This is a contributed discussion on the article ‘Causal inference by using invariant prediction’ by Peters et al.