Workshop Proceedings of the 16th International AAAI Conference on Web and Social Media
This work advances investigations into the visual media shared by agents in disinformation campaigns by characterizing the images shared by accounts identified by Twitter as being part of such campaigns. Using images shared by US politicians' Twitter accounts as a baseline and training set, we build models for inferring the ideological presentation of accounts using the images they share. Results show that, while our models recover the expected bimodal ideological distribution of US politicians, we find that, on average, four separate influence campaigns - attributed to Iran, Russia, China, and Venezuela - all present conservative ideological presentations in the images they share. Given that prior work has shown Twitter accounts used by Russian disinformation agents are ideologically diverse in the text and news they share, these image-oriented findings provide new insights into potential axes of coordination and suggest these accounts may not present consistent ideological positions across modalities.