Information overloading is one of the major problems of the
Information Society, and it is experienced by many people. Information
retrieval is aimed at solving such problem, and hence it is a crucial
discipline of this new era. Despite its centrality, information
retrieval has its own shortcomings: for instance, most of Internet
users have discovered with excitement the information retrieval
systems available on Internet (the so called 'search engines'), but
they have also experienced how often the performance of such services
is too low, very far from an ideal 100%. The lackness of a formal
account is probably one of the most evident of these shortcomings:
concepts like information, information need and relevance are neither
well understood nor formally defined. This paper sketches a cognitive
framework that permits to analyze these central concepts of the
information retrieval scenario.

The cognitive framework consists of concepts as cognitive agents
acting in the world, knowledge states possessed by the cognitive
agents, transitions among knowledge states, and inferences. On the
basis of such framework, information is formally defined as an ordered
pair representing the difference between two knowledge states; this
definition permits to clarify the distinction among data, knowledge
and information and to discuss the issue of the subjectiveness of
information. On this ground, the concept of information need is
examined: it is defined, it is studied in the context of the
interaction between an information retrieval system and a user, and
the well known classification in verificative, conscious topical and
muddled needs is analyzed. On the basis of the above definitions of
information and information need, relevance is formally defined, and
some critical features of this concept are discussed.