KIM Project Conference: Knowledge and Information Management through Life
The KIM Project [1], known in full as Immortal Information and Through Life Knowledge Management: Strategies and Tools for the Emerging Product-Service Paradigm, is a £5.5 million research programme funded primarily by the EPSRC [2] and ESRC [3] and involving eleven UK universities. The purpose of the project is to find robust ways of handling information and knowledge — for example, product models and documentation of design processes and rationale — over the lifetime of project-services such as PFI hospitals, schools and military equipment, as well as enterprise-level strategies for this new way of working.
The first KIM Project Conference was held on 28–29 March 2007 at Loughborough University. The purpose of the event was to allow the Project’s many researchers to present the work they had accomplished so far to the rest of the Project team and to get some initial feedback from representatives of the Project’s industrial collaborators.
Information Systems
The first set of papers dealt with the issues surrounding the organisation and evaluation of information.
Saikat Kundu of the University of Leeds presented a technique for ascertaining the design requirements of information systems in engineering firms that have made the transition from supplying products to providing services. The technique involves analysing a ‘service product’ and producing a service blueprint which details the processes involved and how they relate to one another, both logically and chronologically. This can then be used as the basis of the design of information systems, in much the same way that process models for physical products have been used previously.
Yuyang Zhao of the University of Bath Engineering Innovative Manufacturing Research Centre (IMRC) presented a technique for establishing the quality and value of engineering design information, with the aim of using this as a filter to remove low-value ‘hits’ from search results in engineering information systems. The technique uses a Bayesian network based on a set of information characteristics — accessibility, usability, concurrency, accuracy and trustworthiness — to assess quality, and an additional two factors — relevance and cost — to judge value. The technique relies on absolute judgements being made, but does distinguish between judgements made on a personal level and those made at the enterprise level.
George Gunendran of Loughborough University presented a paper on using topic maps to help organise both structured and unstructured engineering information. Topic maps are a way of overlaying structure on a corpus of documentation. Topics are abstracted from the documents and each topic occurrence recorded as a link; the associations between topics are also recorded, forming the map. The paper showed how this technique could be used to group model parts into model feature families, and to relate particular features to specific manufacturing methods, while at the same time allowing different perspectives on the same set of data.
Information Capture and Representation
The second set of papers dealt with product, process and rationale models, and the automatic capture of metadata and rationale data.
Claudia Eckert of the University of Cambridge detailed some classes of knowledge that need to be preserved across long product lifecycles. Typically there are four stages to a product lifecycle: new product development, product maintenance, product upgrade (e.g. putting new technology into old vehicles) and capability continuation (designing a new product to replace the capabilities of an older product). The information flows that occur along this lifecycle and between products can be complex; for example, in a capability continuation context, information may flow from the most recent product of a similar design, rather than from the product being replaced. The knowledge most needed along the lifecycle includes: an overview of how the various parts of the design interrelate, the margins within which each part was designed to operate and the consequences of exceeding them, the rationale behind certain design parameters being set and the consequences of changing them, and the function-level importance of the design features.
Alastair Conway of the University of Strathclyde presented a vision of fully mobile shared workspaces with integrated tools for information and knowledge capture. The starting point for this vision was the LauLima system, developed as part of the DIDET Project [4]. LauLima is made up of two components: the LauLima Learning Environment, a wiki-based collaboration workspace, and the LauLima Digital Library, a digital object repository that also acts as an archive for wiki pages that have run their course. In essence, the idea is to capture information as it is created, in its raw form, and to apply robust versioning to the information, and automatically to add as much contextual metadata as possible. Combined with sophisticated search facilities, it should then be possible to generate visualisations of design evolution and the underlying rationale on request.
Robert Sung of Heriot-Watt University presented the results of a case study looking at extracting rationale from log files generated by a simple computer-aided design (CAD) system. The system in question was a modified version of the BAMZOOKi software provided by the BBC [5]. BAMZOOKi is a game in which one designs and races a ‘zook’ creature; the particular implementation in the research imposed certain design goals that had to be achieved by the zook, requiring the designer to go through several design-test-modify iterations. Over three hundred log files have been collected and analysed so far.
