Quantifying the Prevalence and Diffusion of Generative AI in Science

Cutting-edge, large language models (LLMs) like ChatGPT have taken the world by storm due to their remarkable proficiency in generating content that resonates with humans. While the popularity of these generative AI models is rapidly increasing, we know little about how scientists actually use them, or how they will influence the future of scientific-knowledge making. This project fills these critical knowledge gaps by bridging expertise in computer science and the sociology of science to achieve three key objectives. First, it will provide much-needed evidence on both overt and covert usage of LLMs in science. Second, it will analyze trends in the usage and prevalence of LLMs across scientific disciplines. Finally, it will leverage this comprehensive understanding to predict the diffusion and adoption of generative AI in scientific networks. The project both builds on and extends beyond the principal investigators’ prior research by offering novel interdisciplinary perspectives on the potentially transformative impact of generative AI in science. By doing so, the project helps prepare the scientific enterprise for the challenges and opportunities presented by generative AI.