A model for retrieving non-solicited information is presented. The model is based on the guided exploration of information spaces. It works by monitoring user actions and formulating a hypothesis of her information needs. This hypothesis is used to locate documents similar to those which the information seeker has found to be interesting. The manipulation of these documents by the user is then used to update the hypothesis of information needs and retrieve a new cluster of documents that the system estimates as being relevant to the user’s information needs. We present a general model for guided information exploration, discuss the functional requirements of the main components of the model, and propose an architecture for implementing it.