The Myth in the Methodology: Towards a Recontextualization of Fairness in Machine Learning