Steels L.: "Components of Expertise", AI Magazine, 11, 2, 1990, pp. 30-49.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... language designed to support knowledge mod elling of problem solving and learning [3, 4] Noos is based on the task method decomposition principle and the analysis of knowledge requirements for methods and it is related to knowledge modelling frameworks like KADS [12] or components of expertise [11] . A method models a way to solve a task. A method can be elementary or can be decomposed in subtasks. These new (sub)tasks can be achieved by corresponding methods in the same way. For a given task there may be multiple alternative methods (alternative ways to solve the task) For instance, a BI ....

Luc Steels. Components of expertise. AI Magazine, 11(2):28 49, 1990.

....hypothesis suggests [Chandrasekaran, 1987] The interaction hypothesis states that the structure of knowledge is geared to, or allows specific problem solving methods. A problem solving method is either a decomposition of a problem solving task, or a procedure (algorithm) to solve a problem [Steels, 1990]. A problem solving method therefore controls the use of inferences by stating (alternative) sub goals. These decompositions are not arbitrary. In a good decompostion, the process of inference making becomes tractable and does not explode, because only a subset of all potential inferences is to be ....

....deriving the fact that next in a pair of bicycles next to one another means that one is less to the right side of the road than the other one, is performed here. The WORLD MODEL is used to construct a SITUATION DESCRIPTION, which is a situation specific model of the case or request at hand. [Steels, 1990] It consists of a list of states and events in terms that are known to the legal reasoning component. This is the second component, and its role is to derive the legal consequences of the SITUATION DESCRIPTION. These legal consequences can be in terms of violations of regulations, or ....

Steels, L., Components of Expertise, in: AI Magazine, Summer 1990, pp. 29-49. 101 Breuker, J.A.P.J.

....processing. The selection of a method among others is dependent upon multiple factors such as the availability of the knowledge (e.g. domain theories or models) required by a particular method and the pragmatic constraints (e.g. explosiveness of the search space) imposed by the task environment [13]. Chandrasekaran [14] in the discussion of the PCM1 method for the generic design task, suggests a scheme of criteria to look at when determining which methods to choose for the various tasks. For example, by examining the availability of the knowledge required by a 1 PCM stands for ....

....that provides the associations between the various qualitative factors and their potential influence on the inherent risk level of the various accounts. The generate plausible hypotheses generic task in the past has been solved by one of the several forms of the Classification method (e.g. [13], 20] The selection of an approach to classification depends on the form of the knowledge available in the domain. The organization of Figure 3 Task decomposition for the Quantitative Assessment subtask. Assessment Search for Potential Explanation Identify Departed Identify Strong ....

Steels, L., Components of Expertise. AI Magazine, 1990. 11(2): p. 28-49.

.... and (3) comparing di#erent role limiting methods, such as propose and revise and cover and differentiate, in the goal of characterizing a taxonomy of methods to guide system modeling [20] PSMs were proposed as standard reasoning procedures for addressing generic tasks in a domain independent way [8, 31, 28, 3, 11]. Over the years, the knowledge engineering community identified and developed PSMs of general usefulness [6, 3] or for specific high level tasks such as classification [8, 23] diagnosis [2] design [29, 22] and planning [33, 32] A PSM that the knowledge aquisition community studied at length ....

....PSM descriptions, while including formal specifications that allow both PSM providers to organize their components into libraries and systems to index and retrieve PSMs automatically. UPML takes its roots in knowledge engineering approaches such as generic tasks [5] components of expertise [31] and CommonKADS [30] principles of software engineering and reuse, as well as recent Semantic Web issues. UPML covers all aspects of modeling PSMs and the way in which they are configured into running applications. Although UPML focuses on PSMs because they are the actual performance components ....

