Unveiling the Causal Effects of Open Access on Paper Impact: An AI Field Altmetrics Perspective

This paper analyzes the causal relationships between OA paper impact indicators using causal inference methods. This article takes the papers in the field of artificial intelligence (AI) as the research sample and adopts descriptive statistical analysis, negative binomial regression analysis (NBREG), and difference-in-difference (DID) model from both OA and non-OA levels. This paper examines both societal and academic impact indicators, analyzes their distribution in AI papers, and explores the advantages of OA. The findings of this article are as follows: 1) In the field of AI, the differences between OA and non-OA papers across various metrics are relatively modest, with data distributions exhibiting high concentration. Nevertheless, OA papers consistently demonstrate higher mean values than non-OA papers in all indicator dimensions; 2) whether OA papers or non-OA papers, each altmetrics indicator has a significant positive influence on Dimensions citations, but the influence of OA papers is significantly higher. Among them, patent mentions has the most significant influence on Dimensions citations in the OA pattern, while policy mentions, news mentions, and Twitter/X mentions have a weaker influence on Dimensions citations in that order; 3) there is a significant causal relationship between OA and news mentions and Twitter/X mentions of papers, that is, OA significantly enhances the mentions of papers on News and Twitter/X platforms. However, the OA altmetrics advantage is not consistently stable, but shows a gradual decline over time.