Csaba Salamon of Heriot-Watt University presented a new technique for watermarking CAD models. The technique involves rotating the co-ordinate system for each face of the model, and using the differences in orientation to encode data. The technique is resilient to most migrations between formats that support boundary representations of surfaces, and does not alter the original model or increase the file size in any way.
Alex Ball of UKOLN, University of Bath, gave a presentation on reliably migrating CAD models into and out of lightweight file formats, which are mainly used for visualising and sharing product model data. Deciding on the most appropriate strategy requires comprehensive information on the capabilities of the various formats and the reliability of the tools available. Even with one migration tool and destination format, there can be different migration routes giving very different results.
This section concluded with a feedback session allowing the industrial representatives present to comment on the papers presented so far. There was strong support for the direction taken by the researchers, with only a few points that the Project needed to take on board.
Learning Through Life
The third set of papers dealt with tools and techniques for supporting design learning and decision making.
Matt Geiss of the University of Bath Engineering IMRC expanded on George Gunendran’s presentation, looking at how topic maps can be used to unify product, process and rationale models to enable ‘double-loop learning’ [6]. The technique involves labelling all occurrences of a topic, such as a given dimension, across the various documents in which it appears, and using associations to show where the dimension came from and how it is used further. Topic maps with common elements can be combined to permit analysis across several similar development projects, allowing one to identify and correct systematic errors and erroneous assumptions.
John Rooke of the University of Salford questioned the existence of a product-service paradigm in the construction industry and suggested that such a paradigm might be provided by the software industry. In software, the design is the product, presenting the industry with a choice between policing the distribution of copies, or offering services around the product. Elements of open source development models may contribute methods for coping with the unpredictability of built environment lifecycles.
Khadidja Grebici of l’Institut National Polytechnique de Grenoble, and on placement at the University of Bath, presented a framework based on information maturity for handling uncertainty in collaborative design. Information maturity is the relative state of the development of the information with respect to satisfying a purpose; thus a set of information may be mature for one purpose and immature for another. This can lead to conflict and misunderstandings between, say, providers and users, or management and operational groups. Capturing how maturity evolves, and how compromises are reached between groups with different goals, is key to life-proofing information and improving collaboration.
Decision Making
On the second day, the first set of papers outlined the Project’s research into aspects of decision making and decision support.
Lauri Koskela of the University of Salford discussed the various approaches that have been taken over the years to integrate the various stages of the product lifecycle: from systems engineering and lifecycle assessment to public-private partnerships and product lifecycle management. This move from treating each lifecycle phase separately to treating them all holistically is not a paradigm shift, but rather a coming together of various approaches over the past fifty years. These approaches need to be unified and any gaps filled.
Murray Sinclair of Loughborough University presented an introduction to lifecycle decision support, with particular reference to organisational complexity. There are several forms of complexity that need to be addressed: the relationships between partners in extended teams and supply chains, the inherent complexity of products and how to realise them, and the structures within individual organisations. Put very simply, the only way to overcome complexity is with wisdom, that is, having someone within the organisation with sufficiently broad and deep knowledge of the systems involved. Unfortunately, Murray ran out of time before he could apply this to decision making systems.
Koray Pekericli of the University of Reading presented an introduction to decision making theory and decision support systems. Within the construction industry, four-fifths of knowledge is tacit, and tends to be passed from master to apprentice rather than between teams, hence decision support systems have tended to be knowledge-based: for example, directories of experts, networks of knowledge leaders and activists, best practice reviews and communities of practice. The Amazon Mechanical Turk [7] represents a possible new model of decision support, based on the principle of artificial artificial intelligence — that is, systems that behave like artificial intelligences but are in fact programmed using an ever-expanding set of simple human decisions. Perhaps by breaking decisions down into millions of small, simple decisions, a more robust decision support system could be produced.