L. Steels. Components of Expertise. AI Magazine, 11:30--49, 1990.

....the interaction of the user with the EDSS. The user may ask it for justifications and explanations of suggested decisions and possibly validation of plausible alternatives to make a better decision. The development of an EDSS as a complex integrated KBS relies on the idea of model refinement [43]. Every stage in the development process involves a relatively straightforward step of transformation from one model to the next. That is, from requirements to conceptual model, from conceptual model to design model, and from design model to code. In Fig. 2, a scheme of the ideal cycle of ....

....cognitive tasks and techniques such as learning, reasoning, knowledge acquisition [97] and distributed problem solving. Also, different AI techniques are combined such as rule based reasoning, case based reasoning and modelbased reasoning. Four levels are distinguished from the domain models [43] point of view: data, knowledge, situations and plans. On the other hand, taking into account the supervision tasks, seven levels are considered: evaluation, diagnosis, supervision, prediction, validation, actuation and learning. This system was developed for the WWTP domain, but it is a general ....

L. Steels, "Components of expertise," AI Magazine, vol. 11, no. 2, pp. 28--49, 1990.

....the knowledge represented in a knowledge base. Ontologies are effective ways to unify terminology in a given domain. That is, they provide a means for common domain modeling. All this integration will lead the EDSS towards the extensive use of domain models in the sense of components of expertise [35]. These models should reflect both the operational and strategic knowledge embedded in the system and used by the people around it. This integration of knowledge struc Fig. 4. Integrating technologies: How tures and methods describes the domain s deep knowledge level. During the integration ....

....processes on local constraints, such as weather conditions, climatic aspects, geographical positions, environmental and or health law regulations, etc. Methods, Models and Tasks are means to ensure desirable Knowledge Based behavior and are crucial elements of the Knowledge Based Systems design [35]. These represent the components of expertise needed to handle the problem. In Fig. 7, we advance a proposal to combine those components of expertise to build EDSS. Advances in the area of Model Based Reasoning are very promising [24] and specific models are to be developed for this field. ....

L. Steels, Components of expertise, AI Magazine 11(2) (1990), 28--49.

....feedforward control, optimal control, adaptive control, etc. One of the difficulties inherent to the development of a knowledge based system is the build up of the knowledge base (i.e. knowledge acquisition) specially when dealing with a wide and complicated field (i.e. ill structured domain) [22, 23]. Acquiring relevant knowledge is a difficult task for a number of reasons. Experts do not necessarily find it easy to formulate how to solve the day to day problems arising in the running of the process. Moreover, they may omit part of the information in their explanation, introduce ....

....information. Useful in ill structured domains. Could be extended in some ways. For example, the treatment of uncertain or approximate reasoning. Nevertheless, KBS do not incorporate some desired features from human intelligence and have some technical difficulties in their development [23, 43]. Most KBS do not learn from experience. The use of experience is a valuable feature to be contemplated in KBS [44] The knowledge acquisition problem. There are some difficulties in the process of extracting the knowledge and experience from knowledge sources [22] Brittleness. Their ....

[Article contains additional citation context not shown here]

L. Steels. "Components of expertise". AI Magazine 11 (2), pp. 28-49, 1990.

....of the hill climbing heuristic sometimes used for searching purposes. There are several knowledge modelling techniques in the literature such as the generic task, proposed by Chandrasekaran [8]also present with some variants in the KADS model [14] and in the model of components of expertise [13]and the role limiting method of McDermott [11] According to these techniques, there are two organizational principles: the task oriented principle and the domainoriented principle. Our approach belongs mostly to the latter. Like AQUINAS [4] DESIRE [6] and Common Kads [10] we handle ....

Steels, L., "Components of expertise", AI Magazine, 11(2): 29--49, 1990.

....[19] In KSM, the developer can create generic abstract knowledge structures that can be applied to different specific domains by duplicating and configuring their components. The Knowledge Reuse Tool (KREST) is another software reuse environment, based on the components of expertise approach [20]. KREST presents a knowledge level description of an application and assists non programmer users in reusing parts to develop applications. Other approaches to the design and configuration of knowledge based systems exist, such as the KADS project [21] 5.1. The PROTG II Approach PROTG II is a ....

