The following are examples of major research challenges which CISA researchers have accepted and are now pursuing with government and industrial funding.

• Knowledge lifecycles: Old fashioned knowledge-based systems tended to be simple, self-contained and of limited importance to business success. Their modern counterparts are often complex, interacting with other systems and may be business-critical. This means that we must provide "joined up" engineering which links the various stages in use of knowledge (from acquisition to decommissioning) and enables us to support these in concert. Through projects like the AKT-IRC we are developing this sort of engineering and developing the theories needed to understand it.

• Model integration: Normally our choice of representational styles and inference systems is conditioned by a particular style of modelling which we believe appropriate to the problem in hand. The ability to choose models appropriate to problems is a prerequisite for engineering but the proliferation of seemingly different models inhibits the development of unifying principles across similar types of problem. Through projects like AKT and I-X we are building the frameworks necessary to develop and share different types of problem-specific model through common underlying representations.

• Agent-based engineering: It is hard to build a multi-agent system and predict accurately what its behaviour will be. Even harder is the task of building an individual agent which will "do the right thing" within someone else's multi-agent system. The nub of the problem is that agent systems are not allowed precise expectations about the integrity of their environment or the reliability of the other agents with which they must interact. This demands that we bring engineering precision to "soft" concepts like negotiation, argument and belief revision, and that we understand how macro-behaviours of multi-agent systems may emerge from micro-behaviours of individual agents. We are doing this through projects like SLIE and I-X.

• Planning and activity management: We are exploring representations and reasoning mechanisms for inter-agent activity support.Planning and acting rationally are key capabilities for intelligent behaviour. The agents may be people or computer systems working in a coordinated and perhaps mixed-initiative fashion. We are exploring and developing generic approaches by engaging in specific applied studies. Applications include crisis action planning, command and control, space systems, manufacturing, logistics, construction, emergency procedural assistance, help desks, etc.

• Intelligent Interfaces: We are researching and developing intelligent multi-modal interfaces which can provide support to user tasks, languages, locations and capabilities. Projects such as GhostWriter, O-Plan and I-X are a basis for this work. Applications include multi-lingual and cell phone support for maintenance and for emergency procedures.