Leiden Declaration on Artificial Intelligence and Mathematics

Signatories

The primary signatory list now lives on the homepage below the declaration text.

Sign Declaration
4162 Signatories
  1. Claude-Michel Brauner ORCID Professor Emeritus, Université de Bordeaux
  2. William Daniel Stephenson ORCID University of Ottawa
  3. Dell Zhang ORCID Chief Scientist, XSci
  4. János Marcell Benke ORCID University of Szeged
  5. Nathan Chen verified email Purdue University
  6. Paul Goldberg ORCID Reuben College; Department of Computer Science, University of Oxford
  7. Riccardo Zuffetti verified email Technische Universität Darmstadt
  8. Michael Reitmeir ORCID University of Bonn
  9. Sihyun Song ORCID Yonsei University
  10. Bernhard Keller ORCID Universtié Paris Cité
  11. Erin Schwertner-Watson verified email University of New Mexico
  12. António Machiavelo ORCID Department of Mathematics, University of Porto, Portugal
  13. Hyungmin Jang ORCID department of mathematics, Yonsei University
  14. Dominick Range ORCID Academic Researcher, University of California, Santa Cruz
  15. Radoslav Harman ORCID Prof., Comenius University
  16. Iliana Moseley ORCID LMU / TUM München
  17. Soumya Dey ORCID Krea University, India
  18. SAUNAK BHATTACHARJEE ORCID PhD Candidate, UNSW Sydney
  19. Christian Risco ORCID The University of Queensland
  20. Cesare Borgia ORCID Student, University of Electronic Science and Technology of China
  21. Martin Man-chun Li verified email The Chinese University of Hong Kong
  22. Ben Marlin verified email Purdue University
  23. Yunfei Gao Master's graduate，Chinese Academy of Sciences
    Comment

    The development of AI has provided effective assistance to various scientific research fields. However, as the declaration states, the validation of results has become a significant challenge. If human experts lack the methods or time to verify AI outcomes, how can we trust the correctness of the results? Meanwhile, as the declaration notes, 'solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight,' we should devote more effort to understanding the principles behind complex and profound issues.

  24. Anthony Rossi ORCID Lucerne University of Applied Sciences and Arts
    Comment

    "I hope we will be able to solve these problems before we leave" - Paul Erdős

  25. Weihao Xia verified email Louisiana State University
  26. Jeanine Van Order ORCID
  27. Gabriel Bonuccelli Heringer Lisboa ORCID Master's student, Universidade de São Paulo
  28. Gábor Pete ORCID Alfréd Rényi Institute of Mathematics
  29. Yaxin Tu ORCID Doctoral, Princeton University
  30. Crichton Ogle verified email The Ohio State University - Columbus
  31. Maria Cristina Pereyra ORCID Professor, University of New Mexico
  32. alejandro r. matta verified email Edgewood College
  33. Matthew Powell ORCID Rice University
  34. Lizhen Zhang ORCID post doc., McGill University
  35. Giovanni Interdonato ORCID Scuola Normale Superiore
  36. Mengwu Guo ORCID Associate Professor, Lund University
  37. Colm-cille Caulfield ORCID Professor of Environmental and Industrial Fluid Dynamics, Department of Applied Mathematics and Theoretical Physics, University of Cambridge
  38. Nathanael Pribady ORCID Doctoral Researcher, University of Calgary
  39. Omar Essakine verified email Université Laval
  40. Matthew Salter Mathematical Association of America
  41. Carlo Cossu Directeur de recherche, Centre National de la Recherche Scientifique (CNRS), France
  42. Ward Nijhuis verified email University of Amsterdam
  43. Yan Xuan ORCID Math, Boston University
  44. Alexander Korbonits ORCID
  45. Sougata Panda Indian Statistical Institute
  46. Ty O Easley ORCID PhD Student, Washignton University in St Louis
    Comment

    As academics, we have a responsibility to take seriously the massive amount of intellectual theft undergirding modern LLM/LRMs. Even when correct statements are generated in response to prompts, the model training process often renders proper attribution impossible. All (current) frontier models (and their predecessors) are trained on nonconsensually obtained data, which means that any work produced with such models fundamentally contradicts prevailing notions of authorship in the academic community. More pressingly, as citizens of the world, we are obliged to thoroughly investigate the environmental and sociopolitical costs of these tools before we use them. Incorporating these tools in our workflows, especially if done uncritically, risks significant harm to our communities, along with those of every global majority. In particular, when measuring the ethical ramifications of LLM/LRM-assisted work, we have a moral duty to weigh the LLM/LRM's (a) relationship to data sovereignty, (b) propagation of environmental harm (including and especially environmental racism), and (c) propensity for political and/or military weaponization alongside any implications of the *content* of the work itself. While the mathematical community is often reluctant and ill-equipped to manage these kinds of ethical responsibilities, the Leiden declaration is an important step towards a more unified and active conscience.

