This work presents an approach in which user profiles,represented by ontologies that were learned by an interface agent,are compared to foster collaboration for information retrieval from the web. It is shown how the interface agent represents the knowledge about the user along with the profiles that were empirically developed. Departing from a specific matching model,briefly presented here,quan titative results were achieved by comparing such particular ontologies in a fully automatic way. The results are presented and their implications are discussed. We argue that an agent’s knowledge can be enriched by other agents navigation experiences,p ossibly helping in connecting and reconciliating distinct knowledge views while preserving a degree of privacy.