In addition to advancing the state of research in qualitative reasoning, the WGQR working group has been intended to increase the awareness of the subject in Europe, especially within ESPRIT. Its method of doing this is by testing the strengths and limitations of current standard representations for qualitative reasoning about processes, and by identifying new paradigms that will encourage new types of applications. It has been set up to promote qualitative reasoning as a technique in Europe by organisingopen workshops and at least one technology transfer meeting to which representatives of European projects that could benefit from the use of qualitative reasoning methods can be invited. .
The project brought together the most active centres of research in qualitative reasoning in Europe in order to define the subject and to extend the range of problems to which qualitative reasoning can be applied. The collaboration has concentrated on six issues: systems with distributed parameters, reasoning about time, spatial reasoning, the drawing of qualitative conclusions from mathematical information, the use of multiple models or views of phenomenon, and intelligent automation. Developments in this work have led to a further concentration on the use of qualitative reasoning in design, particularly where problems are not specified primarily by equations.
APPROACH AND METHODS
Each topic covered requires a specific approach:
-Distributed parameters: modelling a distributed system by a collection of elements behaving as molecule-like units. Division of multidimensional spaces into qualitatively equivalent regions after identifying the conditions expressing qualitative equival ence.
-Temporal reasoning: comparative assessment of different interval-based logic systems and qualitative calculi, in order to select the best ones for capturing common-sense views of durations, extended events and processes.
-Spatial reasoning: development of the technical concept of "place vocabulary" to handle examples of 2-dimensional physical objects with arbitrary 2-dimensional motions.
-Mathematical information: construction of a taxonomy of equation sets and types, and of representations suitable for holding expert knowledge about the interpretation of equations.
-Multiple models: development of description mechanisms that can be used to indicate when it is appropriate to use each type of model.
-Intelligent automation: defining benchmark problems that are realistic in themselves and that also provide reasonable tests of each of the schemes mentioned above.
PROGRESS AND RESULTS
While the Working Group's members have made technical progress on each of the six topics that it originally undertook to study, the main results of its efforts have been:
-The establishment of active links with the qualitative reasoning community and research programmes in North America, to the extent that its continuing contribution to the evolution of qualitative reasoning is recognised as being of world standard.
-A change of focus from the symbolic (algebraic) mathematical style of expressing and solving problems to an approach that takes into account more general frameworks for human qualitative judgements.
A consequence of the latter is that qualitative reasoning is now tackling new types of problems, particularly in design (architectural, electronic, etc), which are not specified primarily by equations. Because of this, the subject is returning to the emphasis that it had when it began in the early 1970s: 'naive physics', or reasoning about behaviour of systems and structures by using informal or everyday terms and axioms to describe them. The Working Group's existence has ensured that this has been an international and not just an American development.
Three types of workshops/conferences have been run by the Group. General European meetings have been organised for researchers and users of qualitative reasoning (Genoa 1991). Technical workshops on individual topics were arranged (eg. distributed parameters, multiple models of single phenomena). Finally, international workshops have attracted significant American and Japanese participation (1990, 1992). Activities and collaborations such as these will be continuing.
It is now appreciated that the full scope of qualitative reasoning requires more of a connection with research and development in mainstream topics in knowledge-based systems than existed in the 1980s. Part of this appreciation has come from the experiences of the Group's members. In particular, qualitative knowledge needs many of the same approaches and results as are used in the general subject of knowledge representation. Hence integration of existing technical achievements of qualitative reasoning with such approaches has become the agenda for the near future. This is a two-way process: it is generally appreciated that software and techniques developed in research by participants in the Group, among others, can extend the usefulness of mainstream knowledge-based systems. This has proved to be relevant not only as a research issue, but also for the extended applicability of knowledge-based systems in commercial and industrial environments.
The main potential of the Group's work is seen as being the extension of qualitative reasoning as a subject: its boundaries are still fluid. For exploitation in applications, existing work (particularly in spatio-temporal modelling) is suitable for early use. In Europe, it simply needs wider awareness among possible users.
EH14 4AS Edinburgh