Understanding the Dynamics of Retracted AI Publications: A Comprehensive Study

Abstract

This current study accounts for most retracted works in Artificial Intelligence, providing a scientometric analysis to cover detailed trends, causes, and implications of those retracted works. This paper performs a bibliometric analysis of withdrawn papers in the AI domain during 2015-2024, tracing trends, networks of authors, and citation impacts. Data were retrieved using the Web of Science by the keyword "Artificial Intelligence, " a dataset containing 1,147 documents published in 100 journals that demonstrates fantastic annual growth of 39.23%. The mean document age is 2.17 years, indicating the research output is relatively recent and part of a strong citation network made up of 25,351 references. The analysis shows high international collaboration, with 20.05% documents holding international co-authorships. Emergent disciplines include Telecommunications, Computer Science Information Systems, and Engineering, the most represented being the core journals: Computational Intelligence and Neuroscience, adding to the landscape of retracted papers. Local contributors are authors and institutes making numerous impactful contributions; the Egyptian Knowledge Bank and King Abdulaziz University are most prominent. Geographical location wise, the scientific production heavyweights are China, India, and Ethiopia. Keywords like "classification" and "system" are often used, showing the durability of the topics in the contemporary trends of research. Findings The results enable critical insights in the dynamics of retracted AI publications based on their importance towards enhanced research integrity and provide a foundation for future investigations into the continually changing landscape of academic publishing in AI.