Real Talk, Virtual Faces: A Formal Concept Analysis of Personality and Sentiment in Influencer Audiences
New Formal Concept Analysis shows virtual vs. human influencer audiences communicate with fundamentally different 'grammar'.
A research team from NYU Abu Dhabi, led by Shahram Chaudhry, Sidahmed Benabderrahmane, and Talal Rahwan, has published a groundbreaking study applying Formal Concept Analysis (FCA) to understand how audiences communicate with virtual versus human influencers. Their paper, 'Real Talk, Virtual Faces,' introduces a novel two-layer analytical framework. The first layer uses FCA with support-based iceberg filtering on weekly-aggregated YouTube comments to extract 'discourse profiles'—bundles of co-occurring sentiment, Big Five personality cues, and topic tags. The second layer applies association rule mining at the comment level to uncover deeper dependencies between these signals that frequency-based analyses miss.
Applied to three matched pairs of virtual and human influencers, the analysis revealed a stark structural divergence. Human influencer discourse consistently coalesced into a single, emotionally regulated regime characterized by low neuroticism anchoring positive sentiment. In contrast, virtual influencer discourse supported three structurally distinct discourse modes. One notable cluster focused on appearance—a topic absent from the human influencer discourse structure despite having similar overall prevalence. Furthermore, topic-specific analysis showed virtual influencer contexts elicited more negative sentiment in psychologically sensitive domains like mental health, body image, and discussions of artificial identity.
The study positions FCA as a powerful, principled tool for multi-signal discourse analysis, moving beyond simple 'what' is said to understand the 'grammar' of how different psychological and topical signals co-occur. The findings demonstrate that virtuality doesn't just change the content of audience reactions but fundamentally reshapes the underlying structure of online discourse, revealing more fragmented and complex audience engagement patterns around synthetic personas.
- Human influencer audiences communicate in a single, stable emotional regime (low neuroticism + positivity), while virtual influencer audiences use three distinct discourse modes.
- A unique 'appearance-discourse' cluster emerged for virtual influencers, structurally absent from human influencer comments despite equal topic prevalence.
- Virtual influencer contexts showed elevated negative sentiment in sensitive domains like mental health and artificial identity, per topic-specific analysis.
Why It Matters
This provides brands and platforms with a scientific framework to measure the qualitatively different—and often riskier—audience engagement generated by virtual influencers.