The notion of `knowledge level' was introduced by Allen Newell more than a decade ago . Ever since it has provided a common perspective for researchers in Artificial Intelligence (AI) and in knowledge systems in particular. Its impact has been tremendous. Newell managed to make explicit what had become common practice in AI, namely talking about intelligent systems in a language of `knowing' and `wanting'. Moreover, he gave this language a role in systems engineering by postulating the knowledge level as a computer systems level to be studied in line with other levels such as the register-transfer level or the symbol level. It is no surprise then that Newell's treatment of knowledge was particularly attractive to the system oriented mind of computer scientists who are, after all, still the majority of AI researchers. The notion of knowledge level has been used most visibly within the so called modelling approaches toward knowledge systems. In these approaches developing a knowledge based system is viewed as the construction of a series of models related to some (problem solving) behaviour. In particular the knowledge level model is a model in terms of the knowledge that rationalises that behaviour. It has become `en vogue' to assimilate the knowledge level idea in any encompassing treatment of knowledge systems. It ties together and to some extend unifies different approaches toward the theory and practice of knowledge systems No doubt taking a knowledge level perspective has greatly improved our understanding of what knowledge systems are and how we can build them. For example, it has provoked a profound shift in knowledge acquisition: rather than extracting knowledge from an expert the aim of knowledge acquisition is to build a consolidated model of an expert's problem solving behaviour in terms of knowledge. Nevertheless the knowledge level is not beyond critique and several authors have pointed out problems with it. Some of these problems required further clarification or minor repairs. Others have been claimed to be unrepairable, which would render the knowledge level useless. When one looks at knowledge level descriptions as they are presently used in knowledge systems then one finds striking differences with Newell's original notions.For example, at the knowledge level according to Newell there is no structure, whereas most models as we now see them are highly structured. How and why are these models different? The aim of this paper is to clarify this and other issues. It is not an introduction to knowledge level modeling but brings together ideas and interpretations to provide a (subjective) bird's eye view on the state of the art and the field's present research directions. This paper starts with a brief recapitulation of Newell's notion of knowledge level Then Sect. highlights the differences with the knowledge level models as they are practiced in contemporary knowledge engineering. An approach to relating the two notions, called two step rationality, is put forward. Section takes a closer look on the structures that one finds in knowledge level modelling. Then Sect. digs deeper into the nature of problem solving itself from a knowledge level perspective while Sect. discusses implications for future generation architectures. In Sect. the methodological role of knowledge level modelling is discussed. Finally, Sect. is a brief conclusion. The Knowledge Level in Practice
The original aim of the knowledge level was to clear up confusion concerning the usage of the terms 'knowledge' and 'representation'. The idea immediately resonated with ongoing research toward understanding and building knowledge systems from a knowledge content (epistemological) perspective. Clancey's model of heuristic classification illustrated the power and scope of competence models that make explicit the kinds of knowledge embodied in a system and their roles in an overall pattern of reasoning. Here are only...
Bibliography: 1. Aamodt A (1991) A knowledge-intensive approach to problem solving and sustained learning. PhD thesis, University of Trondheim, Norwegian Institue of Technology, Trondheim, Norway.
2. Breuker J. A, et al (1987) Model Driven Knowledge Acquisition: Interpretation Models. ESPRIT Project P1098 Deliverable D1 (task A1), University of Amsterdam and STL Ltd.
3. Brown D. C. and Chandrasekaran B. Design Problem Solving: Knowledge Structures and Control Strategies. Research Notes in Artificial Intelligence.
4. Clancey W. J (1985). Artificial Intelligence
5. Rademakers P. and Pfeifer R. The role of knowledge level models in situated adaptive design. Proceedings of the Tenth European Conference on Artificial Intelligence
6. Van de Velde W.( 1999) Proceedings AISB '91: Artificial Intelligence and Simulation of Behavior. London
7. William J. Clancey (1989) Machine Learning, Special Issue on Knowledge Acquisition The knowledge level reinterpreted: Modeling how systems interact.
Please join StudyMode to read the full document