Informatics Programme in
Knowledge Modelling and Systems
Notes - 5-Dec-97
Area of Interest
The programme will focus on the methods and techniques
for the creation and maintenance of knowledge-based models and their
uses for a wider range of purposes.
The areas of interest (non-exclusively) includes:
For tasks such as:
- knowledge acquisition, requirements capture, task analysis
- knowledge representation, ontologies and problem solving methods
- knowledge modelling
- models of users, tasks, processes, capabilities and communication
- adaptive systems
- computer supported cooperative working
- KBS methods and knowledge engineering
In application areas like:
- system synthesis (design, planning, scheduling, configuration and layout)
- system analysis (classification, interpretation, diagnosis and explanation)
- system adaptation (control, learning, execution and enactment)
- sectors: finance, manufacturing, publishing, education, entertainment,
environment, petroleum, aerospace, defence and engineering (civil,
process, software, etc).
- cross-sector areas: information management/intelligent documents,
task and process management/intelligent workflow
This potential Informatics Programme in Knowledge
Modelling and Systems is at the early stages of discussion and, as we
argue below, there is scope for participation across Informatics and
beyond. Much of the value of the programme will stem
from the interaction between applied and theoretical research.
Some Informatics Division members will consider this
programme to be their main affiliation. Others may wish to devote a
smaller proportion of their effort to it. Both types of participation
Those expressing an interest as a main affiliation are:
Those expressing an interest for some part of their research work are:
- John Kingston (AIAI)
- Ann Macintosh (AIAI)
- Dave Robertson (DAI)
- Qiang Shen (DAI)
- Austin Tate (AIAI)
- Others to be included
Research students who have expressed interest in the
- John Lee (HCRC)
- Chris Mellish (DAI)
- Bob Muetzelfeldt (Ecology) [with the possibility of a linked programme
- Others to be included
Regular visitors to the University or regular project collaborators
from other organisations who have expressed interest in the
- Howard Beck (DAI, supervisors Austin Tate and Chris Malcolm)
- Daniela Carbogim (DAI, supervisors Dave Robertson)
- Alberto Castro (DAI, supervisors Dave Robertson and Bob Muetzelfeldt)
- Jessica Chen (DAI, supervisors Dave Robertson, Jussi Stader and Qiang Shen)
- Peter Funk (DAI, supervisors Dave Robertson and Gillian Hayes)
- Yannis Kalfoglou (DAI, supervisors Dave Robertson and Austin Tate)
- Renaud Lecoueche (DAI, supervisors Dave Robertson and Chris Mellish)
- Edjard Mota (DAI, supervisors Dave Robertson, Alan Smaill and Bob
- Steve Polyak (DAI, supervisors Austin Tate and Dave Robertston)
- Gerhard Wickler (DAI, supervisors Austin Tate and Alan Bundy)
- Others to be included
- Jaume Agusti (IIIA, Barcelona)
- Nigel Shadbolt (Nottingham)
- Others to be included
It is becoming common to build models to express our understanding of
problems and systems. Academically, this is a foundational component
of culturally diverse sub-areas (such as requirements engineering,
knowledge acquisition and simulation modelling). These academic
boundaries are less meaningful to industry where it is not considered
remarkable, for example, to use a knowledge representation method
drawn from the symbolic knowledge representation community but draw
conclusions from it using inference mechanisms from the neural network
community. This difference in emphasis, between broad pragmatism in
application and narrowly focused research, is reflected within
Informatics at present. Academic coverage is localised and patchy so
it has not been possible to build or maintain a critical mass in this
area. Our industrial activity in this area (primarily through AIAI)
is well established but is limited in the scope of work it can perform
and in the visitors and students with which it can afford to interact.
A programme combining both theoretical and applied
strands of Informatics could provide academics with more opportunities
to tackle the type of heterogeneous problem which is typical of this
area, whilst ensuring that our applied work has more flexible access
to academic results.
One of the strengths of this programme is its connection
to "the sharp end" of application, and this has normally meant
tackling problems from sources external to the University. Although
the ability to attract external funding is a measure of our success,
perhaps we should be prepared occasionally to take our own medicine.
