This thesis addresses two paradoxes: (1) why empirical studies find that fake news represent only a small share of the information consulted and shared on social media despite the absence of editorial control or journalistic norms, and (2) how political polarization has intensified even though users do not appear especially receptive to fake news. To investigate these issues, two complementary studies were carried out on Twitter and Facebook, combining quantitative analyses of digital traces wit...