Representing Complex Multi–Agent Organisations in UML ⋆ Joaquin Pe˜ na, Rafael Corchuelo and Miguel Toro Dpto. de Lenguajes y Sistemas Inform´aticos Avda. de la Reina Mercedes, s/n. Sevilla 41.012 (Spain) E–mail: joaquinp@lsi.us.es, web page: www.tdg-seville.info Abstract. Interaction has been proved one of the main sources of com- plexity in Multi-Agent Systems (MAS) and many researches are working on techniques to palliate it. Furthermore, Organization modelling tech- niques lies on representing the groups of agents which are related by some kind of interaction. Current UML approaches represent these rela- tionships as a set of binary links usually represented as stereotyped UML associations not providing abstraction tools to manage the complexity derived from interactions. In this paper, we argue for multiparty links in order to increase the level of abstraction of organization models and thus, their ability to manage complexity. keywords: Complex systems, organization modelling, multiparty inter- actions, agent protocol descriptions, UML. 1 Introduction 1.1 Organizations and Complexity The organizational metaphor has been proved one of the most appropriate tools to engineer Multi-Agent Systems (hereafter MAS) being used as the abstraction which guides the analysis and design of MASs, e.g. [14]. Organization of Multi- Agent Systems is usually seen by many researchers as a collection of interacting roles [8], e.g. GAIA [14] or AUML social structures representations [11]. An or- ganization shows the groups of agents formed in the system due to get benefits from one to another in a collaborative or competitive manner. As a matter of fact, it shows that an organization emerges when exists some kind of interac- tion between its participants (either through direct communication by means of speech acts or through the environment). Most researchers agree on that MASs are a special kind of distributed systems (objects with their own threads of execution) with special features where a higher degree of complexity exists than in current Object–Based Software Systems [7]. This complexity of MASs is consequence of their features and mainly of their interacting nature: Complexity is caused by the collective behaviour of many basic interacting agents. James Odell [7]. ⋆ The work reported in this article was partially supported by the Spanish Ministry of Science and Technology under grants TIC97-0593-C05-05