Daniel Geiger of Liverpool University presented the results of a case study looking at the use of communities of practice (CoPs) in an international construction company. While people claimed to value the hi-tech knowledge management tools available to them, they tended not to use them or contribute to them, and instead relied on informal personal networks. Two types of CoP were identified: role-based and interest-based. The role-based CoPs tended to be stable, but were used mainly for top-down dissemination and team building, and therefore lacked innovation. Interest-based CoPs took a bottom-up approach to developing expertise and dealing with issues, and hence came up with more innovative solutions. However, the sustainability of interest-based CoPs was fragile; such groups only survived where they were given a physical meeting space, technological support, and real influence within the company.
Julie Jupp of the University of Cambridge presented a model for comparing decision making processes in design work. Starting from the input conditions, the initial stages of the decision making process are classifying the problem type (from looking at constraints, requirements, specifications, etc.) and then recognising the problem itself, its ambiguities and risks. The central part of the decision making process is perspective development, informed by both internal and external design conditions. Once the synthesis of all these elements has been performed, actions follow, resulting in decisions and their attendant risks and conflicts. The output conditions of the process are then fed back into the input conditions of other decision making processes. This model is particularly suited to identifying scenarios where there are strong dependencies between earlier and later decisions, and between decisions on different levels of the design process.
Incentivisation
The next set of papers were on the subject of giving contractual incentives to contractors to improve knowledge transfer and innovation within engineering projects.
Weisheng Lu of the University of Reading presented some preliminary thinking about incentivising knowledge and information sharing. At present, in the construction industry, knowledge and information doesn’t flow particularly well, leading to solutions being reformulated from scratch and mistakes being repeated. So far, the Project’s research has concentrated on enumerating classes and instances of barriers to knowledge and information flow, and the various incentives that could be used to overcome them. Field work will determine which incentives work in practice.
Will Hughes of the University of Reading was the only speaker to rebel and give a presentation without using slides; this suited his talk, which challenged commonly-held views about incentivisation. Contrary to intuition, contractors can make money on a contract while selling their product at a loss, through careful timing of payments to their own suppliers. Quite often, offering a contractor more money to finish early can be no incentive at all, if the contractor does not have another job to go on to afterwards. Even pain-share/gain-share contractual clauses can be ineffective if the contractor is forced to share too much of its gain with its supply chain and the client.
Jurong Zheng of the University of Bath Management IMRC presented the results of two case studies, looking at how two early Private Finance Initiative (PFI) projects were organised and directed. In both cases, exhaustive bespoke PFI contracts were used, with the emphasis on penalty clauses rather than benefit clauses; the inflexibility of these contracts caused some problems. Some benefit sharing was later introduced, some of it financial and some related to prestige. The issues still to be teased out include the relationships between contractual and relational approaches to running the project, and the role of formal and informal mechanisms for aligning the goals of the clients and the contractors.
Florian Täube outlined some of the research being done by the team at Imperial College, London. Technological advance is making it easier for globally distributed teams and global networks of organisations to work together. Increasingly, Indian firms used to providing services are moving along the supply chain into systems integration, thus challenging firms in the UK. The team are therefore looking at the place of outsourcing and offshoring in through life knowledge management, and how to incentivise work in globally distributed teams.
Selected Industrial Papers
The original intention for this last set of papers was to repeat some choice papers from earlier in the conference for the benefit of industrial representatives who were only attending the final afternoon of the Conference. In the end, there were too many papers accepted for this to be possible, so instead it was a case of saving the best, most industrially relevant, and most general papers till last.
The session began with Project Director Chris McMahon of the University of Bath Engineering IMRC explaining the mission of the Task 1.1 team, which is to come up with a unified model for product, process and rationale. The team’s research covers the whole spectrum of documentation, across various levels of structure, various ways of working collaboratively, and various stages of design and product lifecycle.
