Editorial: the Sisyphus-VT initiative

This special issue contains results of the second ‘‘Sisyphus’’ study carried out in the knowledge acquisition community in the period 1992 – 1995 . The Sisyphus initiative arose at the 1990 European Knowledge Acquisition Workshop in Driebergen (The Netherlands) . To explain the rationale behind this initiative it is necessary to give a brief synopsis of the recent history of the knowledge-engineering field . The traditional approach towards knowledge engineering (KE) was that of a ‘‘mining’’ activity : knowledge was ‘‘extracted’’ from available sources such as text books and experts , and subsequently ‘‘transferred’’ to the target system . This view proved to have serious limitations . Firstly , it turned out that it was dif ficult to maintain the resulting knowledge base . It was not clear when the knowledge base was complete , and introduction of new knowledge pieces (typically production rules) could have unpredictable ef fects on the problem-solving process . Secondly , the resulting systems were unable to explain their reasoning behaviour in a way that was understandable for humans . Thirdly , it was unclear how one could reuse (parts of) a system for a new application . With the growing complexity of systems , these limitations became more important . Clancey was one of the first researchers to analyse these shortcomings based on his experiences with the NEOMYCIN system (Clancey , 1983) . The limitations of so-called ‘‘first-generation’’ knowledge engineering led during the second half of the 1980s to a shift from the ‘‘mining’’ paradigm to what has become known as the ‘‘model-based’’ knowledge-engineering paradigm . Two features lie at the heart of model-based knowledge engineering :