1 Agents acquiring Resources through Social Positions: An Activity-based Approach George Lekeas and Kostas Stathis Intelligent Computing Environments Department of Computing, City University Northampton Square, LONDON EC1V 0HB. {g.k.lekeas, k.stathis}@city.ac.uk 1. Introduction The main assumption of this paper is that we will need to use artificial agent societies [1, 11] in order to manage intelligently the resources and the interactions envisaged in socio- cognitive grids. We are motivated by scenarios where location-independent smart-services provide the social and cognitive resources that agents will need to acquire in order to satisfy their users' needs. More concretely, we are investigating the situation where an agent who is trying to achieve a goal in a specific environment realises that it needs to join an artificial agent society in that environment (or possibly in another environment), which possesses the resources necessary for the agent to achieve the goal [14]. For example, if you travel to a foreign place for business purposes, you would expect your agent to join the artificial agent society of a train station, so that you can receive advice on how to buy a train ticket [5]. We start from the work described in [14] but we extend it one step further as follows. For an agent to achieve a goal it needs to enter a society with a specified social position [12]. By social position, we mean assigning to an instance of an agent a non-empty set of roles. By role we mean an expected set of behaviours that "tell" the agent how it should behave in a specific society whose organisation relies on that role; for example if an agent has a social position that contains the role receptionist of a hotel, then that role tells the agent that it should not steal credit card numbers. Following the society classification presented in [4], we are interested in semi-open societies, where membership is by application only. In this context, we are investigating in detail the activities between (a) agents, which apply to artificial societies for membership and (b) the societies, which will respond to the agent’s application for the requested social position. For this purpose, we take an activity-based approach and we propose a set of UML activity diagrams to model the required interactions. Although it has been argued [6] that UML is not sufficient for modelling agent activities (because of agents' richer properties [9] and the fact that UML does not have clear semantics [10]), we use UML as in [9] to visualise interaction processes between agents, by representing concurrent and asynchronous processing and expressing simultaneous communications with several correspondents. It is also important to add that here we are particularly concerned with giving a top-level picture of the entry of an agent to a society, the life-cycle of the agent while a member of that society, and the exit of the agent from that society, rather than investigating the meaning of the interactions that take place in the