Too Good to be False: How Photorealism Promotes Susceptibility to Misinformation

초록

Recent advances in generative AI enable the production of photorealistic images that blur the line between authentic and synthetic content, raising concerns about misleading users when misused. We examined how the realism of AI-generated images affects users’ credibility judgments of misinformation and explored two key cognitive mechanisms: realism heuristic and synthetic heuristic. A between-subjects experiment (N = 253) revealed that highly realistic AI-generated images significantly increased the credibility of misinformation compared to less realistic images, and this effect was primarily driven by the realism heuristic, as photorealistic images were perceived as portraying reality. Additionally, photorealistic images tended to inhibit activation of the synthetic heuristic, which we newly proposed in this paper (i.e., the perception that images are artificially created or edited). Contrary to previous findings, our data show that regular users of generative AI and social media are more susceptible to misinformation. We propose identifying and labeling AI involvement in image creation and increasing literacy. © 2025 Copyright held by the owner/author(s).

키워드