Background

---

With the fast growth of the Internet, it is now being used in every aspect of human life. In particular, the online delivery of courses has led to a brand new field that all educators, both the advocates of traditional education and the developers of new methods, have to face—e-education, in which the learning process occurs in a distributed manner. Not only can the instructors and learners be geographically distributed, but also the sources of courses and the assembly and delivery of instructional material can be managed dynamically by coordinating distributed entities.

How can agent systems make distributed online learning possible? First, because of their adaptability, agent systems can adjust the instructional material according to the changing demands of a learner. An agent can reason based on its existing knowledge and past experience to determine the best action in the given circumstances. Moreover, it can accommodate new situations by learning. Second, as an agent is mobile, it can deliver a course dynamically. That is, by migrating from one running environment to another, the agent makes a “mobile” campus possible, allowing users to have a flexible schedule and a changing study venue. Third, collaborating agents make possible the merging, comparing, and optimizing of knowledge. This capability will change the process of preparing teaching material and the nature of the learning process, and should help decrease the amount of time required to complete a course.

We found that the dynamic nature of distributed agents in e-learning environments makes these agents ideal objects for modeling by Gamma languages. The concurrency and automation require that the modeling language not have any sequential bias and global control structure. As well, the dynamic nature and nondeterminism of the interaction between an agent and its environment are suited to a computation model with a loose mechanism for specifying the underlying data structure. Therefore, the chemical-reaction metaphor provides a framework for the specification of the behavior of an agent. Data, which move around the Internet, can be well modeled by chemical reactions, which are represented by the multiset in Gamma languages. In addition, the architectural features of a multiagent system can be succinctly captured by higher-order multiset processing mechanisms.

Gamma languages have been used in the study of coordination programming (Holzbacher, 1996) and have proved to be powerful tools for describing and reasoning about the overall properties of the coordination system’s architecture, such as, the connection styles of the coordinating entities in the system. The major difference between a distributed agent system and a coordination system is the transitional properties of the agents: their mobility, automation, and adaptability. To capture these properties, we used the higher-order extension of the Gamma language, which effectively separates interagents’ operations from state transitions on an agent in a local environment.

---