The programme of MEDLAR II aims to refine and develop the MEDLAR technologies, seeking:

- the synthesis of specific capabilities like abduction, analogy, planning and model building in a given combination of logics
- the development of the general framwework to produce specific reasoning capabilities
- exploration of the architecture for specialisation, implementation and integration in a generic practical reasoning agent.
A general framework is being developed for mechanizing deduction in the logics of practical reasoning (the philosophical term for logics of time, action, belief, knowledge and duty). The concept of a practical reasoner is being developed, an agent capable of acting autonomously and interacting flexibly with its real world environment.

So far, the outline of a general deductive framework for mixing various logics and calculi has emerged and is being developed to deal with specific reasoning capabilities.
Scientifically we seek to advance and disseminate a framework for knowledge representation and reasoning which is cognisant of the principles and practice of mechanised deduction in the logics of practical reasoning.

In practical terms we must reconcile the vision of a generic conceptual system, the "MEDLAR Practical Reasoner" with more concrete individual perspectives, so that we can develop small demonstrators for case studies in natural language dialogue and in robotic reasoning.

In MEDLAR II the concept of a practical reasoner will be developed, an agent capable of acting autonomously and interacting flexibly with its real-world environment. A human practical reasoner is distinguished from an artificial machine in many aspects, but for the purpose of the project we are chiefly concerned with two. The first is the rich variety of common sense reasoning methods - temporal, modal, probabilistic, relevant, abductive, hypothetical etc. The second aspect is an ability to interface and join together different modes of reasoning into one complex system: a practical reasoning agent.

APPROACH AND METHODS

The project will be motivated by case studies in natural language dialogue, the safety and security of databases and in robotics. It will address certain key steps in knowledge representation and reasoning with the logics of practical reasoning, namely the synthesis of specific reasoning capabilities like abduction, analogy, planning and model building. We seek to place these capabilities within an integrating framework which can be shown to have a basis for implementation.

Five tasks, co-ordinated by management meetings and scientific workshops are intended to deliver both scientific progress for the wider community and demonstrations of feasible application.

POTENTIAL

Future work in the MEDLAR II project includes the design of a MEDLAR Practical Reasoner, a concept which can be specialised in many ways but which will in practice be specialised to the needs of specific case studies in the project. Nevertheless, progress suggests that new technologies for abductive logic programming, efficient model building on parallel machines, and other novel approaches in the areas of the case studies will have potential for exploitation in more applied industrial projects.