  47. Peter Goetz ORCID Professor, Cal Poly Humboldt
  48. Valeria de Paiva ORCID Principal Researcher, Topos Institute
    Comment

    New technologies have always brought risks, unintended consequences, and new opportunities. Large language models and chatbots are no exception—and they are, in many respects, a genie that is already out of the bottle. There is no realistic way to turn back the clock. I am therefore happy to sign the Leiden Declaration in support of Open Science, proper open publishing—including diamond open access—and the rights and autonomy of individual researchers. These principles become even more important as new technologies transform how research is produced, evaluated, disseminated, and credited. The challenge before us is not to prevent technological change, but to shape it responsibly. We should acknowledge both its opportunities and its risks, while making sure that researchers and the broader scholarly community are protected from its unintended consequences. The genie is out of the bottle; our task now is to make sure that the people, institutions, and values at the heart of research are not left behind. (written with the help of a chatbot)

  49. Volodymyr Nekrashevych ORCID Professor, Texas A&M University
  50. Alejandro Kocsard verified email Instituto Nacional de Matemática Pura e Aplicada - IMPA
  51. Javier de Lucas ORCID Associate Professor, University of Warsaw
  52. Mentzelos Melistas ORCID University of Twente
  53. Fabrizio Del Monte ORCID Assistant Professor, University of Birmingham
  54. Alun Wyn-jones Mathematical Models, Inc. (owner), independent researcher
    Comment

    I support the Leiden Declaration. I would add that AI is becoming a destructive influence on the education of our children. However, I believe that point 03 under Recommendations for Policymakers is the most important immediate goal since otherwise we will be overwhelmed by commercial interests and individuals in control of AI resources.

  55. Shaowu Zhang ORCID PMA, California Institute of Technology
  56. Ioan ROXIN ORCID Professeur émérite, Université Marie et Louis Pasteur (Université de Franche-Comté)
  57. Subhadip Chowdhury ORCID University of Chicago
  58. Andor Lukacs ORCID Lect. Dr., Universitatea Babeş-Bolyai
  59. Karen Gunderson ORCID University of Manitoba
  60. Marcus Waurick ORCID University Professor; Chair for Partial Differential Equations, TU Bergakademie Freiberg
    Comment

    The recent developments concerning the Navier—Stokes millennium problem have shown that this declaration is dearly needed and adhering to it is required by all bodies mentioned. Indeed, not so much an apparent solution (as of now still pending peer-scrutiny) of one of the Clay problems but rather the discussion that came with it show the complicated entanglement of mathematical research, personal interests, the question of resources and the tendency of AI companies being potentially problematic as being more or less commercial interest led endeavours in the whole discourse that is the centre of this declaration. Hence, I wholeheartedly signed this declaration.

  61. Shayaan Emran verified email Johns Hopkins University
  62. Sug Woo Shin ORCID UC Berkeley
  63. John S. Nolan ORCID Visiting Assistant Professor, The Ohio State University
  64. Riccardo Costantini "Physics PhD student, Universität Regensburg"
  65. Eduardo Souza Fraga ORCID Full Professor, Universidade Federal do Rio de Janeiro
  66. Konstantinos Avgerakis verified email Technische Universität München
  67. Gabriel Soto ORCID Profesor Titular, Universidad Nacional de la Patagonia San Juan Bosco
  68. Diogo Arsénio ORCID New York University
  69. Paula Truöl ORCID Postdoc, University of Glasgow
  70. Maxine Calle ORCID HCM postdoctoral scholar, University of Bonn
  71. Michael Wallner ORCID TU Wien
  72. Federico Vigolo ORCID Junior Professor, Georg-August-Universität Göttingen
  73. Joel Pulikkan verified email Rutgers University
    Comment

    The future will wish they had lived in the past.

