Results

• The Albert II language (an agent-oriented language for requirements specification with real-time temporal constructs) has been defined intuitively and formally defined respectively in [5] and [6].
• Features of the Albert II language has been validated through the handling of a medium size case study with real-time, concurrency and distribution aspects [7].

• An analysis of the differences between the concept of an ``agent'' and the popular concept of an ``object'' with regard to the specification of information systems has been made [19]. The agent-based approach promises to provide a semantically richer framework for specifying distributed, heterogeneous information systems.
• For the specification language TROLL operational semantics for prototypical execution of specifications have been considered especially for the concept of object interaction [8,9].
• The development of a basic calculus for expressing and proving properties of objects and its use for designing information systems have been presented in [3,2]. Although this calculus was originally intended for object specification frameworks, it could be extended to a more advanced, agent-based specification framework for information systems due to its fundamental conception.
• For object specification languages like TROLL an extension which captures the evolution of object behaviour has been proposed and discussed in [20]. Whereas usually the behaviour of objects is fixed during specification time, this extensions allows to change the behaviour of objects during runtime of a system in a certain restricted way. Thereby, a more agent-like behaviour is obtained because, for instance, the change of behaviour can be understood as a change of goals for an agent.

• Results have been obtained on modal logic for the specification of the behaviour of multi-agent systems. As basis a blend of epistemic and dynamic logic is employed augmented with a special ability operator. Especially work has been done on specifying knowledge producing actions such as observations, communication, default reasoning and belief revision [10,11,15,14].
• An integrated framework for both ought-to-be and ought-to-do constraints has been proposed and investigated [1]. Furthermore, work has been done on the assessment of the infamous paradoxes of deontic logic with respect to the use of this logic in computer science and AI applications. Special focus has been put on a solution of the Free Choice paradox by means of so-called contexts. The logics considered are based on dynamic logic [17,4].

• A first combination of Petri nets and deontic logic has been carried out by modelling deontic states in Petri Nets [12].
• Another, more directly implementable approach is [16].

• Proposal of a new notion of guarded command based on the distinction between permission and obligation guards and of a new class of properties --- willingness, their formalisation in branching-time logic, and results on compositionality [18].
• Refinement techniques for actions in the context of object/agent-based systems specified in temporal logic [13].