Pricing Information Goods in Distributed Agent-Based Information Filtering

摘要

Most approaches to information filtering taken so far have the underlying hypothesis of potentially delivering notifications from every information producer to subscribers; this exact information filtering model creates efficiency and scalability bottlenecks and incurs a cognitive overload to the user. In this work we put forward a distributed agent-based information filtering approach that avoids information overload and scalability bottlenecks by relying on approximate information filtering. In approximate information filtering, the user subscribes to and monitors only carefully selected data sources, to receive interesting events from these sources only. In this way, system scalability is enhanced by trading recall for lower message traffic, information overload is avoided, and information producers are free to specialise, build their subscriber base and charge for the delivered content. We define the specifics of such an agent-based architecture for approximate information filtering, and introduce a novel agent selection mechanism based on the combination of resource selection, predicted publishing behaviour, and information cost to improve publisher selection. To the best of our knowledge, this is the first approach to model the cost of information in a filtering setting, and study its effect on retrieval efficiency and effectiveness.