Steels L. Components of Expertise. AI Magazine 1990; 11 (2):29--49.

....knowledge are likely to change over time. Thus, inflexible tools will make maintenance a nightmare. To break this inflexibility, several research groups have recently developed architectures in which problem solving methods are composed from small grained reusable components [5] 8] 18] 32] [36] [42] While these groups share many similarities in their approach to composing problemsolving methods, they differ markedly on the degree of support they provide to the task of acquiring Corresponding author. email: tu camis.stanford.edu; fax: 415 725 7944 domain knowledge from the application ....

....frameworks for modeling and implementing flexible knowledge based systems. In the following, we shall review the work of four groups who have research goals similar to those of PROTG II. Steels and his colleagues at Free University of Brussels have developed a components of expertise framework [36] and have implemented a KREST workbench that supports the construction of knowledge based systems in that framework [37] The componential framework has three perspectives on an activity to be automated: task, information sources, and methods . Information sources encompass data sources coming ....

L. Steels, Components of expertise, AI Magazine. 11 (1990) 30--49.

....adaptable. Using the facilities of Mecano, knowledge engineers can adapt the knowledge editors produced by Mecano to the characteristics of specific tasks, domains, and users. Although other researchers of knowledge acquisition tools are examining means to develop multiple method architectures [6,12], they have generally disregarded the problem of developing knowledge editors that are suited for individual tasks and domains. As a consequence, even when a knowledge editor can be built for a given task, its usefulness may be highly constrained because the interface of the knowledge editor does ....

Steels, L. 1990. Components of expertise. AI Magazine 11(2):30--49.

.... a knowledge level description to be constituted of different types of knowledge, and that these forms of knowledge play different roles during the problem solving process, and require different structuring principles [Wielinga and Schreiber, 1993] For example, Steels in Components of Expertise [Steels, 1990] partitions knowledge into domait models, tasks, and problem solvitg methods. Wielinga et al. divide CommonKADS knowledge [Wielinga 1992] into domai kowledge, task kowledge, iferece kowledge, and strategic kowledge. In order to describe an application at the knowledge level it is necessary to ....

.... Task concept into Task Structures [Chandrasekaran and Johnson, 1993] A comprehensive extension of these classifications is proposed by Puppe in [Puppe, 1993] The separa tion between problem solving and domain has been further elaborated and refined by Steels concept of Components of Expertise [Steels, 1990] and CommonKADS expertise models already described [Wielinga et el. 1992, Wielinga et el. 1994] To support the implementation of CommonKADS models of expertise, a suite of problem solving types is proposed by Breuker [Breuker, 1994b, Breuker, 1994a] and ontologies for describing tasks and ....

L. Steels. Components of expertise. AIMagazine, 11(2):29 49, 1990.

....distinction between theory and practice is commonplace in many domains close to cognitive science. However, we will show that practice is more context dependent than theory. In the AI field, the previous distinction is intertwined with the discussion about deep and surface knowledge (see e.g. [34]) It is commonly admitted in AI that deep knowledge refers to models and causal explanations that goes back to nature laws, whereas the surface knowledge is represented by practical rules that can be acquired from people performing efficiently a given task (human experts) Thus, we see that, on ....

Steels L. (1990) Components of Expertise, AI Magazine, Summer, 28-49.

....between theory and practice is commonplace in many domains close to cognitive science. However, we will see that practice is more dependent of context than theory. In the AI field, the previous distinction is intertwined with the discussion about deep and surface knowledge (see e.g. [14]) It is commonly admitted in AI that deep knowledge refers to models and causal explanations that goes back to nature laws, whereas the surface knowledge is represented by practical rules that can be acquired from people performing efficiently a given task (human experts) Thus, we see that, on ....

Steels L. (1990) Components of Expertise, AI Magazine, Summer, 28-49.

