Alan Rector
Users are always right...even when they are wrong: Making Knowledge Representation Useful and Usable
Users - domain experts in whatever field - have problems to solve and knowledge that they use to solve them. They are the experts on the nature of those problems, the knowledge used to solve them, and the criteria for success. Users are experts in the "problem space", although like any experts, much of their expertise is tacit.
The Knowledge Representation community is expert on the "solution space". They are experts in solving problems right - at least right in terms of their own criteria. Unfortunately, their criteria for being right do not always match users', and KR tends to be divided into subcommunities, each expert at using particular tools, each with its own criteria for rightness. All too often - perhaps inevitably - users' problems are made to fit a subcommunity's favourite tools rather than the tools being chosen to fit users' problems. Furthermore, subcommunities may insist that users reformulate their problems in terms of the subcommunity's criteria and solutions, in which the user rightly has little expertise.
Knowledge Representation may be poised to become a mainstream technology, stimulated by the Semantic Web, successes in bioinformatics and other fields. If it is to become so, it will be through a much closer partnership with users and users' intermediaries to solve their problems, with KR tools where appropriate, with other tools where not. Much more research effort needs to be devoted to understanding matching KR tools to user tasks and to articulating the strengths, weaknesses, and trade-offs of different approaches so that they can be explained and discussed with users. Since uses' problems are unlikely to fall neatly factored along the lines of individual KR technologies or even to be susceptible to complete solutions, effort on hybrid and heuristic systems is needed.
In this talk we discuss issues arising out of experiences working with users at the "knowledge-face" and suggest some high priority areas for outward looking research, explanation, and action.
Biography
Alan Rector is Professor of Medical Informatics in the Department of Computer Science at University of Manchester. He was presented with the first award for time achievement in Health Informatics by the British Computer Society in 2003. He received his BA in Philosophy and Mathematics from Pomona College, his medical training at the universities of Chicago and Minnesota where he obtained his MD, and his PhD in Medical Informatics from the University of Manchester. He has been a visiting senior scientist at Stanford University and consultant to the NHS Information Authority, Hewlett Packard, and the Mayo Clinic. He is a member of the JISC Committee for the Support of Research, the National Cancer Research Institute Board for Bioinformatics, the Joint NHS/Higher Education Forum on Informatics, and the Board of the Academic Forum of the UK Institute for Health Informatics. He is also active in HL7, the main standards body for health informatics, and on the board of HL7-UK. Over the past twenty-five years he has led a series of projects on clinical decision support, medical records, and medical terminology. His work on clinical terminology and ontologies provided a key stimulus for the technologies which underpin the use of ontologies for the Semantic Web.