  74. Shubham Sinha ORCID Postdoctoral Fellow, University of British Columbia
  75. Dipramit Majumdar ORCID Indian Institute of Technology Madras
  76. Jaco Ruit ORCID Postdoctoral fellow, The Hong Kong University of Science and Technology
  77. William Sheppard ORCID University of California, Santa Barbara
  78. Assad Oberai ORCID Professor, University of Southern California
  79. Emilio Annoni ORCID Quandela (France)
  80. Nathan Hayes ORCID PhD, University of Illinois Urbana-Champaign
  81. Andrew Robinson ORCID Professor, University of Melbourne
  82. Woo Young Lee ORCID Professor Emeritus, Seoul National University
    Comment

    Upon hearing the news that AI model solved the Navier-Stokes Millennium problem: Why do we devote ourselves to solving the long-standing open problems ? Is it the solution itself that truly matters, or is it the journey—the experience of roaming freely through the world of thought? We may find an answer in the following words from Morris Kline’s foreword to Bertrand Russell’s celebrated book <An Essay on the Foundations of Geometry>: “The insolvable problems continually entice the human mind and the greatest intellectual pleasures are derived from the constant striving after elusive truth. The works of great minds keep the problems before our eyes and prevent us from sinking into mental lethargy.“ Indeed, the true meaning of mathematics lies in keeping the human spirit awake.

  83. François Lalonde Université du Québec à Montréal
  84. Patrick Gambill verified email Washington State University
  85. Alex Dunbar ORCID Hale Visiting Assistant Professor, Georgia Institute of Technology
  86. Thomas Nikolaus ORCID Professor, University of Münster
  87. Kevin Gomez ORCID Tandon School of Engineering, New York University
  88. Amaury Lambert ORCID Ecole Normale Supérieure, Paris, France
  89. Takis Konstantopoulos Professor of Mathematics, University of Liverpool
    Comment

    1) Mathematics has traditionally been one of the most democratic intellectual activities. Talent, curiosity, and hard work matter far more than wealth or status. AI risks changing this. If access to the best mathematical tools depends on the ability to pay, then economic privilege becomes a significant advantage. The mathematical community should explicitly RESIST A FUTURE IN WHICH ACCESS TO MATHEMATICAL INQUIRY IS DETERMINED BY WEALTH. 2) AI use by beginners deserves caution. Effective prompting, evaluation of answers, and detection of errors require mathematical expertise that most students do not yet possess. For them, AI can encourage dependency and superficial understanding rather than genuine learning. Universities should be more critical of claims that AI is universally beneficial in mathematics education. 3) AI can be a useful encyclopedic aid, but successful use presupposes prior mathematical experience. Without a solid foundation, users may be unable to distinguish insight from plausible-sounding error. 4) AI can only be as good as the material on which it is trained. It performs best in areas with strong traditions of rigor and worst where low-quality work is common. A serious concern is a feedback loop: poor material produces more poor material. Garbage in, garbage out. Today's systems draw on centuries of carefully accumulated mathematical knowledge. Will future systems be able to say the same? 5) Mathematics is not primarily about obtaining answers. It is about achieving UNDERSTANDING. A proof is valuable because it provides insight and explanation, not merely because it establishes truth. Papers should communicate human understanding, not merely machine output. AUTHORSHIP AND RESPONSIBILITY MUST REMAIN HUMAN, EVEN WHEN AI HAS BEEN USED AS A TOOL.

  90. Dario Faro ORCID Università degli Studi di Milano
  91. Noel Arteche ORCID Doctoral student, Lund University
  92. Ioan Marcut ORCID Professor, University of Cologne
  93. Christopher Deninger verified email Westfälische Wilhelms-Universität Münster
  94. Olivier Faugeras ORCID Emeritus Researcher, Inria Centre de Recherche Sophia Antipolis Méditerranée
  95. Jan-Willem van Ittersum ORCID Postdoc, University of Amsterdam
  96. Paulo Carvalho ORCID Associate Professor, Universidade do Minho
  97. Aleksandr Trufanov verified email Université de Montréal
  98. Jacopo De Simoi verified email University of Toronto
  99. Tae-Geun Kim ORCID Postdoctoral Researcher, Fudan University and RIKEN iTHEMS
  100. (Alex) Jiayi Zhang ORCID PhD Student(Math and Theoretical Physics), Durham University