The "what" problem: the emergence of new goals in a robot

Biological and cognitive systems have the capability of developing new goals during phylogenesis of species or during ontogenesis of single individuals. On the other hand, current artificial cognitive systems focus on how achieving a given fixed set of hard-wired goals. They search an optimal solution of a problem, given a set of goals and a set of optimization criteria. They look for "how" to achieve a given goal. Natural agents develop new goals in order to cope with partially unknown and ever changing environment. They must find "what" they want to achieve and not only "how". The development of new goals on the basis of the interaction with the environment is here defined the "what" problem. The development of a collection of goals permits to redefine the concept of Umwelt in what could be considered the ideological Umwelt of an agent. The objective of this paper is twofold: i) to outline the "what" problem and ii) to describe a robotic architecture capable of addressing it.