REF: a Practical Agent-Based Requirement Engineering Framework Paolo Bresciani 1 and Paolo Donzelli 2 1 ITC-irst Via Sommarive 18, I-38050 Trento-Povo (Italy) bresciani@itc.it 2 Department of Computer Science — University of Maryland College Park - MD (USA) donzelli@cs.umd.edu Abstract. Requirements Engineering techniques, based on the fundamental no- tions of agency, i.e., Agent, Goal, and Intentional Dependency, have been recog- nized as having the potential to lead towards a more homogeneous and natural software engineering process, ranging from high-level organization needs to sys- tem deployment. However, the availability of simple representational tools for Requirements Engineering still remains a key factor to guarantee stakeholders involvement, facilitating their understanding and participation. This paper introduces REF, an agent-based Requirements Engineering Frame- work designed around the adoption of a simple, but effective, representational graphical notation. Nevertheless, a limited expressiveness of the graphical lan- guage may constrain the analysis process, reducing its flexibility and effective- ness. Some extensions are proposed to enhance REF capability to support require- ments engineers in planning and implementing their analysis strategies, without affecting however REF clarity and intuitiveness. 1 Introduction Agent- and goal-based Requirements Engineering (RE) approaches have the potential to fill the gap between RE and Software Engineering [5, 4]. The concepts of Agent, Goal, and Intentional Dependency, in fact, applied to describe the social setting in which the system has to operate, lead towards a smooth and natural system development process, spanning from high-level organizational needs to system deployment [4]. Goals are valuable in identifying, organizing and justifying system requirements [14, 2], whereas the notion of agent provides a quite flexible mechanism to model the stakeholders. However, the concrete application of such approaches has been until now limited only to few case studies. Several causes of this still immature adoption of agent- and goal-based paradigms for RE may be identified. Below we consider only two of them. First, although the notion of goal is central in some RE consolidated approaches like i* [15], GBRAM [1,2], and KAOS [8], an integrated and comprehensive require- ments analysis methodology, clearly linked, or link-able, to the subsequent phases of software development, still is an open issue. At best of our knowledge, only the Tropos methodology [5, 4] fully addresses this issue. Yet, not full consideration has been given