Doctoral research by Davide Livrieri

Model: Generative image AI + Intervention

Structured visual prompting ➔ Theme: Multimodal gendered disinformation

Visual and multimodal disinformation is a growing part of online information disorder: images and video reinforce false or misleading content and shape both credibility and emotional response. This case study examines gendered multimodal disinformation — content combining text with AI-generated images to convey stereotypical, sexist portrayals of women in politics.

Experimental stimuli, designed to resemble real social posts, test how these representations affect perceived credibility and attitudes toward the person targeted. A later phase will develop visual and media-literacy tools to help audiences recognise manipulated imagery.

"Even a tightly structured prompt doesn't guarantee the model generates exactly the stereotype the researcher intended to test — the image can carry unintended signals of its own."

Davide Livrieri, Uninettuno 

Supervisors: Alessandro Pollini (Uninettuno), Francesca D'Errico (University of Bari), Marinella Paciello (University of Rome - La Sapienza)