Publishing date: 11 June 2026 Immagine notizia Immagine Paragrafo A new international study published in PNAS (Proceedings of the National Academy of Sciences), one of the world's most prestigious scientific journals, examines the potential of artificial intelligence to support the assessment of research reproducibility in quantitative social sciences.The study, entitled “AI-assisted teams outperform AI-led teams but not human-only teams in assessing research reproducibility in quantitative social science”, was conducted by the Institute for Replication (I4R), an international network dedicated to promoting the verification and robustness of scientific research.Article link: https://www.pnas.org/doi/10.1073/pnas.2524747123Among the authors is Raffaello Seri, Full Professor in the Department of Economics at the University of Insubria, Director of the InsIDE Lab, and Coordinator of the Department of Excellence Project 2023–2027.The research involved 288 researchers organized into 103 experimental teams: 33 groups composed exclusively of human researchers, 35 teams supported by artificial intelligence tools, and 35 groups in which AI played a predominant role with minimal human intervention.The aim of the study was to evaluate whether generative AI tools, such as ChatGPT, can effectively contribute to assessing the reproducibility of quantitative scientific studies in the social sciences. This process is considered essential for ensuring the reliability of research, yet it has traditionally been highly demanding in terms of time, effort, and resources.The findings show that AI-assisted teams and teams composed entirely of human researchers achieved similar performance across most analytical tasks, including reproduction rates exceeding 90%. However, human researchers proved more effective at identifying critical errors during the scientific verification process.The study also found that teams relying primarily on AI systems achieved a substantially lower reproduction rate of 37%. This suggests a potential future role for artificial intelligence as a preliminary automated screening tool, particularly in contexts where a full human review would be prohibitively costly.According to the authors, the results indicate that artificial intelligence can provide promising support for scientific validation processes, while also confirming the central role of human expertise and judgment in the reliable verification of research findings.The study contributes to a growing international body of research examining the impact of artificial intelligence on the production, validation, and dissemination of scientific knowledge.