Knowledge-Level Speech Acts

In this paper we try to answer the following questions: Is it possible to program solely at the level of an agent communication language? If this is the case, ii) what requirements and conditions need to be taken into account? We argue that, although a number of languages defining abstract communication primitives have been proposed in the past few years, knowledge-level programming can only be supported if a number of careful assumptions on the communication primitives and the underlying architecture are made, including asynchronous communication mechanisms, reliable message passing, and non-blocking primitives. To have a more rigorous understanding of these issues we proceed in a formal way. First, we postulate a set of requirements that an agent communication language should satisfy to be regarded as knowledge-level. These requirements also include some of the typical properties that distributed systems should satisfy in general. Then, we define a weak speech act based language, and we show that a synchronous version of the language does not satisfies requirements for knowledge-level programming. Finally, we show how an alternative asynchronous version of the language can be defined, which avoids the aforementioned problems. To prove these results, we introduce a general framework to reason on communication and concurrency aspects in the context in speech acts based languages. 1. This paper has been partially supported by the Italian Ministry of Universities (MURST) and by the ESPRIT BRA Project 9102-Coordination. 2. Department of Computer Science, University of Bologna, Via Mura Anteo Zamboni, 7, 40127 Bologna, Italy.