earticle

영어

This study investigates the potential of generative AI tools to create pictorial health warnings (PHWs) for cigarette packaging. Traditionally, PHWs have utilized real photographs to depict the health risks associated with smoking—such as lung cancer—in order to evoke fear and discourage tobacco use among the general public. This research explores whether generative AI can serve as a viable alternative or supplement to traditional methods in producing images that elicit fear and disgust based on factual information. Using DALL·E 3, the image generation tool integrated with ChatGPT-4o, a set of experimental visual mock-ups was produced based on four types of image generation prompts. These prototypes were analyzed in terms of visual design elements— including color, tone, and brightness—as well as their effectiveness in eliciting fear. To deepen the analysis, a focused group interview (FGI) was conducted with design experts for this analysis. Furthermore, the study examined the fear-evoking effects of each image type based on the prompt and identified the visual elements responsible for eliciting fear responses. Images were generated based on four prompt types, each aimed at producing fearoriented visual content: Emotion-Appealed Photograph Images, Fact-Appealed Photobraph Images, Emotion- Appealed Illustration Images, Fact-Appealed Illustration Images. Each image type exhibited distinct visual characteristics. Among the color elements, red was used most frequently (16 times), followed by black and yellow (9 times each). Regarding tone, “deep and dark” was most common (5 times), while “Clear and strong contrast” and “bright and clean” each appeared 4 times. In terms of brightness, “dark tone” was the most prevalent, appearing in 7 images. In the survey assessing which prompts were most effective in eliciting fear, the Emotion-Appealed Photograph Images(EAPI) type was found to be most effective for laryngeal cancer, lung cancer, and heart disease. For oral cancer, both EAPI and FAPI were similarly effective. When asked which visual elements influenced fear perception the most, participants identified "visual style"—referring to the type of image expression—as the most influential factor across all four diseases. With more nuanced and strategically crafted prompts, AI-generated images could potentially intensify emotional responses. This limitation highlights the need for further refinement in prompt design to ensure comprehensive coverage of various health conditions relevant to tobacco use.