This work reports about ongoing research that abstracts and generalizes a previously discussed, advanced software architecture, particularly suitable to represent crucial components of Intelligent Tutoring Systems. A distributed ITS framework is extended in terms of social categories of intelligent actors. The basic social entities are discussed, explaining how their activities are organized in a distributed environment performing concurrent computations. In order to highlight the benefits of the architecture, we describe in detail the diagnostic module, considered traditionally at the same time necessary for modeling the student’s behaviour and therefore tune the dialogue to the user’s needs, but also computationally too complex. We deepen our concerns and present our results showing how to enhance the available tools and languages for needs and problems specific to ITSs. Within this framework, we focused our attention on the design and implementation tools that become a priority: (a) because the underlying primitives imply a mental model of the virtual machine that has a non trivial influence with what the developer will privilege in her/his design and implementation, (b) because the primitives of the tool will eventually map or not onto the real technologies available, such as multiprocessors, interfaces with human agents and the network seen as a world-wide computer. Among the available software design tools, actor languages are certainly the most desirable ones to start with even if they are not easy to use.
