GENERATIVE ARTIFICIAL INTELLIGENCE IN CO-AUTHORSHIP NETWORK ANALYSIS
Keywords:
artificial intelligence, bibliometrics, co-authorship networksAbstract
This paper explores the application of generative artificial intelligence (GenAI) and bibliometric tools in the analysis of academic co-authorship networks. The central research problem investigates how different GenAI approaches and specialized software compare in terms of pertinence, methodology, completeness, and performance when analyzing co-authorship networks through the use of the same contextualized prompt. The general objective was to comparatively analyze these tools in the identification of authorship and co-authorship patterns in scientific articles. The methodology adopted was a comparative and exploratory study, with a qualitative and quantitative approach. It consisted of applying an identical prompt to the three GenAI tools (ChatGPT, Gemini and Manus) and submitting the same corpus to VOSviewer, through a procedure compatible with the software’s operation, in order to analyze 627 articles from BRAPCI, focusing on data cleaning, author extraction, publication counting, network construction, metric calculation, and visualization. The data generated were submitted to technical-methodological comparative analysis. The main results revealed convergence in fundamental findings, such as the number of unique authors, network density, and the largest connected component, which validates the overall structure of the network. Specific discrepancies emerged in the exact number of edges and in the methodologies for community detection. ChatGPT presented greater analytical completeness and better methodological explicitness, while VOSviewer maintained its prominence in specialized visualization.