A thesis isn't enough for a PhD

The piece examines how traditional PhD assessment - coursework, a possible comprehensive exam, and above all a written dissertation followed by a public defense - has long reduced doctoral qualification to the thesis itself. Defenses are often formalities and committees rarely fail candidates in public, so a strong written document has typically been sufficient even when oral performance was weak. That model is now threatened by AI: large-language models can generate long, thesis-like texts with minimal prompting, making the dissertation a less reliable indicator of a student’s independent mathematical development or original contribution.

The response proposed and endorsed by prominent institutions is to move beyond text-as-proxy and implement regular, multi-faceted in-person evaluations conducted by multiple faculty members. These assessments must be rigorous rather than perfunctory and should yield formal reports that join the student’s dossier and inform future hiring. The practical upshot is a shift toward verification of thinking and mathematical maturity through sustained, documented faculty interaction and testing, rather than awarding doctorates primarily on the basis of a written document that can be produced or heavily assisted by AI.