@article{quadras-etal-2026-assessing,
    title = "Assessing the Persuasive Effect of {AI}-Generated Image Support of Arguments",
    author = "Quadras, Mackwyn  and
      Stede, Manfred  and
      Wachsmuth, Henning",
    editor = "Piperidis, Stelios  and
      Bel, N{\'u}ria  and
      van den Heuvel, Henk  and
      Ide, Nancy  and
      Krek, Simon  and
      Toral, Antonio",
    journal = "International Conference on Language Resources and Evaluation",
    volume = "main",
    month = may,
    year = "2026",
    address = "Palma de Mallorca, Spain",
    publisher = "ELRA Language Resource Association",
    url = "https://preview.aclanthology.org/ingest-lrec/2026.lrec-main.641/",
    pages = "8085--8095",
    abstract = "Argumentation is, at its core, an inherently verbal activity. Yet, other modalities may support arguments, one of which are images. In the argument mining community, this combination has not received much attention yet. While a few previous works studied whether images can make argumentative texts more effective in persuading people, the images that were considered matched the texts loosely only, or they were heavily text-based themselves. In this paper, we take the step to study to what extent the persuasive effect of textual arguments can be supported by images specifically created for this purpose. For a consistent experiment design, we combine NLP with image generation to synthesize both arguments and images with generative AI, for five controversial topics and for two rhetorical strategies. In two consecutive user studies, we first determine the best-matching image for each argument and then compare the perceived effect of bare textual arguments to those that are supported by an image. Our results suggest that the images may increase the persuasive effect of argumentative texts, but with variance across topics."
}