Social User Agents for Dynamic Access to Wireless Networks

We present a dynamic and adaptive decision model for an autonomous user agent whose task is to dynamically negotiate and procure wireless access for a mobile user. A user is assumed to have cognitive and motivational costs associated to providing subjective preference in- formation to the agent. Therefore the task of the per- sonal agent is to dynamically model the user, update its knowledge of a market of wireless service providers and select service providers that satisfies the user's expected preferences based on minimal, or missing, information that is derived from a simple user interface. In this paper we show how this user modeling problem can be repre- sented as a Markov Decision Process. Adaptive rein- forcement learning solutions are then evaluated for two subclasses of tractable MDPs via simulations of some representative user models.