Complex software systems are extraordinarily difficult to understand and to manipulate. This work extends earlier work on visual programming systems [] [] to support visualisation of program structures using self organising systems and virtual reality techniques.

Object visualisation is enabled by building a virtual world through which the user can navigate to explore and manipulate the object space. The view of the objects adapts so that they may be seen as distinct objects, agglomerations (eg. a cluster of objects may appear as a translucent mass with the individual objects barely visible) or as iconic representations.

The organisation of the objects in space is controlled by forces between the objects (determined by the relationships between these objects). The emergent structure that appears can be startling and can give considerable insight into the system's organisation.

Further work will expand upon these existing ideas and may involve the development of a more complete programming environment. Other aspects of the behaviour of systems will also be included, for instance, the incorporation of animation to show the dynamic behaviour of systems.

This large project is establishing a high speed communications network within the city. The project is wide ranging and will provide services for education and training, arts and leisure, council services etc. The network will provide a full range of multimedia services, including digital video. At present, the network infrastructure is being established along with an initial set of services.

It is intended that this framework will provide an infrastructure within which a wide range of research and development projects will be undertaken. In particular we will investigate the use of novel HCI techniques and other advanced technologies to provide access to appropriate information from within and without the system. It is also likely that a proportion of the work will involve evaluation of the use of the system rather than just development.

This project is being undertaken in collaboration with HR Wallingford and the school of Civil Engineering and is funded by EPSRC/DTI. The main objectives of the work are to improve upon existing user interface designs and, in particular, to devise enhanced visualisation mechanisms and to ensure that there is support for the users' deep tasks. Hydraulic models (for instance of rivers) are complex and their proper use depends upon the user having the appropriate skill and knowledge. In reality this may not always be the case, but also an unnecessary burden is placed upon the user who with existing systems needs to work at an unreasonably low level - for instance, when performing a sensitivity analysis of a model the user will need to explore the space around the solution to ensure that the model is stable. It is providing support for tasks such as this and in allowing the user to visualise the behaviour of the system, that the work is focussed.

Introduction: Virtually all man-machine interaction could make use of models of the users' goals and knowledge. This project is looking at the recognition of users' plans and from these the inference of the users' goals and knowledge.

Objectives: An observer can make inferences from the sequence of actions that a (computer) user performs. Most plan recognition systems are based upon the idea that the user generates a plan and then executes it step by step. A more realistic model is for the plan to be reviewed and perhaps modified (or abandoned) at each step, depending upon the response of the system to user behaviour.

This work seeks to build a recogniser which incorporates this approach, albeit modelled from the point of view of an observer rather than that of the user. The recogniser does not seek to model the user, but rather to build explanations of the users behaviour. These explanations are necessarily incomplete and unreliable, but represent a best guess as to what the user is doing.

Output of the recognition process is intended to feed (for instance) a user modeller, which will then be used to assist a tutor/help system since the tutor can relate advice to the user's experience.

This work is concerned with producing generating hyperdocuments which are adapted to the needs of individual users or groups of users. Work reported in [] shows how a general knowledge representation system can be used to represent domain specific knowledge from which multiple documents can be generated. More recently we have taken a similar approach to generate multiple HTML documents from information held in a database.

The work has looked primarily at off-line adaptation to cater for the needs of multiple classes of user. The delivery of the document has been achieved through a standard off-the-shelf system. Future work will look at more dynamic adaptation and will also build upon our other work on visualisation and user modelling

Work on ITSs and AI in Education has involved both practical developments (eg. in the CASCADE project []) and more long term research (eg. the PITT intelligent tutor for economics [] and work on plan recognition and user modelling).

Future work is expected to involve the use of neural networks to model some of the components of a tutor []. This work is likely to be undertaken in the context of some of the other projects in which we are involved.