Harvard University has released the first formal recommendations for redesigning PhD mathematics education in response to the rapid rise of artificial intelligence, following a landmark summit held 17–18 September 2026.
The Summit on PhD Math Education in the Age of AI brought together 24 senior mathematicians and educators from leading US institutions—including MIT, Stanford, Princeton, UC Berkeley, and Columbia—to address a challenge now facing universities worldwide: how to ensure that doctoral graduates possess genuine mathematical understanding and originality, not just the ability to produce AI-assisted research outputs.
The summit’s recommendations, published online and circulated to the global mathematics community, call for a fundamental shift in assessment and training. Key proposals include:
- De-emphasising written homework and dissertation text as sole evidence of understanding, since AI can now generate plausible answers and even novel proofs.
- Prioritising in-person examinations and oral defences to directly assess a candidate’s depth of knowledge and ability to explain their work.
- Broadening the definition of originality in doctoral research, recognising that AI can rapidly generate technical results, and focusing instead on creative inquiry and the ability to direct research independently.
- Encouraging multifaceted evaluation for hiring and advancement, moving beyond reliance on publication records or written submissions.
The summit’s organisers stress that AI should be used to accelerate human understanding, not to bypass it. They warn that traditional proxies for expertise—such as submitted documents or even the production of a “world’s first” proof—are no longer reliable indicators of a mathematician’s formation. Instead, the focus must shift to cultivating and verifying independent, responsible inquiry, including the ability to collaborate productively with AI tools.
These recommendations are likely to influence policy and practice far beyond the US. Universities in Aotearoa and across the Asia-Pacific are already grappling with similar questions about academic integrity, research assessment, and the safe, effective integration of AI into advanced education. The summit’s call for staged pilots and explicit recognition of the costs and opportunities of AI in doctoral training provides a practical roadmap for institutions seeking to harness AI’s benefits while safeguarding the quality and trustworthiness of mathematical research.
Uncertainty and Next Steps
The recommendations are described as a “first draft,” with organisers inviting feedback from the global mathematics community. The summit’s outcomes highlight both the promise and the risks of AI in research: while AI can open new avenues for exploration, it also threatens to erode the traditional markers of expertise and originality if not carefully managed. The coming months will see further debate and likely pilot programmes as universities worldwide seek to adapt.
For Aotearoa and the Asia-Pacific, the summit’s work offers a timely reference point as local institutions consider how to maintain academic standards and foster genuine innovation in the AI era.
Sources:
note.com (27 September 2026)
terrytao.wordpress.com (25 September 2026)
AI disclosure: Automated systems assisted with research and drafting. Gemini independently peer-reviewed this article before publication. Sources are linked for verification.
