A requirements elicitation framework for agent-oriented software engineering.

The hypothesis of this research is as follows: "Conceptual modelling is a useful activity for the early part of gathering requirements for agent-based systems." This thesis examines the difficulties of gathering and expressing requirements for agent based systems, and describes the development of a requirements elicitation framework. Conceptual modelling in the form of Conceptual Graphs is offered as a means of representing the constituent parts of an agent-based system. In particular, use of a specific graph, the Transaction Model, illustrates how complex agent concepts can be modelled and tested prior to detailed design specification, by utilising a design metaphor for an organisational activity.Using an exemplar in the healthcare domain, a preliminary design framework is developed showing how the Transaction Agent Modelling (TrAM) approach assisted the design of complex community healthcare payment models. Insight gained during the design process is used to enrich and refine the framework in order that detailed ontological specifications can be constructed, before validating with a mobile learning scenario. The ensuing discussion evaluates how useful the approach is, and demonstrates the following contributions: Use of the Transaction Model to impose a rigour upon the requirements elicitation process for agent-based systems; Use of Conceptual Graph type hierarchies for ontology construction; A means to check the transaction models using graphical inferencing with Peirce Logic; Provision of a method for the elicitation and decomposition of soft goals; The TrAM process for agent system requirements elicitation.