From Agents to Artifacts Back and Forth: Operational and Doxastic use of Artifacts in MAS Michele Piunti Universit` a degli studi di Bologna DEIS - Bologna, Italy Email: michele.piunti@unibo.it Alessandro Ricci Universit` a degli studi di Bologna DEIS - Bologna, Italy Email: a.ricci@unibo.it Abstract— Recent approaches in Multi-Agent Systems are focusing on providing models and methodologies for the design of environments and special purpose tools supposed to ease pro- gramming in the large and scale up growing complexities. Among others, the Agents and Artifacts (A&A) approach introduced the notion af artifact as first class abstraction providing agents with external facilities, services and coordination medium explicitely conceived for promoting their activities. In this paper we analyse A&A systems by focusing on the functional roles played by arti- facts. In particular, we here investigate the function of artifacts once they are employed in the context of societies of cognitve agents, i.e. agents capable to reason about their epistemic and motivational states. In this context, a twofold kind of interaction is envisaged. On the one side, artifact rapresentational function allows agent to improve epistemic states, i.e., by representing and sharing strategic knowledge in the overall system (doxastic use). On the other side, artifacts operational function allows agents to improve the repertoire of actions, i.e., by providing additional means which can be purposively triggered by agents to achieve goals (operational use). Some of the outcomes of this approach are discussed along with test cases showing agents engaged in goal-oriented activities relying on the transmission of relevant knowledge and the operations provided by artifacts. I. I NTRODUCTION The artifact abstraction has been recently introduced in Multi-Agent System (MAS) [13] and MAS programming [20] as a basic building block to model and design agent environments and, more exactly, agent work environments. The notion of work environment used here refers to that part of the MAS – so developed by MAS designers and developers – which is perceived and used by agents as a first-class entity of their world, and which provides suitable functionalities and services that agents can exploit to ease their individual and social activities [18]. Artifacts – as in- troduced by the A&A conceptual model – can be conceived as basic module to structure such work environments, rep- resenting non-autonomous computational objects 1 that agents can dynamically instantiate, share and use as resources and tools to support and promote their activities. Mutuating the notion of ecosystems 2 or human societies, where individuals are supposed to behave and interact by means of shared 1 The notion of object is used here in its general term, meaning a dynamic entity with a proper identity 2 Introduced by Cristopherson in [5], ecosystem has been defined as “a natural unit consisting of all plants, animals and micro-organisms (biotic factors) in an area functioning together with all of the non-living physical (abiotic) factors of the environment”. Fig. 1. A metaphorical representation of a MAS according to A&A. knowledge, cultural transmission, memes [6], A&A states a clear separation of concern between the entities dwelling in a MAS: whereas agents can be considered as the proactive actors of the systems, exhibiting a purposive and autonomous behavior, artifacts are the non-autonomous entities, providing agent with facilities and special purpose tools to be exploited as external resources in order to serve a functional role [14]. According to this view, a MAS is designed and programmed in terms of an ensemble of agents that play together in a common (work) environment not only by communicating through some high-level Agent Communication Languages, but also co- constructing and co-using different kinds of artifacts, organised in workspaces (Figure 1 shows a metaphorical picture of a MAS in this perspective). The main source of inspiration underlying this view comes from human societies and research works in Activity Theory [12], remarking the fundamental role that play artifacts in our society in mediating and supporting human work, in particular cooperative work. Besides A&A, CARTAGO [20] has been introduced as a platform and infras- tructure providing a concrete computational model to program artifacts and a distributed runtime for executing artifact-based work environments, making it possible for agents developed on different agent platforms to dynamically join and work inside such environments [18].