[Paper Review] Do Facial Expressions Predict Ad Sharing? A Large-Scale Observational Study

Daniel McDuff, Jonah Berger|arXiv (Cornell University)|Dec 21, 2019

Consumer Behavior in Brand Consumption and Identification43 references4 citations

TL;DR

This large-scale observational study investigates whether facial expressions during ad exposure predict online sharing behavior. Using automated facial coding on thousands of viewers, it finds that while smiles increase sharing, certain negative expressions—like nose wrinkles (disgust)—also boost sharing, challenging the idea that only positive emotions drive virality.

ABSTRACT

People often share news and information with their social connections, but why do some advertisements get shared more than others? A large-scale test examines whether facial responses predict sharing. Facial expressions play a key role in emotional expression. Using scalable automated facial coding algorithms, we quantify the facial expressions of thousands of individuals in response to hundreds of advertisements. Results suggest that not all emotions expressed during viewing increase sharing, and that the relationship between emotion and transmission is more complex than mere valence alone. Facial actions linked to positive emotions (i.e., smiles) were associated with increased sharing. But while some actions associated with negative emotion (e.g., lip depressor, associated with sadness) were linked to decreased sharing, others (i.e., nose wrinkles, associated with disgust) were linked to increased sharing. The ability to quickly collect facial responses at scale in peoples' natural environment has important implications for marketers and opens up a range of avenues for further research.

Motivation & Objective

• To investigate whether facial expressions during ad viewing predict subsequent online sharing behavior.
• To test whether emotional valence alone determines sharing, or if specific facial actions matter independently.
• To explore the role of distinct facial action units (e.g., smiles, lip depressors, nose wrinkles) in predicting virality.
• To evaluate the scalability and real-world applicability of automated facial coding in natural viewing environments.
• To challenge the assumption that only positive emotions drive information sharing in social media contexts.

Proposed method

• Employed scalable automated facial coding algorithms to analyze facial expressions in real time from thousands of participants.
• Collected facial response data during exposure to hundreds of advertisements in natural viewing environments.
• Mapped specific facial action units (e.g., AU6 for smiles, AU15 for lip depressor, AU9 for nose wrinkles) to emotional states.
• Used multilevel regression models to predict ad sharing behavior based on facial expression patterns.
• Controlled for demographic and contextual variables to isolate the effect of facial responses on sharing.
• Validated findings using a large, diverse dataset collected across multiple viewing sessions and ad types.

Experimental results

Research questions

• RQ1 Do facial expressions during ad exposure predict whether users will share the ad online?
• RQ2 Is the relationship between emotion and sharing driven solely by emotional valence (positive vs. negative)?
• RQ3 Which specific facial action units are most predictive of increased ad sharing?
• RQ4 Do certain negative facial expressions, such as disgust, paradoxically increase sharing?
• RQ5 Can automated facial coding in natural settings reliably predict social transmission of digital content?

Key findings

• Facial expressions significantly predict ad sharing behavior, even after controlling for demographic and contextual factors.
• Smiles (facial action unit AU6) were consistently associated with increased sharing, supporting the role of positive affect.
• Lip depressor (AU15), associated with sadness, was linked to decreased sharing, aligning with valence-based expectations.
• Nose wrinkles (AU9), associated with disgust, were unexpectedly linked to increased sharing, challenging the valence-only hypothesis.
• The relationship between emotion and sharing is more nuanced than simple positive/negative valence, with specific action units having divergent effects.
• Automated facial coding in natural viewing environments provides a scalable and reliable method for predicting social sharing behavior.