The Ethics of Seeing: Contextual Moderation of Altered Image Acceptability

Digital communication environments increasingly normalise both photo-editing and generative AI, intensifying concerns about whether images continue to function as credible records of reality. This study investigates how audiences evaluate unedited, Photoshopped, and AI-generated images in social media, news, and advertising contexts, and tests whether ethical judgments about truthfulness account for variation in acceptance of altered images. In a mixed factorial online experiment, 709 adults recruited via Prolific were randomly assigned to one of three alteration conditions in which selected images were unedited but were described to be as either unedited, Photoshopped, or AI-generated. Participants viewed three images (one per context) and rated acceptance of use, perceived ethical violation, perceived authenticity, and perceived deviation from truth on multi-item Likert scales. Results show that image-editing in the news context was consistently evaluated as least acceptable and most ethically problematic, with AI-labelled images eliciting the highest ethical concerns. Across context conditions, alteration labels reduced acceptance mainly by increasing perceived deviation from truth, which heightened ethical violation, whereas perceived authenticity did not explain this pathway. Overall, the findings suggest that acceptance of altered images primarily reflects a moral appraisal of truthfulness and underscore the need for appropriate governance and legislation that ensures perceived fairness, transparency, and context-appropriate use of AI and edited imagery.