Purpose – A key task in many Collective Intelligence systems is to represent people and resources in some computational form that is general enough to accommodate different needs and diverse sources of information. Also, it should take into account the inevitable imprecision deriving from the necessity of representing complex phenomena in feasible ways. The purpose of our work is to define an abstract model for representing both users and resources with the main features of generality, adaptivity and capability of handling imprecise information. This model is called Fuzzy Information Filter (FIF). Design/methodology/approach – A FIF is based on the homogeneous description of users and resources in terms of fuzzy metadata, i.e. attributes whose values are defined in terms of fuzzy sets. Metadata define elementary filters, which can be combined in order to define composite filters that could better represent complex profiles. Furthermore, both elementary and composite filters can be adapted to incoming data by means of learning rules based on Possibility Theory, eventually relaxed to comply with specific application needs. Originality/value – The adoption of metadata to represent resources in an homogeneous way makes possible the use of the proposed model within semantic web applications. The use of fuzzy sets to define metadata enables imprecise assignment of values to attributes coming from subjective judgments, perception-based knowledge, etc. When a resource is given, the application of a FIF assesses how much the resource is compatible with the information stored in the filter. The resulting degree can be used to rank resources, e.g. with respect to the interests of a user. This approach could be exploited to attenuate the problem of information overload in large repositories of resources. Furthermore, the technology of fuzzy sets makes FIF models intelligible to analysts and inclined to integration with expert knowledge. Finally, the possibility of automatically adapting filters using incoming data enables the realization of personalized and adaptive systems. Practical implications – FIF models can be applied to represent user profiles in personalized applications where the use of fuzzy sets is a competitive advantage, as in personalized e-learning systems, recommendation systems, etc. More advanced applications could profit of the peculiar representation of profiles in terms of complex filters to produce, for example, clusters of people with a similar profile. On the overall, FIF introduces flexibility and adaptivity in web-based applications that use social networks and resources to expose intelligent behaviors.