AI-driven peer recommendation systems can enhance creativity in social networks

Can AI-driven peer recommendation engines elevate people’s creative performances in self-organizing social networks? Addressing this question requires overcoming challenges in (i) data collection (e.g., explicitly tracing how inspiration flows among individuals) and (ii) intervention design (e.g., developing approaches that stimulate creative ideas without amplifying redundancy). We trained a machine learning model that predicts people’s ideation performance from semantic and network-structural features, and integrated it into SocialMuse, an AI-driven peer recommendation system that maximizes predicted creative outcomes when suggesting connections. In controlled online experiments (N = 420), networks using SocialMuse (i) outperformed AI-agnostic controls across multiple creativity measures and (ii) exhibited greater decentralization, spreading inspiration sources and potentially helping ideas stand out. While the intervention was tested in controlled laboratory networks, our study offers actionable insights and proof-of-concept design implications for intelligent peer recommendation engines that foster creativity in online social platforms.