Generative Artificial Intelligence for Tobacco Health Promotion
Mené sur 614 participants (âge moyen : 20,5 ans), cet essai évalue, par rapport aux publicités existantes émanant des agences de santé officielles, l'efficacité de publicités de sensibilisation aux dangers du vapotage générées par des jeunes à l'aide de l’intelligence artificielle
Leung and colleagues examined the perceived message effectiveness (PME) of artificial intelligence (AI)–generated, youth-codesigned vaping awareness social media ads compared with existing ads from official health agencies. The authors also examined how different source labeling (eg, “made with AI” vs “made by the World Health Organization [WHO]” vs “made with AI by the World Health Organization [WHO]”) was associated with PME. The study addresses an important gap in understanding the effectiveness and perception of generative AI (GenAI)–created, youth-codesigned vaping prevention ads, a relatively unexplored area in research. In a prestudy phase, the GenAI-generated ads were codesigned with youths. This procedure included AI generation of vaping cessation ads, followed by 9 focus groups and interviews to refine the ads and select a final set of 25 ads. Next, antivaping campaign ads from existing health agencies were selected using those from the WHO and the US Food and Drug Administration. Following the prestudy phase, an analytic sample comprising 614 youths aged 16 to 25 years completed the study. The authors used a mixed factorial experimental design. Participants were randomized to 1 of 4 source labeling conditions: control condition, AI condition, health agency condition, and combined condition. Then each participant viewed 50 ad content sources: 25 AI-generated vs 25 existing health agency. The ad source was the within-subjects factor, and the source labeling was the between-subjects factor. The outcome of interest was the PME Scale for Youth, which comprised 5 subitems. Of the 5 items, 2 examined effect perceptions as they relate to thoughts about vaping (eg, very bad idea or very good idea) and intent to vape (eg, strongly discourages me or strongly encourages me). Then 3 items assessed ad perceptions: attention-grabbing, informativeness, and convincingness.
JAMA Network Open , éditorial en libre accès, 2025