We discuss problems of current centralized Information Retrieval (IR)-approaches in view of a dynamic change in the culture of the Web towards a mobile, social Web of Participation where the paradigms of decentralization and autonomy play an increasingly important role. From the discussion of human principles of IR, existing approaches from Geo-IR, P2P-IR and the increased importance of spatio-temporal referencing of information, we justify the proposition of a distributed, Multi-Agent IR architecture with spatio-temporal reference as the primary classification criterion as an alternative solution to the future requirements in IR. We also discuss implicitly social and semantic spatio-temporal small world structures which play an important role in this concept. In a second part we discuss the elements of our architecture and their relations to the justifying theoretical considerations of the first part in more detail and present the processes in our approach as well as a qualitative evaluation with the help of a prototypical implementation of key aspects.

Schlüsselwörter: Information Retrieval, Spatio-Temporal Information Retrieval, Distributed Information Retrieval, Social Information Retrieval, Multi-Agent-System, Distributed Spatio-Temporal Index, Peer-to-Peer-Protocol, Small World, Toblers First Law