The aim of the University is to advance and disseminate knowledge and
understanding through research and teaching. We have been slow to
exploit the possibilities of formal knowledge representation and
modelling to help us understand and manage parts of this process.
Other UK universities are already in pursuit of this goal (a recent
example being the Open University's new Knowledge Media Institute) and
our programme would be well placed to explore what could
be achieved at Edinburgh, both within and outside of Informatics.
The ACM Workshop on
Strategic Directions in Computing Research describes (from a US
perspective) key directions for future research in Artificial
Intelligence (AI), Formal Methods (FM), Human Computer Interaction
(HCI) and Software Engineering (SE). Although the style of
description varies between areas, several interests within the
Knowledge Modelling Programme recur as key directions in
different areas. Some of these are given below, simply to illustrate
the scope for interaction across Informatics. This list is not
definitive of the programme's aims.
- Composition of larger models from smaller elements. In FM this is
considered an issue of specification composition. In SE the elements are
distributed components. In AI a key concern is to construct knowledge
- Decomposition of problem descriptions. In FM this is the problem of
decomposing complex properties into simpler, local ones. In SE the
driving force is the Web, which raises new demands for highly distributed
decomposition and re-assembly of modular systems. In AI we have growing
interest in the development of ontologies for describing problems and
problem solving methods in general of in specific domains.
- Abstraction from the complexity of real situations to idealised
descriptions which it is useful to model. In FM this is viewed as a
prerequisite for appropriate specification and verification. In SE a key
issue is the formalisation of informal domain-specific notations. AI
tends to look to psychological theories of problem solving and to use these
as clues to the essential features of problem descriptions.
- Reuse of knowledge. In AI this is the longstanding goal of
knowledge expressed using standard representations and re-deployed for
different purposes. In SE a more prominent problem is to determine design
patterns with properties which are invariant across different uses.
- Merging of disparate models. A major motivation for HCI is the rise
of "ubiquitous computing", which requires non-specialists to interact at an
appropriate level with a wider range of different systems. SE emphasises
the construction side of this problem, through the need for multiple views of
designs. AI makes use of methodologies which support multiple perspectives
- Coping with large search spaces. Model checking is a major growth
area for FM. In AI there is a persistent demand for large, high
performance knowledge based systems. SE anticipates a need for mobile
agents with models of their locality and the ability to self-limit search.
- Evolution of designs. A long term goal of AI is to understand the
relationship between the "rationality" of designs and that of systems. In
SE a vehicle for this is through the use of patterns and abstractions,
while in FM there is a need for better understanding of appropriate
- Integration with existing practices. In FM and SE a key aim of
standard practice is to maintain parity between specification and
system. As well as the use of mathematical foundations and logic
theories, AI tends to concentrate on the possibilities for automation
of scientific methods by formalising specialist expertise.
- Focussed systems. A priority for SE is to develop systems for
domain-specific programming. HCI sees opportunities for "end-user"
programming, where non-specialists can undertake limited forms of high
level design. AI views the construction of programming assistants as a
key domain of application.
- Co-operative systems. The aim of allowing formal systems themselves
to co-operate is raised by both SE (in terms of co-operating agents) and AI
(in terms of agent societies).
- Visualisation of information. The need for visual query languages is
raised by HCI. In SE this is an issue for design, where different
visualisations may be needed for different views on a problem. AI
emphasises the need for multiple modality communication.
- Analysis of effectiveness. In HCI there is a growing demand for
participatory design, in which clients are more closely involved in
design life-cycles. FM raises the problem of ensuring that gain from
formalisation increases incrementally with effort. SE emphasises the need
for methods which reinforce best practice, whilst a key topic in AI is the
selection of the most appropriate approaches for standard purposes.
- Lightweight methods. A concern in FM is that methods should be easier
to introduce and produce pay-back earlier. SE suggests that one way of
reducing development cost may be through domain specific languages. A
trend in AI is to "intelligence prosthesis", which uses formal modelling to
enhance rather than replace human skills.
These proposals do not have official backing and express suggestions
for wider discussion.
The latest version of this document
is available on-line in HTML form at
updated: Thu Aug 27 15:06:01 1998
by Austin Tate (email@example.com)
and David Robertson (firstname.lastname@example.org).