Much of the work on validation and verification of knowledge based systems (KBSs) has been done in terms of implementation languages (mostly rule-based languages). Recent papers have argued that it is advantageous to do validation and verification in terms of a more abstract and formal specification of the system. However, constructing such formal specifications is a difficult task. This paper proposes the use of formal specification languages for KBS-development that are closely based on the structure of informal knowledge-models. The use of such formal languages has as advantages that (i) we can give strong support for the construction of a formal specification, namely on the basis of the informal description of the system; and (ii) we can use the structural correspondence to verify that the formal specification does indeed capture the informally stated requirements. This paper has been submitted to the Journal of Human Computer Studies (formerly the Journal of Man Machine Studies)....

The safety-critical nature of the application of knowledge-based systems to the field of medicine, demands the adoption of reliable engineering principles with a solid foundation for their construction. Logical languages with their inherent, precise notions of consistency, soundness and completeness offer such a foundation, thus promoting scrutinous engineering of medical knowledge. Moreover, logic techniques provide a powerful means for getting insight into the structure and meaning of medical knowledge used in medical problem solving. Unfortunately, logic is currently only used on a small scale for building practical medical knowledge-based systems. In this paper, the various approaches proposed in the literature are reviewed, and related to different types of knowledge and problem solving employed in the medical field. The appropriateness of logic for building medical knowledge-based expert systems is further motivated.

This paper focuses on the current functionality of OaSiS and discusses the use of safety related knowledge identified from an extensive study of oncology protocols and from discussions with clinicians, pharmacists and medical informaticians. The safety critical nature of the domain imposes requirements on software designers and implementers to ensure that the translation from paper to computerised protocol is completed thoroughly and correctly [9]. OaSiS has been implemented within RED, a project funded by the UK DTI and SERC "Safety Critical Systems" research programme. A major influence on OaSiS is the work at Stanford University on the ONCOCIN [32], EON [26] and OPAL [25] family of computer systems. In the OaSiS prototype, this is particularly evident in the user interface and its combined use of graphical, form-based 1 An earlier version of this paper was presented at the 2nd International Conference on the Practical Application of PROLOG, London, 1994. 2

A generic architecture for autonomous agents is presented. In common with other current proposals the agent is capable of reacting to and reasoning about events which occur in its environment, execute actions and plans in order to achieve goals in its environment, and communicate with other agents. The work described here proposes certain advances on other systems, notably the ability to reason about and make decisions under uncertainty, including decisions about competing beliefs and alternative actions. The framework is grounded in a non-classical decision model, the 'domino' model. This is formalised to ensure continuity with classical decision theory and avoid ad hoc features. The domino model is embodied in a well-defined knowledge representation language, R 2 L, which explicitly supports the central concepts of decisions and plans, and associated constructs of goals, arguments, commitments, obligations and constraints. The availability of such a language provides a sound basis ...

a rule-based expert system. This expert system checks results to ensure their consistency, identifies whether the results come from arterial or venous vessels, and then produces an interpretation of their meaning. This `crisp' expert system was validated, verified and commercially released, and has since been installed at twenty two hospitals all around the United Kingdom. The assessment of umbilical acid-base status is characterised by uncertainty in both the basic data and the knowledge required for its interpretation. Fuzzy logic provides a technique for representing both these forms of uncertainty in a single framework. A `preliminary' fuzzy-logic based expert system to interpret error-free results was developed, based on the knowledge embedded in the crisp expert system. Its performance was compared against clinicians in a validation test, but initially its performance was found to be poor in comparison with the clinicians and inferior to the crisp expert system. An automatic tuni

This paper presents a general discussion of the role of formal methods in Knowledge Engineering. We give an historical account of the development of the field of Knowledge Engineering towards the use of formal methods. Subsequently, we discuss the pro's and cons of formal methods. We do this by summarising the proclaimed advantages, and by arguing against some of the commonly heard objections against formal methods. We briefly summarise the current state of the art and discuss the most important directions that future research in this field should take. This paper presents a general setting for the other contributions in this issue of the Journal, which each deal with a specific issue in more detail. 1 Historical growth of Knowledge Engineering towards Formal Methods Although the history of KBS technology and Knowledge Engineering (KE) is well documented in a number of places in the literature ( e.g. [42, ch.2]), in this section we will give an account of the development of K...

An assessment of neonatal outcome may be obtained from analysis of blood in the umbilical cord of an infant immediately after delivery. This can provide information on the health of the newborn infant and guide requirements for neonatal care, but there are problems with the technique. Samples frequently contain errors in one or more of the important parameters, preventing accurate interpretation and many clinical staff lack the expert knowledge required to interpret results. The development and validation of an expert system to overcome these difficulties is described. The initial development utilised conventional `crisp' logic within the rule base and this system was evaluated to commercial release. This expert system validates the raw data, provides an interpretation of the results for clinicians and archives all the results, including the quality control and calibration data, for permanent storage. Subsequent development went on to incorporate fuzzy logic into part of the expert system knowledge base, but tests of this preliminary fuzzy system showed that it performed worse than the original crisp expert system. A tuning algorithm was then employed to modify the fuzzy model and this process resulted in improved performance to a level comparable to clinicians and superior to the crisp system. Finally, the entire knowledge base was converted to utilise fuzzy logic and this `integrated' fuzzy expert system was validated against international expert opinion.

method and a safety strategy