Angi Voß

MODEL-K for prototyping and strategic reasoning at the knowledge level

To close the gap between knowledge level and symbol level, the MODEL-K language allows to specify KADS conceptual models and to refine them to operational systems. Since both activities may be arbitrarily interleaved, early prototyping is supported at the highest level. Systems written in MODEL-K contain their conceptual model, making them more transparent, easier to communicate to the expert, to explain to the user, and to maintain by the knowledge engineer.

OFFICE-PLAN: Tackling the Synthesis-Frontier

Analytic problems like diagnosis or classification have been intensively studied and their abstract structure is well known. For synthetic tasks like design, configuration or assembly such models are still missing. They would help us to understand the nature of the problem, the types of knowledge involved, to guide the knowledge acquisition process and to serve as a specification for the implementation. Based on concrete experiences from modelling and building an office planning system we give a first, tentative description of a conceptual model for assembly tasks.

Reflection and competent problem solving

Most of today’s knowledge based systems are far from being competent. They do not degrade gracefully, they provide no basis for adequate explanations, tutoring or really cooperative problem solving. We present an approach that aims at improving the competence of given problem solvers. For that purpose, we view competence assessment and improvement as a reflective activity which is essentially situated at the knowledge level. Our approach suggests a terminology and basic inference steps to describe competence assessment and improvement.1

Controlling Generate & Test in Any Time

Most problem solvers have a one-dimensional stop criterion: compute the correct and complete solution. Incremental algorithms can be interrupted at any time, returning a result that is more accurate the more time has been available. They allow the introduction of time as a new dimension into stop criteria. We can now define a systems utility in terms of the quality of its results and the time required to produce them. However, optimising utility introduces a new degree of complexity into our systems. To cope with it, we would like to separate the performance system to be optimised from utility management.

Solving the office allocation task in reflective ASSIGN

Abstract We show how we solved the Sisyphus task with an existing assignment problem solver, ASSIGN. The system is written in MODEL-K, a language for mechanizing KADS models of expertise. MODEL-K was extended to build reflective systems that reason about other problem solvers. Our assignment system has reflective components to simplify the problem, to solve it incrementally and to relax inconsistent constraints in the problem definition.

Ausführbare konzeptionelle Modelle

Auf dem ersten Workshop Informationssysteme und KI haben wir versucht, einen Uberblick uber die Wissensakquisition zu geben [Vos 90]. Dabei spielte die KADS-Methodologie eine nicht unwesentliche Rolle. Ganz nebenbei erwahnten wir, das wir in der GMD eine Sprache entwickelt haben, mit der konzeptionelle KADS-Modelle aufgeschrieben und so erweitert werden konnen, das sie ablauffahig werden. Diese Bemerkung erntete die groste Resonanz, und so ist uns immer wieder gegangen, egal, wo und woruber wir vorgetragen haben. Unsere Schlusfolgerung daraus ist, das es einen “Markt” fur KADS- Operationalisierungen gibt.

A Layered Algebraic Specification Technique for Expert Systems

Following the KADS’ knowledge acquisition methodology, expertise should be modelled on different layers, and recently they proposed to use algebraic specifications for each. As the base of each layer, I propose four different interface specifications describing the vocabulary, the operations and the data assumed on the layer below. The layers can then be connected by filling in the interface specifications, which can typically only be done by meta-level references to the layer below. The advantages of this technique wrt. the original conceptual models are: a richer vocabulary, a clear distinction between the layers, a smooth transition to operational representations.