Quantifying the disruptiveness of a paper by analyzing how it overshadows its successors

The disruption index (DI) proposed by Funk and Owen-Smith (2017) is a practical metric that has been widely used to identify and analyze disruptive research. However, it suffers from several limitations, such as susceptibility to authors’ manipulation, a narrow focus on the local citation network, and unreasonable convergence characteristics. To address these shortcomings, we propose a novel overshadowing disruption index (∆DI), based on the DI, that captures the disruptive quality of a focal paper by examining its overshadowing impact on its successors. Using 359 highly cited, 443 moderately cited, and 40 Nobel Prize-winning physics papers as research objects, we analyze the evolutionary trajectories of ∆DI and demonstrate its rationality via the statistical methods and GPT-4. Specifically, ∆DI presents a decay trend converging to zero, indicating that the disruptive impact of a paper declines over time. By analyzing papers’ research content via GPT-4, we further explain the decay trend from the perspective of semantic analysis. Additionally, we comprehensively examine ∆DI’s statistics and unveil its correlation with common DI-based metrics. Finally, we systematically verify the effectiveness of ∆DI by scrutinizing the relationship between ∆DI and future scientific impact. Our results show that ∆DI exhibits better predictive power than DI and DI₅, and the combination of ΔDI and DI performs the best in predicting scientific impact.