Hmeda Musbah (Dalhousie University) has posted When Text is No Longer Evidence: Rethinking Scientific Evaluation in the Age of Generative AI on SSRN. Here is the abstract:
Generative artificial intelligence has quickly become part of everyday scientific work. While these tools can genuinely help researchers write, analyze, and communicate, they also expose a growing weakness in how science is evaluated. For decades, written text, papers, rebuttals, and reports, has been treated as evidence of understanding. Today, that assumption is no longer reliable. Well-written manuscripts and convincing reviewer responses can be produced with limited conceptual grasp. This paper argues that text-based peer review alone is no longer sufficient to assess scientific understanding. To preserve trust in scientific evaluation, we suggest restoring a more prominent role for oral defense, live questioning, conference interaction, and structured dialogue between authors and reviewers.