Abstract
The rationalist conception of public action, which aims to measure the performance of education systems and relies on facts and evidence to do so, is gaining ground. This concept is intended to legitimize public action, yet skepticism towards educational policies and authorities, and the experts who advise them, is now evident. To shed light on this paradoxical double observation, we will defend the thesis that what is measured is not necessarily what matters from the point of view of the beneficiaries of public action. Based on an examination of the policy implemented in French-speaking Belgium to regulate the use of repetition, we will show that the quantitative evidence (such as that provided by meta-analyses) on which this policy bases its cognitive legitimacy has in fact contributed to its ineffectiveness and discredit in the field. This is due to a failure to pay sufficient attention to and take into account the practical, situated realities of school actors - those that matter most to them. To avoid the " revenge of contexts ", the rationalist conception of public action should therefore be complemented by a comprehensive perspective that anchors factuality in teachers' experience and judgment and, in so doing, recognizes and draws on contextual expertise and professional knowledge.