Exploring the Impact of AI-Generated Images on Political News Perception and Understanding

Abstract

In political news articles, images play an important role in conveying news content and attracting readers' attention. As image generation technology has been developed and its use has increased, this study investigated the effect of image generation of political news articles on readers' perception and response to the articles. We first examined the primary elements that characterize political news articles and used the latest text-to-image model with these elements to generate images appropriate for liberal/conservative media articles. The results of the user study with 102 participants showed that the generated images reflected the content of the articles better than the original images and facilitated the understanding of the articles. In particular, conservative participants preferred generated images in factual reports, and liberal participants preferred generated images in biased reports, suggesting a careful consideration of the use of generated images in information delivery. Our study opens up new possibilities for the use of AI in journalism, where providing fair and clear information to readers and reducing political polarization is important.