Preventing profiling for ethical fake news detection

Liesbeth Allein, Marie-Francine Moens, Domenico Perrotta

Abstract: Highlights•Ethical fake news detection in terms of user data integration.•Profiling-based decision-making is prevented while leveraging social knowledge.•Social science-based losses integrate social context during model optimisation.•The losses assume identity correlations between news texts and Twitter spreaders.•Context-inspired classifiers learn statistically better discriminating latent spaces.