Integrating Frameworks in Human-Computer Interaction: A Meta-Analytic Comparison of Human and Virtual Influencers With Varying Levels of Anthropomorphism

Virtual influencers (VIs) are computer-generated characters that leverage artificial intelligence (AI) to imitate, interact with, and influence humans. While researchers extrapolated frameworks in human-computer interaction to examine this novel phenomenon, this nascent scholarship is undermined by its fragmented theoretical perspective. Past studies also yielded conflicting results when comparing the persuasiveness of VIs to human influencers (HIs). A meta-analysis of 75 experimental studies (N = 27,711) revealed that VIs were generally less persuasive than HIs, particularly in shaping user perceptions and attitudes. VIs were also less persuasive than HIs when they had a low follower count or when their follower count was not displayed. Considering the diversity of VIs with varying levels of anthropomorphism, another meta-analysis conducted using 29 experimental studies (N = 6,497) revealed that human-like VIs were generally more persuasive than anime-like VIs, specifically in shaping user perceptions and promoting behavioral intentions. Theoretical contributions and practical implications are discussed.