Mathematical Illustration Expands Research Beyond Formal Proof

A specialized workshop held in January 2026 at the Institut Henri Poincaré in Paris convened mathematicians, historians, artists, and educators to redefine the role of illustration in mathematical research. Organized as part of a dedicated trimester, the event sparked a broader academic dialogue about the future of mathematical practice in an era increasingly dominated by artificial intelligence. The central inquiry regarding rigorous mathematical illustration has since evolved into a pressing discussion on how the discipline measures and values human understanding beyond automated theorem production. Leading scholars argue that the traditional reliance on theorem generation as the primary metric of mathematical success is becoming obsolete. With AI capable of verifying complex proofs and identifying novel structures, such as the recent discovery of the three-dimensional aperiodic monotile known as Chair44, the historical scarcity that once made deep theorems a reliable proxy for insight has vanished. Experts emphasize that mathematics has never been merely a verifiable registry of results. Rather, it is a scholarly community engaged in interpretation, digestion, and contextualization of knowledge. The Leiden Declaration on Artificial Intelligence and Mathematics, released alongside these discussions, formally acknowledges this shift, urging a recalibration of academic credit toward human-centric intellectual labor. The workshop explored how visual and spatial representations function as active research tools rather than mere explanatory aids. Drawing on historical training methods from nineteenth-century institutions like the École Polytechnique, participants highlighted that drawing and modeling have long served as cognitive processes for discovery. Contemporary practitioners, organized under the Illustrating Mathematics community, demonstrated how computational shaders and algorithmic visualization can reveal hidden mathematical structures. In one documented case, an unintentional geometric distortion in a shader unexpectedly exposed a deeper topological property, illustrating how creative manipulation can drive research forward. However, participants also cautioned that AI-assisted visualization risks creating opaque outputs, where the generative process itself remains unexamined, distancing creators from the underlying mathematical logic. To address these challenges, the panel proposed a formal framework for evaluating illustrations grounded in their functional affordances, properties that computational proofs can rigorously validate. They also drew parallels with academic art criticism, advocating for communal review processes that examine how visual representations are received, interpreted, and iteratively refined. This approach aligns with growing movements like the Geometry Labs United network, which integrates undergraduate experimentation with faculty mentorship, and recent mathematics education research emphasizing tactile and visual learning. As the mathematical community navigates this transition, the consensus emerging from the IHP trimester is clear: institutional recognition must expand to honor exposition, visualization, pedagogy, and collaborative sense-making. By embedding these practices into academic valuation systems, the field can preserve its humanistic core while leveraging new technologies. The ongoing discourse aims to establish a more resilient mathematical ecosystem, one that prioritizes conceptual clarity, aesthetic insight, and collective understanding alongside formal proof.