Over the last few years, driven by the growth of open data and e-governments, there has been an increasing demand for transparency as well as accountability. These two concepts require particular information to be easily accessible, most common for datasets being the creator, the license, the modification dates, etc. According to us, the expression of such specific information need is central to the definition of these concepts. This leads to consider their assessment as a particular case of completeness, one aspect of data quality. In order to make the approach applicable to any case requiring information, we first formalize the notion of information need based on a hierarchical structure. Then we define completeness of a knowledge graph with respect to an information need, in such a way it takes into account preferences of the users which can vary significantly. The proposal is used to define a first measure of accountability of a knowledge graph, which enables to evaluate several RDF graphs of the linked open data. It appears that most of them do not provide accessible metadata about themselves. The others get various accountability results, including good ones, showing that linked data can be very suitable for accountability and more broadly transparency.