....various knowledge systems performing the task. Gruber and Cohen [33] identified uncertain reasoning as a generic activity and proposed generic methods and representations to handle this problem. Musen s work on Protege [54] was quite close in spirit to DSPL. Steels work on componential frameworks [66] was to come later, but it also shared the spirit of the GT work (we discuss this approach later in the paper as well) Thus a community was emerging with similar goals, with approaches that shared some ideas and differed on others. Specifically, they shared two features. First, they identified ....

....made specific commitments for a vocabulary is, as we discussed above, in their proposal for a typology of primitive goal types. As we said, our preference is for these terms to emerge empirically from studies of classes of tasks. 4 . 3 The Componential Framework In Steels componential framework [66] systems are described from three perspectives: 1) the model perspective; 2) the task perspective; and 3) the method perspective. This framework is in many ways similar to the task structure framework that we have described. The model perspective corresponds to the knowledge states of problem ....

Steels, L.: Components of expertise. AI Magazine, 11(2):28-49, 1990

....the tasks that need to be solved, the methods that will accomplish those tasks, and the knowledge of the application domain that those methods need. A description of a system along these lines is often referred to as a knowledge level description, and more recent research in knowledge acquisition #Steels, 1990; Wielinga et al. 1993# has clearly demonstrated the usability of this approach. The original knowledge level idea has undergone some modi#cations over the years, from Newell s highly intentional, purpose oriented way of describing a system #Newell, 1982#, to a more structured and usable type of ....

Steels, L. #1990#. Components of expertise. AI Magazine, 11#2#:29#49.

....of) problems or tasks rather than as a model of general intelligence; ii) the introduction of structuring elements at this level; and (iii) a top down analytic approach to problem system decomposition. The main modelling methodologies, such as Chandrasekaran and Johnson (1992) Musen (1992) Steels (1990), Wielinga et al. (1992) can all be seen as organizing models around three main types of components: knowledge models containing domain knowledge necessary to solve a problem, which are structured according to one or several ontologies; tasks or (types of) problems to be solved; and Draft. To ....

Steels, L. (1990). Components of expertise, AI Magazine 11(2): 28--49.

....from the implementation details. We illustrate our approach using the problem solving method Propose Revise. 1 INTRODUCTION Many approaches in knowledge engineering cope with the problem of constructing a library of reusable problem solving methods (PSMs) Chandrasekaran, 1983; Clancey, 1985; Steels, 1990; Musen et al. 1994; Breuker and Van de Velde, 1994) Here, we consider problemsolving methods in the sense defined by McDermott (McDermott, 1988) in which a PSM is a strong method with a constrained control structure. These methods are called role limiting methods because their force is in ....

Steels, L. (1990). Components of expertise. AI Magazine, 11(2):28--49.

....Examples of such control knowledge are the order in which observations must be obtained during diagnostic reasoning, or the order in which components must be configured during design reasoning. Many Knowledge Engineering methodologies provide some form of expressing the control knowledge in a KBS [28, 20, 3, 27]. A more recent, and less explored approach to dealing with the intractability of KBSs is the development of anytime algorithms [19] An anytime algorithm gradually approaches the perfect solution. As runtime increases, the quality of the solution increases. The algorithm can be interrupted at ....

L. Steels. Components of expertise. AI Magazine, Summer 1990.

No context found.

Steels L.: "Components of Expertise", AI Magazine, 11, 2, 1990, pp. 30-49.

No context found.

Steels, L., Components of expertise. In: AI Magazine, summer 1990.

No context found.

STEELS, L. (1990). Components of expertise. AI Magazine, 11(2), pp. 30-49.

No context found.

L. Steels, Components of expertise, AI Magazine 11(2) (1990), 28--49.

No context found.

Steels, L. "Components of expertise". AI Magazine 11 (2), pp. 28-49, 1990.

No context found.

#77# L. Steels. Components of Expertise. AI Magazine,

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC