Interactive Case Memories as Knowledge Mediators between Strategic Planners
and Technical Simulation Systems in the Electricity Supply Industry
1. Abstract
This proposal is to investigate the use of Case Based Reasoning (CBR) [Kolodner,
93] in the form of Interactive Case Memories (ICMs) [Dearden, 95] as a
mediating technology for communicating knowledge between two domains of
expertise, each of which offers a distinct perspective on common
subject matter. We hope to uncover principles governing the use of CBR
in this mediating capacity, and to improve our understanding of the issues
arising when embedding case-based assistance to support the users of complex
systems.
As a case study, we aim to collaborate with EA Technology Ltd., via
ERCOS funding, in developing and evaluating an ICM to mediate between a
suite of tools that support the construction of simulations of electricity
supply networks based on detailed engineering knowledge, and strategic
planners working within the industry.
Strategic planners in the industry are concerned with the same artefacts
as power systems engineers, but their expertise may typically relate
more to financial planning and customer service issues and may not include
detailed knowledge of power systems. The ICM will be used to provide access
to, and interpretation of, the technical knowledge about the engineering
and mathematical modelling of electricity supply networks needed to permit
the strategic planners to construct and interpret detailed simulations.
The results of such simulations would be valuable in strategic planning
because of the need to evaluate the consequences of different investment
and maintenance policies.
This application of ICMs will extend knowledge of the use of case based
reasoning technology within user-interfaces to complex tools; will contrast
with previous work on interactive knowledge based systems which typically
assume that the user is either a novice in the domain, or an expert in
the same domain as the knowledge within the system; and will evaluate a
new approach to supporting the users of simulation tools.
2. Background
In most applications of interactive knowledge based systems, the user is
assumed either to be a knowledge worker within the same domain of expertise
as the system, as in medical and legal expert systems[Keravnou & Washbrook,
89], or to be a novice, as in the case of simple fault diagnosis systems.
In contrast, in this project we propose to investigate an application of
an ICM to act as a mediator between two knowledge domains that have only
a limited overlap. The two knowledge domains we intend to study are the
technical simulation of power systems, and the strategic planning of electricity
supply services. These two domains share common units of discourse, but
offer substantially different perspectives. Constructing and interpreting
simulations of technical systems is typically time consuming and expensive.
It may require detailed knowledge of: the simulation language; the domain
to be modelled; the mathematical modelling techniques employed; and an
understanding of the issues of validation and qualification of the simulation
results [Murray-Smith, 94]. As a result of this complexity, strategic planners
in technical domains are often unable to benefit from the understanding
and information that simulation might provide [Kiviat, 91]. EA Technology
are currently developing a suite of tools that is intended to support simulations
of power systems based on detailed electrical engineering knowledge. We
hope to collaborate with EA Technology by investigating ways in which an
ICM can act as a mediator delivering the knowledge required to develop
and interpret such simulations to strategic planners in the electricity
supply industry.
One proposal that has been made to simplify simulation is the development
of domain specific simulation facilities that provide the user with simple
building blocks that correspond to significant concepts in a domain. This
approach is intended to permit the construction of coarse grained simulations
without requiring detailed knowledge of the simulation language. Agentsheets
[Repenning, 93] is an interface development environment in which domain
specific simulation facilities can be simply constructed. However, coarse
grained, low-fidelity simulations alone are insufficient to support critical
business planning decisions. To support such decision making, strategic
planners will require support in moving from coarse grained simulations
within a domain specific environment, to the editing and interpretation
of hi-fidelity simulations that make use of additional technical knowledge.
We propose to exploit work on CBR and ICMs in order to provide the necessary
support. ICMs may be an appropriate technology for a number of reasons:
Firstly, the construction of meaningful and informative simulations
requires consideration of a large space of possible configurations of the
various elements (lines, transformers, customers etc.) of the power system.
Also, for each element of the system to be simulated, a range of parameters
must be considered, and the way in which the proposed strategy will alter
them must be described. ICMs have been suggested as an appropriate method
for supporting users investigating large search spaces [Dearden, 95]. Also
the construction of technical descriptions of network components is likely
to be easier for strategic planners to perform using a strategy of retrieve
and adapt than by attempting to construct the description from first principles.
Some preliminary work on embedding a design catalogue within a design environment
has been conducted at Colorado [Nakakoji, 93].
Secondly, an ICM can be characterised as an extensional representation
of a relation between sentences in two languages, a language of problems
and a language of solutions [Dearden, 95; Jankte, 93]. The problem of knowledge
mediation can similarly be considered as explicating the relation between
two languages of discourse. The extensional (i.e. case-based) representation
of such a relation has the advantage over intensional representations,
such as might be provided by a rule-based system, of being better suited
to problems where the domain knowledge is incomplete [Dearden & Bridge,
93; Chi & Kiang, 93].
Thirdly, we hypothesise that many of the components and parameter evolution
projections that will be required within simulations to evaluate a particular
policy, will share important similarities with elements from prior simulations.
By using an ICM to organise information from previous simulations, we hope
to exploit this repetition by providing access to precedent cases. Likewise,
in interpreting the output of simulations, the results of previous simulations
may be informative. This may be particularly important where a previous
similar simulation indicates a potential risk of undesirable consequences.
Finally, the system will need to communicate a very diverse body of
knowledge to the user. The use of precedents and cases may provide a natural
encapsulation of this knowledge for pedagogic purposes.
We believe that the results that we obtain will generalise not only
to other technical domains where simulation would be a valuable technique
for strategic planning, but also to other problems involving the
transfer of expertise between knowledge workers in different domains. For
example, supporting the users of typesetting systems who have expertise
in document authoring, or supporting users in managing the complex system
software that is now supported on personal computers.
3. Scientific / Technological Relevance
The simulation community recognises the need to provide improved support
for non-expert users of simulation facilities [Kiviat, 91; Schmidt
& Gomes,92; Murray-Smith,94]. CBR, and in particular ICM technology,
is increasingly popular for stand alone advisory systems. ICMs have
been applied to help desks[Dearden & Bridge, 93], technical fault diagnosis[Richter
& Wess, 91], design assistance [Domeshek et al., 94], and tutoring[Hirashima
et al., 93]. However, to date, only a small amount of work has been conducted
on the specific issues that arise for human interaction with CBR systems
[Dearden, 95]. In particular, we are not aware of any previous work that
seeks to apply CBR systems as knowledge mediating tools.
CBR has been used in the context of simulation by Wendel [94]. However,
Wendel's system used only a loose coupling between the interactive simulation
environment and the CBR system, and the intended users were experts within
the same domain as the knowledge represented within the simulator.
Our proposal aims to extend knowledge about the application of CBR
and ICM technology as knowledge mediators, and to develop new ways of supporting
the users of simulation tools.
4. Relevance to Beneficiaries
The chief beneficiaries of this project will include (not necessarily in
priority order):
the information technology industry, which will benefit from the investigation
of new ways of supporting users of complex computer systems;
the energy industry, and other network service industries which will
benefit from a better understanding of the relationships between strategic
planning and technical simulation of networks;
researchers interested in improving the user interface to simulation
facilities;
other researchers investigating the use of case based techniques
for tutoring and educational applications.
In particular we plan to collaborate with EA Technology Ltd. through
ERCOS funding. EA Technology hope to exploit our results in developing
user support for a simulation facility for electricity distribution companies
and other network utilities.
5. Dissemination and Exploitation
It is hoped that the results of this work will be directly utilised
through technology transfer to our industrial partners (EA Technology Ltd.
and Regional Electricity Companies).
Additionally, results will be disseminated by means of publications
at national and international conferences. In particular through the UK,
European and International Case Based Reasoning Workshops and decision
support journals; through UK and international HCI / CHI conferences and
HCI journals; through ACM SIGSIM and IEEE TCSIM conferences and simulation
publications; and through electrical engineering journals, UK workshops
organised by the IEE and international conferences on electricity distribution
such as CIRED.
6. The Programme
Aims
The objective of this proposal is to develop and evaluate an ICM to act
as a mediator between two domains of expertise both of which include
very specific and advanced technical knowledge and skills, but where the
intersection between the knowledge and skills of the two domains is relatively
small.
We have a particular application in mind as a case study in order to
investigate this general problem.. We propose to construct and evaluate
a prototype ICM to support strategic planners in the electricity supply
industry interacting with a suite of simulation tools based on technical
engineering models. The ICM will provide knowledge mediation so that strategic
planners can use the simulation facilities offered in exploring the consequences
of their decisions.
The existence of the simulation facility for a real world application
will permit empirical investigation of the effectiveness of ICMs as knowledge
mediators.
Goals
The project goals are listed below.
To develop a pair of knowledge models to record key elements of domain
knowledge used by engineers in simulating electricity supply networks,
and by strategic planners in evaluating the performance of such systems
in commercial and economic terms, which may include safety and environmental
factors.
To design and construct a domain specific interface for use in constructing
low-fidelity simulations of electricity distribution systems organised
around the knowledge models used by strategic planners. The success of
this work should be judged by the degree to which users with some basic
knowledge of power systems, and particularly strategic planners from the
industry, can construct simple simulations. The simulations produced
at this point are not intended to be high fidelity simulations; rather,
the purpose is to demonstrate the practicality of designing a user interface
to permit simple simulations to be constructed in this domain.
