Abstract: [Purpose/Significance] This paper explored the inherent characteristics of citing behavior, which is of great significance for designing scientific evaluation indicators, discovering the laws of scientific exchange and predicting the evolution of disciplines. [Design/Methodology] In this paper, the Microsoft Academic Graph was used as the data source and an algorithm attribution frameworkwas constructed based on explainable machine learning technology. Then references of 20,116 articles from 18 LIS journals were analyzed by our algorithm attribution framework for exploring the impact of various factors on citing behavior of scholars in this field and these factors’dynamic changing trend from 2000 to 2019. [Findings/Conclusion] The algorithmic attribution framework can effectively reveal the main considerations when LIS scholars citing others’work. It also can be used for discovering the change trend of each factor, and has the ability to explore the quantitative relationship between each factor and the citation probability. [Originality/Value] The conclusions of this study can provide valuable insights for understanding the citing behavior of LIS scholars and the dynamic change of their citation behavior. The introduced algorithm attribution framework provides a new method for the exploratory research on factors influencing the citing behavior of LIS scholars.

Key words: Factors influencing the citing behavior, Reference analysis, Algorithm attribution, Explainable machine learning