SODA: A Roadmap to Artefacts

An artefact for MASs is an entity not driven by an inner goal (as agents are), but used by agents to achieve their own goals. In this paper, we assume agents and artefacts as first-class entities in MAS engineering, and claim that agent-oriented methodologies should exploit these two abstractions as the basic bricks for the whole engineering process. As a first testbed, we take the SODA agent-oriented methodology and draw a possible roadmap for its extension toward the notion of artefact. 1 Agents and Artefacts for MAS Engineering Agents never live alone. Agents coexist with other agents in a MAS (multi-agent system) within an environment where they act and interact. Independently of the specific agent definition adopted—among the many available—, the agent abstraction alone is not enough to fully model the environment in a natural way. In fact, many environmental items, simply, are not agents: instead, they are something inherently different, entities (objects, instruments, tools) that are to be used by agents, rather than agents themselves. Following the lexicon originally introduced by Activity Theory [1] and later borrowed by MAS coordination [2], we refer to such items as artefacts. Artefacts are objects explicitly designed to provide some function, which guides their use [3]. Typically, artefacts take the form of objects or tools that agents share and use to support their activities, and to achieve their (individual and social) objectives. By adopting a cognitive perspective over systems [4], agents are the entities of a system that are characterised by some goals to be pursued, whereas artefacts are the entities that are not intrinsically characterised by a goal (they are not goal-oriented). Instead, artefacts are characterised by the concept of use, where an agent using an artefact for its own goals implicitly (and temporarily) associates an external goal to the artefact itself. Coordination artefacts are a case of particular interest in the context of agent societies, where they are usually exploited to achieve or maintain a global behaviour which is coherent with the society’s social goal [5,6]. As such, a coordination artefact is an essential abstraction for building social activities, in that it is crucial both for enabling and mediating agent interaction, and for governing the social activities by ruling the space of agent interaction. Indeed, at a closer sight, any activity carried on by the components of a system—individually or cooperatively—cannot be really understood (sometimes, even conceived) without considering the artefacts that govern the components’ actions and interactions [6]. More precisely, on the one hand, coordination artefacts mediate the interaction between individual agents and their environment (including other agents); on the other, they capture, express and embody those parts of the environment that are to be designed and controlled in order to support agent’s activities. In the end, along with agents, artefacts constitute the basic building blocks both for MAS analysis and modelling, and for MAS development and actual construction—i.e., real first-class abstractions available to engineers throughout MAS design and development process, down to run-time. So, agents and artefacts can be assumed as the two fundamental abstractions required to model and shape the structure of MASs: a MAS is made by agents speaking with other agents and using artefacts in order to achieve their goals. However, in order to show this conceptual framework worthwhile, and prove its effectiveness, we should be able to show how a MAS could be actually built using agents and artefacts: for instance, experimenting with some well-founded agent-oriented approach adopting agents and artefacts as its basic abstractions. The fact is that, no known agentoriented methodology does this today: as a result, to put our framework to test, we are forced either to invent a new agent-oriented methodology, or to extend an existing one with the notion of artefact. Among the many others, SODA [7] (whose basics are recalled in Section 3) is an agent-oriented methodology for the analysis, design and engineering of agent-based systems, specifically focussing on inter-agent issues. As such, SODA heavily relies on the notion of coordination model [8], which deeply influences its abstractions and mechanisms, leading engineers to build MAS social infrastructure on top of a coordination infrastructure. So, in particular, social rules are expressed as coordination laws, and embedded into coordination media. Since coordination media are to be seen as a sort of ancestors of coordination artefacts (and more generally of artefacts for MASs [9,3]), it seems quite natural to choose SODA as our testbed methodology for extension toward artfacts. In fact, the original SODA formulation does not include artefacts explicitly as such, mostly because the full development of the artefact notion in the MAS context is subsequent to the first articulated definition of SODA [7]. Furthermore, a recent research development has lead us to introduce in SODA a simple layering principle (called zooming) which makes it possible to scale the representation details with the complexity of tasks description (SODA+zoom [10]). This development has clearly shown that a re-formulation of SODA in terms of artefacts could not be delayed any longer, for both theoretical and practical reasons, apart from obvious reasons of coherence and conceptual elegance. Accordingly, this paper traces a roadmap toward the forthcoming SODA+artifacts. First, we shortly outline the most relevant features of artefacts and discuss a possible taxonomy (Section 2), then we recall SODA basics and briefly present the novel SODA+zoom (Section 3). In Section 4 we show the impact of the introduction of artefacts in the SODA approach, sketching a possible re-formulation which also relies on SODA+zoom. In particular, we investigate the meaning and the consequences of introducing the artefact notion onto the above zooming principle (Subsection 4.1), and onto the several SODA models—role, society, resource, and interaction models (Subsections 4.2 through 4.4). Conclusions and further steps are finally drawn in Section 5. 2 Artefacts: Features and Classification The sources for a theory of artefacts can be found in a number of different research fields, ranging from organisational / psychological theories [1] to anthropology [11,12], and obviously including the area of coordination models [13]. Such a theory, first developed for coordination artefacts [5,6] then generalised to artefacts for MASs [3,9], is shortly sketched in the remainder of this section. In particular, Subsection 2.1 outlines the main features of artefacts for MAS, and lists a number of desirable artefact properties. Subsection 2.2 outlines a possible taxonomy for artefacts, meant to be used both as a classification criterion and as a model for a well-principled methodology for agent-oriented engineering. Finally, Subsection 2.3 discusses in principle the impact of the notion of artefact upon an agent-oriented methodology.