Rob Bracewell of the University of Cambridge introduced his Design Rationale Editor, DREd. This tool captures rationale in the form of structured spider diagrams, and is already being used with great enthusiasm in industry. In order to increase the tool’s range, Rob is working on bi-directional linking with other software: already it is possible to link DREd maps to Word documents and vice versa, and to create labels dynamically in DREd maps from data in Excel worksheets. Forthcoming improvements will see similar integration with OpenOffice.org and a version that runs on Linux.
Alastair Conway of the University of Strathclyde presented the results of an experiment to compare the value of traditional minutes of design meetings with ‘multimedia minutes’: video and audio recordings with coded transcripts. The key point of comparison was the number of decisions recorded by each method, and how many of these decisions were ‘critical’, that is, impacting directly on the direction of the design process. On both points, the multimedia minutes recorded between two and three times as many design decisions, even when the traditional minutes were enhanced by sticky-note annotations describing resources used, sketches made and gestures made. On the down side, coding the transcripts took twenty times longer than producing traditional minutes.
Stuart Green of the University of Reading presented an overview of the various different interpretations and implementations of the product-service paradigm. Within the literature, the three main interpretations of product-service are: as a shift from delivering a product to providing a service; as a move to providing more integrated, customised solutions to meet customer need; and as a change of emphasis from a customer’s immediate practical needs to a customer’s own service commitments. In practice, firms are involved in a mixture of product delivery and product-service contracts at any one time, and are having to balance their need for stability and the demand for dynamism.
Stewart Johnstone of Loughborough University presented an analysis of the human resources implications of product-service provision using the analogy of hunters and farmers. The culture of product provision in civil engineering is rather like hunting: having a set goal and time frame, working pragmatically, giving it one’s best shot then moving on. The culture of maintenance contractors is rather more like farming: cultivating relationships, building an understanding of the client’s needs and keeping a long-term perspective. The challenge in PFI contracts and the like is finding a way of joining these two cultures up: it is as much about influencing mentalities as it is about organisational structures.
Mike Lewis of the University of Bath Management IMRC characterised the product-service paradigm as being about buying complex performance outcome. The research agenda on the procurement of complex services has three dimensions. The first concerns temporal aspects of inter-organisational issues such as governance modes, passing of incentives along the supply chain, and innovation. The second concerns the comparison and interaction of inter- and intra-organisational issues such as information flow, outsourcing and offshoring. The final dimension concerns the temporal aspects of intra-organisational issues such as organisational structure and the management of contracts in the face of rapid staff turnover.
Closing Plenary Session
The final closing session was another industrial feedback session similar to that held on the first day. Among the topics considered were e-mail (increasingly an important source of evidence in legal proceedings), whether good knowledge management could stifle innovation, and a reminder of the importance of the bottom line. Industry will always manage to get the job done; the issue is whether the techniques and tools coming out of the Project can save time and money over and above current recovery methods or re-doing from scratch.
Conclusions
This was a densely packed conference and a source of some exhaustion to all concerned. With so many papers to get through, timings had to be strictly controlled, and so this was the first conference I know of where over-running speakers were actually given a red card and sent off. Nevertheless, it was informative and enjoyable, with high-quality presentations and promising research on display. The KIM Project Conference was particularly valued by the industrial collaborators, and no doubt a similar event will be held again later in the Project.
References
- The KIM Project Web site http://www.kimproject.org/
- The Engineering and Physical Sciences Research Council Web site http://www.epsrc.ac.uk/
- The Economic and Social Research Council Web site http://www.esrc.ac.uk/
- The Digital Libraries for Global Distributed Innovative Design, Education and Teamwork Project Web site http://www.didet.ac.uk/
- The BAMZOOKi Web site http://www.bbc.co.uk/cbbc/bamzooki/
- For information on double-loop learning, see: Argyris, C., On Organizational Learning. 1999, Oxford: Blackwell.
- The Amazon Mechanical Turk Web site http://www.mturk.com/