Measures of information are a key concept and in the heart in Artificial Intelligence. The well-known measure of information, the Shannon entropy, is commonly used in several domains to value the information associated with a set of probabilistic events. By extension, it is also used to provide information on a distribution of values. This aim of measuring information has been extended to the fuzzy sets theory and to its extension. Indeed, several entropies of interval-valued fuzzy sets (IVFS), or, equivalently, entropies of Atanassov intuitionistic fuzzy sets, have been proposed. However, all these entropies takes only into account the form of the set rather than a probabilistic information that could be associated to it. In this paper, we state the requirements that a probabilistic entropy of IVFS should fulfilled.