Abstract: Scientific claims and informal discussions about science-related issues have proliferated across online platforms, where they are increasingly mobilized within manipulative narratives and disinformation campaigns. How, why, and when do people invoke science in online debates; what intended or unintended effects do these practices produce? In this talk, I consider these questions through the lens of how scientific discourse is used—or instrumentalized—to legitimize misleading information and to insert potentially harmful narratives in social media. I then explore how artificial intelligence, and in particular transformer-based language models, can contribute to identifying and countering such dynamics, rather than amplifying them. I conclude by turning to the role of emotions and affective language in scientific disinformation, and by discussing the challenges that emotionally charged discourse poses for current state-of-the-art models.

Short BIO: Konstantin Todorov is associate professor at the University of Montpellier (maître de conférences HDR, hors classe) and a researcher at the LIRMM laboratory within the ADVANSE team. See his website for more informations.
Website: https://www.lirmm.fr/konstantin-todorov/

Un lien de connexion Zoom sera affiché sur la ( page des séminaires LFI ) le jour du séminaire.

Ce séminaire est organisé conjointement avec le Chapitre Français de l’IEEE Computational Intelligence Society.