Dual roles of higher-order interactions in temporal cooperation dynamics

Higher-order interactions are ubiquitous in real-world collaborative systems, yet existing studies often consider temporal dynamics and higher-order interactions separately, limiting the understanding of cooperation in realistic group environments. To bridge this gap, we develop an empirical evolutionary game framework based on temporal higher-order networks using two real-world face-to-face interaction datasets: Office 2013 and Office 2015. Specifically, we show that higher-order interactions exhibit a condition-dependent dual role: under a weak dilemma ( δ 1 = 0.1 ), cooperation frequency f C reaches 0.73 compared with 0.16 in pairwise networks, whereas under a strong dilemma ( δ 1 = 0.4 ), cooperation collapses to 0 compared with 0.12 in pairwise networks. Randomization experiments further reveal that the association between temporal burstiness and cooperation differs between pairwise and higher-order interaction systems. Finally, we extend the heterogeneous activity-driven model to derive an analytical defection threshold for temporal higher-order networks, providing a theoretical framework for understanding cooperation collapse in evolving group interactions.