Core principles of responsible generative AI usage in research

Abstract

In a rapidly evolving Generative Artificial Intelligence (GenAI) landscape, researchers, policymakers, and publishers have to continuously redefine responsible research practices. To ensure guidance of GenAI use in research, core principles that remain stable despite technological advancement are needed. This article defines a list of principles guiding the responsible use of GenAI in research, regardless of use case and GenAI technology employed. To define this framework, we conducted an anonymised Delphi consensus procedure comprising a panel of 16 international and multidisciplinary experts in AI, social sciences, law, ethics, and scientific publishing. After three rounds of independent rating and feedback, the panel reached consensus on eight sequentially ordered principles required for responsible GenAI usage: Regulations, Data Security, Quality Control, Originality, Bias Mitigation, Accountability, Transparency, and Broader Impact. For the clear reporting of adherence to these principles, we created a checklist allowing active implementation into the research process. With these efforts, we aim to guide everyday research, support the development of further specified regulations, policies, and guidelines, and promote discussion about GenAI use in research. Supplementary Information: The online version contains supplementary material available at 10.1007/s43681-025-00768-8.

Ähnliche Arbeiten