The detachment of responsibility: how and when AI-powered digital avatars undermine behavioral intentions in cancer screening persuasion

As artificial intelligence technologies advance, humanlike digital avatars are increasingly used in health communication. Yet, whether—and how—avatars embodying real humans influence the persuasive impact of health messages remains largely unexplored. Drawing on communicative responsibility theory, this study proposes that, compared with real human communicators, audiences perceive digital avatars as less capable of assuming communicative responsibility, thereby weakening their persuasive effectiveness. Furthermore, this negative effect is expected to be stronger when the avatar represents a high-expertise identity (i.e., a doctor) rather than a low-expertise identity (i.e., a blogger). Using pancreatic cancer screening as the persuasion context, we conducted a 2 (communicator type: digital avatar vs. human) × 2 (source identity: high-expertise vs. low-expertise) between-subjects experiment with 211 Chinese participants. The findings revealed that (1) messages delivered by the digital avatar led to lower behavioral intentions than those delivered by its human originator; (2) perceived communicative responsibility negatively mediated this relationship; and (3) the negative effect of a digital avatar on perceived communicative responsibility was stronger when it represented a high-expertise identity rather than a low-expertise identity, which in turn led to lower behavioral intentions. These results advance theoretical understanding of avatar-mediated persuasion and offer practical guidance for the design of AI-based health communication.