For brands on social media, guessing which post will spark a like, comment or share can feel like trial and error. A new artificial intelligence model aims to replace that guesswork with sharper prediction — and clearer insight into why people engage.
Tengteng Ma
The study, co-authored by Tengteng Ma, assistant professor of information systems at the University of South Florida, introduces a graph neural network with an attention mechanism for customer engagement prediction, known as GACE. The research appears in Information Systems Research.
Instead of relying on manually selected features, the model maps the web of relationships among users, brands and posts. Using Facebook data that included more than 660,000 engagement actions from over 91,000 users across 47 brands, the researchers found that GACE outperformed leading prediction methods. It improved accuracy by more than 7% compared with the strongest benchmark and delivered gains across precision, recall and overall predictive power.
A key innovation is its use of theory-driven “metapaths” that capture different drivers of engagement, such as similarity between users, brand loyalty and affinity across related brands. An attention mechanism then learns which factors matter most for each user, offering interpretable insights — not just predictions.
The results suggest practical value: better targeting can lower marketing costs, boost engagement and reduce the risk of irritating uninterested users. By learning how users, brands and content connect, AI can predict engagement more accurately — and help firms spend smarter on social media.
Authors: Tengteng Ma, University of South Florida; Yuheng Hu, Yingda Lu, and Siddhartha Bhattacharyya, University of Illinois at Chicago.