The design of a computational model for information retrieval that provides coordination services for collaborative agents is described. Requirements resulting from recent technological developments are identified. Coordination problems in terms of information retrieval are defined. Limitations of previous models based on fuzzy sets are discussed to justify the need for a new approach. The results of psychological experiments are applied to making design decisions so that the proposed system simulates human behavior. The design framework comprising fuzzy proximity networks and typicality measurement algorithms, is described in detail. It is shown how the system takes users' personal profiles into consideration when processing their queries. Learning algorithms to update the networks and to construct a keyword hierarchy automatically are suggested.