Computational reflection is the activity performed by a computational System when reasoning about (and by that possibly affecting) itself. This paper presents an introduction to computational reflection (thereafter called reflection). A definition of reflection is presented, its utility for knowledge engineering is discussed and architectures of languages that support it are studied. Examples of such procedural, logic-based, rule-based and object-oriented languages are presented. The paper elaborates on the design of these languages and the reflective functionality that results, elucidating concepts such as procedural reflection, declarative reflection, theory relativity of reflection, etc. The paper concludes with an assessment of outstanding problems and future developments in the area.

The expert systems business has been through a terrible period in 1987–88. In the USA, the “Gang of Four”—the new venture software companies which once led the industry—have all run into losses and cut back staff more or less severely. Symbolics, whose high priced Lisp workstations have made it the biggest single company in artificial intelligence (AI), has hit an equally dramatic reversal of fortunes.

This paper reviews a wide range of knowledge acquisition techniques in the context of attempts to achieve a systematic methodology. These have been poorly documented by expert system builders, who are often inclined to overvalue textbooks and the ways experts themselves claim they solve problems. No one method has a universal advantage; each has some value. Techniques should be selected to suit the domain, the task, the expert and the knowledge engineer. Knowledge acquisition involves creating a conceptual model of expert knowledge and reasoning, from analysis of data elicited by these techniques. A survey of the literature indicates increasing emphasis on tools for knowledge acquisition, used directly by experts. Several projects currently directed towards providing a proper epistemological foundation for knowledge acquisition are discussed and compared. None has yet produced a complete epistemologically sound methodology; however, recognition of the need to create a conceptual model at the knowledge level (rather than the symbol level) is an important advance. An extensive bibliography is appended.

This paper is a review of certain non-monotonic logics, which I call default non-monotonic logics. These are logics which exploit failure to prove. How each logic uses this basic idea is explained, and examples given. The emphasis is on leading ideas explained through examples: technical detail is avoided. Four non-monotonic logics are discussed: Reiter's default logic, McCarthy's circumscription, McDermott's modal non-monotonic logic, and Clarks's completed database. The first two are treated in some detail. The recent Hanks-McDermott criticism of non-monotonic logic is discussed, and some conclusions drawn about the prospects for non-monotonic logic. Recommendations for further reading are given.

The level of abstraction of much of the work in knowledge-based systems (the rule, frame, logic level) is too low to provide a rich enough vocabulary for knowledge and control. I provide an overview of a framework called the Generic Task approach that proposes that knowledge systems should be built out of building blocks, each of which is appropriate for a basic type of problem solving. Each generic task uses forms of knowledge and control strategies that are characteristic to it, and are in general conceptually closer to domain knowledge. This facilitates knowledge acquisition and can produce a more perspicuous explanation of problem solving. The relationship of the constructs at the generic task level to the rule-frame level is analogous to that between high-level programming languages and assembly languages in computer science. I describe a set of generic tasks that have been found particularly useful in constructing diagnostic, design and planning systems. In particular, I describe two tools, CSRL and DSPL, that are useful for building classification-based diagnostic systems and skeletal planning systems respectively, and a high level toolbox that is under construction called the Generic Task toolbox.

Distributed Artificial Intelligence has been loosely defined in terms of computation by distributed, intelligent agents. Although a variety of projects employing widely ranging methodologies have been reported, work in the field has matured enough to reveal some consensus about its main characteristics and principles. A number of prominent projects are described in detail, and two general frameworks, the System conceptual model and the agent conceptual model, are used to compare the different approaches. The paper concludes by reviewing approaches to formalizing some of the more critical capabilities required by multi-agent interaction.

The knowledge engineering practices of one experienced company are outlined, commenting on differences in emphasis in some of the knowledge elicitation and AI research literatures.

Expert systems and hypertext are technologies which are coming together. This paper reviews the strengths and weaknesses of expert systems and hypertext. Then the complementarity of the two is both argued and illustrated. Translating expertise from documents into expert systems is difficult, but hypertext systems exploit the information in documents. Expert systems don't explain their decisions as well as some people would like, but hypertext is basically designed to provide explanations. On the other hand, people have difficulty deciding what links to follow in hypertext, but an expert system is designed to help people make decisions. When a hypertext reader is confused as to what step to follow next, an expert system might review the path taken thus far and suggest the appropriate next step. An annotated bibliography of the principal literature is provided.

The paper I wrote for the Knowledge Engineering Review was intended primarily to give a balanced review of different techniques developed in (or imported into) Artificial Intelligence to deal with uncertain knowledge.

The oft quoted inadequacy of von Neumann architectures for AI applications has regularly been used to justify the design of special purpose parallel machines. In particular, the von Neumann computational model has been criticized as being unsuitable for parallelism because of the memory access bottleneck. For the design of a new machine both top-down and bottom-up methodologies have drawbacks. The middle-out strategy, working both up and down from an intrinsically concurrent high-level programming language as a means of both representing and processing knowledge provides an attractive way of providing a symbiosis between software and hardware. The longest established and most well-founded symbolic method for the representation and manipulation of knowledge is logic. A notable result of the last decade, work on mechanical theorem proving was that a subset of predicate logic, Horn Clauses, can form the foundation of a computational model. The execution model of Prolog, the first popular language based on Horn Clauses, was designed for efficient evaluation on von Neumann architectures. An alternative computational model, more suitable for expressing reactive systems but retaining Prolog's affinity for metaprogramming, has given rise to a new class of languages, concurrent logic languages. One among many of these languages, FGHC (flat Guarded Horn Clauses), was developed by the Japanese as the kernel of their Fifth Generation initiative. A background familiarity with Prolog would be helpful in understanding this article.

To summarize, Software engineering Environments provides a useful overview of recent progress in software engineering tools and methods; particularly from a UK perspective. The majority of the contributions are well-written but the general paucity of introductory material and the level of technical detail in some chapters might prove daunting to readers unfamiliar with software engineering. It would have been helpful if the editor had made more effort to draw out the common themes and important issues; either by linking sections of the book with short discussion (as Hawley did for Artificial Intelligence Programming Environments) which helps to smooth over the changes in prose style or by cross-referencing points made by different authors. As it stands the result is rather a peculiar hybrid between a conference proceedings and a multi-author text book.

I thank the commentators for their time, and generally positive remarks on the promise of the task-specific approach, in particular the generic task (GT) proposal.

Johnson and Zualkernan would like a methodology for mapping domain knowledge onto one or more generic tasks so as to solve problems efficiently. This stage of problem and domain analysis in which the kind of reasoning that goes on needs to be analysed in a vocabulary of generic tasks is very important, and in our laboratory we identify this as the epistemic analysis stage. For specific classes of problems we have developed guidelines on how to perform this mapping. For example, Bylander and Smith (Bylander and Smith, 1985) describe a set of criteria and guidelines for mapping medical knowledge into CSRL-like structures for diagnostic reasoning. Similarly Brown (1984) describes criteria for mapping design knowledge into DSPL-like structures.

Literature relevant to the design and development of graphical interfaces for knowledge-based systems is briefly reviewed and discussed. The efficiency of human-computer interaction depends to a large extent on the degree to which the human-machine interface can answer the user's cognitive needs and accurately support his or her natural cognitive processes and structures. Graphical interfaces can often be particularly suitable in this respect, especially in cases where the user's “natural idiom” is graphical. Illustrated examples are given of the way in which graphical interfaces have successfully been used in various fields with particular emphasis on their use in the field of knowledge-based systems. The paper ends with a brief discussion of possible future developments in the field of knowledge-based system interfaces and of the role that graphics might play in such developments.