Onderzoekslijnen

Design & Evaluation of Intelligent Systems

1. Introduction

Modern technology creates new possibilities to cater for our basic information and communication needs. It also supports sharing by two or more participants of the same professional or consumer application, based on real and/or virtual presence.

Technological advancement, however, creates its own problems: potentially useful applications may become so complicated that their benefits are overshadowed by their complexity of operation. One way of achieving usability and ease-of-use is to replace overt user action by anticipatory system behavior: the application is then sufficiently intelligent to support the user in a proactive way.

System intelligence implies knowledge of the user, his/her goals and intentions, and the context of use. It also requires information about the application domain. All these pieces of information are inherently dynamic. Therefore the essence of system intelligence is adaptivity and adaptation. This is the foundation of contextually and situationally appropriate system behavior.

The user, too, has to be adaptive: he/she should learn which way of interacting with the system leads to optimal performance.

In its second phase, IOP-MMI will address the following key question: Which dynamic knowledge (of one another) should systems and users acquire and apply in order to optimally achieve their goals?

2. Focus of the MMI programme

The IOP-MMI programme will focus on the Design, Implementation and Evaluation of Intelligent Systems. This choice is in line with the general topic of Ambient Intelligence that is stimulating a large part of the European ICT research efforts, both in the 6th Framework and in ITEA. The focus is also in agreement with one of the dominant areas of interest in international conferences such as CHI, INTERACT, HCI International, Mobile HCI, Work with Computer Systems, Int. Conference on Systems, Man and Cybernetics, etc.

The programme will restrict itself to multi-user applications, either human-human interaction supported by an intelligent system or multiple users interacting with such a system. Non-networked, single-user PC applications are outside the scope of the program.

It may be helpful to relate the focus of the IOP programmeto Figure 1 below, borrowed from the ACM SIGCHI Curriculum. The figure summarises the many disciplines that, together, cover most of the combined expertise that is required to design a proper man-machine interaction solution. In essence, the MMI programmewill focus on the upper right-hand corner of that figure: U3 Human-machine fit and adaptation.

Figure 1: The domain of User-System Interaction and the focus of IOP MMI.

Other areas in the overall domain are explicitly not in the focus of the main generic research questions of this tender; although most of these other topics have to be covered according the recommended user-centered design approach. For instance, knowledge and know-how pertaining to H1, H2, H3, C1 and C4 are considered as a given, of which the state-of-the art can be applied in the design of intelligent systems. User-system dialogue (C2-C3) and its management (C5) will be part of any implementation, but the emphasis should be on the nature of the dialogue and the meta-communication between user and system, not on the language and speech technologies that support this interaction. The same holds for the lower part of the diagram: implementation and evaluation are part of user-centered design methodology, but no novel techniques will be developed.

3. Adaptivity: generic research topics

Adaptivity is widely recognised as the major characteristic of Intelligent Systems. It is a prerequisite to meet the needs of heterogeneous user categories, to guarantee optimal information load, and optimal offering of services and multimedia content. The system behavior, through which adaptivity takes place, is characterized by several attributes: constituents, determinants, goals and rules.

•  adaptivity constituents - those aspects of the interaction that can be adapted, e.g. information content, presentation primitives (interaction techniques, I/O devices, media and modalities)
•  adaptivity determinants - those factors that guide the adaptivity process, e.g. adaptivity based on user characteristics, or on the nature of the task, the context or the application;
•  adaptivity goals - the particular objectives that are to be served by the adaptivity process, e.g. comply with users behaviour in space and time, minimize number of errors, optimize efficiency or effectiveness;
•  adaptivity rules - those rules that drive the instantiation of the adaptivity constituents, based on the state of the adaptivity determinants, and the overall design rationale of the particular application/service.   

Against this general background, IOP-MMI specifically invites project proposals that address research topics dealing with adaptivity at the levels of the overall system architecture, the media and the content, as well as context-dependent multi-modal interaction, and the observed/ measured users behavior. Sub-areas to investigate are:

User > System Adaptation

The required amount of users instruction/training to build the proper users expertise to operate the system according the specified procedures, protocols, and laws. This has to consider the [dynamic] man-machine function allocation level and the overall level of automation.

System > User Adaptation

The way and the amount of adaptation of the system to the users has to be investigated. Specific issues are: the nature and amount of context awareness of the system; when and how user profiling has to be done; what and how the system memory contains user information derived from sensor-based users behavior observation and analysis.

User

A particular meta-communication between user and system is necessary to provide the users with control over the adaptive system: customization of the system behavior (system input), sufficient explanatory transparency about the actual system state (system output), agent-based user support for this kind of meta-communication, e.g. virtual assistance, recommendation component, etc.

User 

The multi-user interaction with an intelligent, adaptive system is primarily targeted to control continuous technical (e.g. process control industry, aviation, etc) and/or social processes (e.g. child care, health care and monitoring, security, etc).

4. Applications and platforms

Intelligent systems can support a wide variety of stationary and mobile applications in the professional and non-professional domains, so that the generic questions raised above can be applied to problems in the industrial and the consumer space.

The design and evaluation of intelligent systems requires an infrastructure that is not readily available in the Netherlands. Therefore, it is the intention of IOP-MMI to create in one or more locations a software/hardware platform that the projects will have to use to demonstrate and evaluate their applications (even if they may have been developed off-line).

Next to offering a unique infrastructure, the platforms will create convergence in the applications and thus enhance the focus in the programme. Further benefits are that the platform(s) will promote the realisation a few national centres of excellence for MMI research. In turn, these will help to safeguard continuity in the research domain after the IOP programme has ended. Finally, the platforms will also be vehicles to facilitate the demonstration and transfer of research results to business and industry.

The (minimal) specification of the application platforms is under discussion in a Working Group with representatives of industry and academia.