The use of generative AI in scientific tasks, exemplified with ChatGPT in biostatistics
Dr. Dennis Dobler, RWTH Aachen University, Germany, Computational Statistics Teaching and Research Unit
Generative artificial intelligence (AI) offers itself for various scientific tasks including writing, summarizing, analyzing, illustrating, coding. When appropriately used, this can lead to a greater productivity of the scientist. However, reliance on AI carries the risk of inaccuracies due to potentialhallucinations that may produce fabricated "facts", leading to erroneous statistical statements and conclusions. Such errors could compromise the high precision and transparency which are fundamental to academia. We will explore both the risks andopportunities AI offers in this connection. One key question is: "What is the role of expert knowledge in this new era of AI?" We will exemplify these matter with the help of different tasks in biostatistics which are similarly relevant in other fields , e.g., sample size planning, meta analyses, simulation studies, translation between programming languages. To this end, ChatGPT 4o, o3, and o4-mini have been used. This talk is based on the following open access "tutorial in biostatistics" paper (with commentary):
Reference: D. Dobler, H. Binder, A.-L. Boulesteix, J. B. Igelmann, D. Köhler, U. Mansmann, M. Pauly, A. Scherag, M. Schmid, A. Al Tawil, S. Weber (2025). ChatGPT as a Tool for Biostatisticians: A Tutorial on Applications, Opportunities, and Limitations. Statistics in Medicine, 44(23-24), e70263, doi: 10.1002/sim.70263
Commentary by B. Zhu (2025). Biostatisticians Meet AI: Navigating Shifts While Preserving Principles. Statistics in Medicine, 44(20-22), e70271, doi: 10.1002/sim.70271
*This seminar is available for RECR Credit, 1.0 Hour. Attendance will be verified and a survey must be completed afterwards with well thought out responses to receive RECR Credit.