sketches and blackboard diagrams; still others paced the floor and related representative dialogues. Interestingly, this knowledge engineer was able to ask the experts “dumb” questions which would have irritated the expert if they had been posed by an individual with an engineering background.
Acquiring the human relations skills described above, while being highly desirable, is not necessarily a prerequisite for a successful Al development project. A dedicated staff, thoroughly familiar with their organization’s activities and working in a supportive environment may be able to develop a useful knowledge system.
A relatively small division of a large corporation had as a primary function the high-volume production of components with demanding specifications. The engineering staff did little computer development. Emphasis was on applying computers to a complex and sophisticated manufacturing process. Competitive pressures were intensifying and the plant manager determined that productivity would be improved by more “intelligent” monitoring of a vital process. Without the support of experienced AI specialists, the plant manager and a staff member did the knowledge elicitation and representation, coded the software in a high-order language, and implemented a knowledge system on a commercial computer. The resulting small expert system was instrumental in improving the productivity of a critical manufacturing process.
This success story is not representative of the majority of Al development projects. More frequently, skilled personnel and appropriate resources are required in order to meet project goals.