Abstract: [Purpose/Significance] Revealing the influence of various user attribute characteristics on the diffusion effects of scientific papers on social media can aid the formulation and optimization of scientific communication strategies, thereby facilitating the effective diffusion of scientific papers. [Design/Methodology] From the perspective of computational communication research, we collected a large-scale dataset of scientific paper diffusion on Twitter（now known as X）, comprising 52,219 papers and 148,935 users involved in the diffusion. By employing complex network analysis methods, we comparatively analyzed the differences in diffusion effectiveness and the structural characteristics of diffusion networks among users with different attributes from three aspects: social platform influence, preferences, and group diversity. [Findings/Conclusion] The influence of user attributes on the diffusion effects of scientific papers does not operate directly; instead, it is mediated by the cascades that users initiate. Specifically, papers shared by highly influential users and by members within interest-based communities, as well as papers that receive early exposure to a more diverse user base, are more likely to trigger cascades that are broader, deeper, and more viral, thereby substantially improving diffusion effects. [Originality/Value] Incorporating the perspective of computational communication research into Altmetrics research can help researchers gain a deeper and more comprehensive understanding of the diffusion process and patterns of scientific papers on social media, thereby revealing the mechanisms behind the diffusion of scientific papers.

Keywords: Altmetrics, Computational communication research, User attributes, Diffusion effects, Diffusion network