To construct an ICM to support users in developing high fidelity simulations
within the framework provided by the domain specific interface. The ICM
will assist users in searching a store of previous simulations and components
used in previous simulations, and adjusting the descriptions of these components,
in order to develop new simulations. The success of this work should be
judged by the ease with which the intended users, i.e. strategic planners
from regional electricity companies, can construct and edit simulations
to investigate a given problem, by using the ICM to search for previously
used components and parameter settings.
To extend the ICM to assist users in interpreting the technical results
of simulations in terms of their implications for customer service, and
to assist in evaluating the quality and reliability of the knowledge provided
by their simulations. The success of this work package should be judged
by evaluating the appropriateness of the knowledge fragments delivered
to the user investigating the outputs of the simulation. This evaluation
will require the co-operation of strategic planners to make use of the
system, and simulation experts from within EA Technology Ltd. to evaluate
the suitability of the knowledge fragments delivered.
To investigate generic design principles that govern the delivery of
on-line assistance using CBR techniques. This work will be the most difficult
to assess since it represents the abstraction of the key design lessons
from the project. The best assessment that would be possible would be the
testing of these techniques by application in a new domain. However,
that would be beyond the scope of the current project proposal.
References
R T Chi & M Y Kiang, 1993. Reasoning by co-ordination - an integration
of case-based and rule-based reasoning systems. Knowledge-Based Systems
6(2), pp103 - 113.
Andrew M. Dearden, 1995. The use of formal models in the design of
interactive case memories. DPhil thesis, Department of Computer Science,
University of York
Andrew M. Dearden & Derek G. Bridge, 1993. Choosing a Knowledge
Based System to Support a Help Desk. Knowledge Engineering Review, 8(3),
pp201 - 222.
E. A. Domeshek and J. L. Kolodner and C. M. Zimring, 1994. The Design
of a Tool Kit for Case Based Design Aids, in Artificial Intelligence in
Design: 94. J. S. Gero (Ed.), Kluwer.
T. Hirashima & T Niitsu & A Kashihara & J. Toyoda, An Indexing
Framework for Adaptive Setting of Problems in ITS,. in Artificial Intelligence
in Education: Proceedings of AI-ED 93, Paul Brna and Stellan Ohlsson and
Helen Pain(Eds.), pp90 - 97, Association for the Advancement of Computing
in Education.
E. T. Keravnou and J. Washbrook, 1989, What is a deep expert system:
An analysis of the architectural requirements of second-generation expert
systems, Knowledge Engineering Review, 4(3), pp205 - 233.
Philip J. Kiviat, 1991. Simulation Technology and the Decision Process,
ACM Transactions on Modelling and Computer Simulation, 1(2) pp89 - 98.
Janet L. Kolodner, 1993. Case-based Reasoning, Morgan Kaufmann.
Klause P. Jankte, 1992. Case Based Learning and Inductive Inference.
Technical Report, BMFT Verbundprojekt GOLSER 08/92, University of Leipzig.
D. J. Murray-Smith, 1994. Enhanced Environments for the Development
and Validation of Dynamic System Models, Mathmod, Vienna: Proceedings of
the IMACS Symposium on Mathematical Modelling, I. Troch & F. Breitenecker
(Eds.), pp9 - 13.
Kumiyo Nakakoji, 1993. Increasing Shared Understanding of a Design
Task Between Designers and Design Environments: The Role of a Specification
Component, PhD thesis, Dept. of Computer Science, University of Colorado,
Boulder.
A. Repenning, 1993. Agentsheets: A tool for building domain oriented
dynamic visual environments. PhD thesis, Department of Computer Science,
University of Colorado at Boulder.
Michael M. Richter and Stefan Wess, 1991. Similarity, Uncertainty and
Case-Based Reasoning in Patdex", Automated Reasoning: Essays in honour
of Noody Bledsoe, Robert S. Boyer (Ed.) Kluwer, pp249 - 265.
Heinz W. Schmidt & Ben Gomes, 1991. ICSIM: An Object-Oriented Connectionist
Simulator, Technical report TR91-048, ICSI, Berkeley.
Oliver Wendel, 1994. Case Based Reasoning in a Simulation Environment
for Biological Neural Networks, in Topics in Case Based Reasoning,
Selected Papers from the First European Workshop on Case-Based Reasoning
- EWCBR-93, Stefan Wess and Klause Dieter Althoff and Michael. M. Richter
(Eds.), pp424 - 435, LNCS 837, Springer Verlag.
