Abstract
Purpose: This study aims to map the conceptual structure and thematic evolution of research on responsible artificial intelligence (AI) applications in scholarly publishing. By employing a bibliometric approach combined with social network analysis, the study seeks to provide a comprehensive, evidence based overview of the field—identifying its key thematic clusters, their developmental trajectories, and the structural gaps within the existing literature. The findings are intended to inform the design of future governance frameworks by offering an empirically grounded conceptual foundation, while also highlighting areas requiring complementary qualitative or context-specific investigations.
Methodology: This applied study was conducted using a bibliometric approach combined with social network analysis and keyword co-occurrence analysis. The analysis was based on 22,918 documents indexed in the Web of Science Core Collection, published between 2020 and 2025. To map the thematic structure and assess the maturity, evolution, and dynamics of re-search domains related to the responsible use of AI in scholarly publishing, thematic maps were generated using the Bibliometrix R-package (Biblioshiny interface). Social network analysis indicators, including centrality and densi-ty, were employed to identify and cluster high-frequency keywords extract-ed from the articles. Multiple Correspondence Analysis (MCA) was applied to construct the conceptual structure of the field, revealing latent dimensions and associations among thematic clusters. VOSviewer software was used to visualize the keyword co-occurrence network and its structural properties.
Findings: The analysis revealed five distinct conceptual clusters: (1) ethical management and research integrity—encompassing principles such as trans-parency, accountability, and trustworthiness in AI adoption; (2) operational applications—focusing on data analysis, systematic review automation, aca-demic writing, peer review, and decision support; (3) post-publication impact assessment—centered on bibliometric indicators, citation analysis, and scientometric evaluation; (4) domain-specific applications—particularly in medicine and healthcare, involving clinical decision support, diagnostic systems, and radiology; and (5) search and knowledge discovery—driven by large language models, natural language processing, and retrieval-augmented generation. The thematic map positioned the operational applications cluster (cluster 2) in the first quadrant (motor themes), indicating its high centrality and density—reflecting a mature, well-established research domain. The post-publication impact assessment cluster (cluster 3) was situated at the boundary between the first and third quadrants, suggesting that while the field is gaining centrality and relevance, it has not yet achieved sufficient internal cohesion and structural consolidation. The medical and healthcare cluster (cluster 4) appeared in the second quadrant (highly developed but isolated), indicating its specialized nature and limited integration with other thematic clusters. The search and knowledge discovery cluster (cluster 5) was positioned in the fourth quadrant (basic and emerging themes), reflecting its rapid growth following the emergence of ChatGPT and other generative AI tools in late 2022. Notably, the ethical management cluster (cluster 1) was located in the lower-left quadrant (emerging or declining themes), revealing a critical paradox: while ethical discourse is highly frequent in the literature, it remains structurally underdeveloped—implying that principles are broadly accepted, yet practical implementation mechanisms are still imma-ture. Temporal analysis of keyword dynamics further indicated a significant shift in research focus: during 2023–2025, scholarly attention moved from purely technical exploration toward normative regulation, policy formulation, and governance-oriented inquiry.
Conclusion: The findings reveal that the field of responsible AI in scholarly publishing has undergone a fundamental transformation from a technical-innovative concern to a governance-ethical paradigm. Five distinct conceptu-al clusters were identified, each representing a different layer of the scholarly publishing ecosystem: meta-governance (ethical principles and research integ-rity), operational-functional (data analysis, writing, peer review), post-publication evaluation (bibliometric and scientometric assessment), domain-specific applications (medicine and healthcare), and knowledge discovery (large language models and retrieval systems). A critical paradox emerged from the analysis: while ethical principles such as transparency, accountabil-ity, and trustworthiness are widely discussed and broadly accepted, their practical implementation mechanisms remain structurally underdeveloped. The thematic map positioned the ethical management cluster in the emerging or declining quadrant, indicating that despite high frequency in the literature, operational frameworks for enforcing these principles have not yet matured. This gap between accepted principles and enforceable mechanisms represents the central challenge facing the scholarly publishing community. Further-more, the analysis revealed a significant operational-normative disconnect: the operational applications cluster, while mature and well-established, shows limited structural linkage with the ethical management cluster. This suggests that AI tools are being extensively integrated into research work-flows without adequate transparency, disclosure, or accountability mechanisms—a finding with important implications for research integrity. Similarly, the emergence of large language models in knowledge discovery high-lights those ethical risks begin at the earliest stages of the research cycle, necessitating transparency requirements that extend beyond writing and publication. These findings suggest that future governance frameworks should adopt a layered, ecosystem-based approach that addresses normative, operational, and contextual dimensions simultaneously. Such frameworks must emphasize mandatory transparency and disclosure, discipline-sensitive policy differentiation, and systematic enhancement of AI literacy among all stake-holders.