Key points are not available for this paper at this time.
The rapid development of artificial intelligence (AI) allows the creation of photorealistic images with simple text prompts. As a result, proliferation of AI-generated visual misinformation is an increasing concern. However, how AI-generated visual misinformation influences our beliefs and whether corrections are effective remains to be seen. In study 1, we examined discernment between real headlines and headlines accompanied by AI-generated images (AI headlines), and how cognitive reflection, conspiracist beliefs and digital literacy correlated with belief in AI headlines. In study 2, after initial belief ratings of headlines, corrective information was provided, during which participants rated their surprise to a headline’s correction or confirmation. Beliefs were recorded immediately afterwards and after a 1-week delay. Subjective ratings of image realism, evidence and surprise were collected as predictors for initial belief and belief regression. Results showed that more realistic images and images that provided stronger evidence for the headline led to greater belief in both real and AI headlines, and that higher cognitive reflection ability and lower conspiracy beliefs were linked to improved discernment. Belief regression did not differ between real and AI headlines, and was primarily determined by how surprising the correction or confirmation was. Our findings suggest that people rely on image qualities such as its realism and how much evidence it provides to make judgements about headline veracity. Corrections were effective at reducing belief in AI headlines, thus emphasizing the importance of identifying and labeling AI content.
Guo et al. (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: