Breakthrough Study Reveals ChatGPT’s Unique Approach to Ancient Mathematical Challenge

Researchers Examine AI’s Problem-Solving Method in Doubling the Square, Echoing 2,400-Year-Old Philosophical Debates

Breaking Research: ChatGPT and the Doubling the Square Challenge

Date: 18th September 2025

In a fascinating exploration of the interplay between artificial intelligence and mathematics education, researchers have uncovered that ChatGPT—an advanced AI chatbot—exhibits behavior akin to that of a student grappling with classical mathematical concepts. This study, published in the International Journal of Mathematical Education in Science and Technology, revisits an age-old challenge: doubling the area of a square—a problem that traces back to the philosophical dialogues of Plato.

An Ancient Problem Resurfaces

In the annals of educational history, Plato’s dialogues featuring Socrates teaching an uneducated boy how to double the area of a square stand out as one of the earliest recorded experiments in mathematics learning. The boy initially stumbles, mistakenly believing that doubling the length of each side would suffice. Socrates, through guided questioning, leads him to understand the true solution: the sides of the new square should be equal to the diagonal of the original square.

The new study was conducted by Dr. Nadav Marco, a visiting scholar at the University of Cambridge, and Andreas Stylianides, a Professor of Mathematics Education at Cambridge, who sought to plunge ChatGPT into this mathematical inquiry. They aimed to discern whether the AI would regurgitate its pre-existing knowledge of Socrates’ solution or develop its own adaptively.

ChatGPT’s Surprising Approach

Contrary to expectations, ChatGPT did not simply recall the classical solution. Instead, it exhibited a surprising improvisational style, sometimes opting for an algebraic solution more akin to modern mathematical practices rather than the geometrical reasoning employed in Plato’s time. During their experiments, the researchers first prompted ChatGPT by emulating Socrates’ Socratic questioning style, then introduced errors and new variations of the problem to gauge its response.

At one point, ChatGPT even made mistakes reminiscent of common student errors, reinforcing its learner-like characteristics. As Dr. Marco noted, “When we face a new problem, our instinct is often to try things out based on our past experience.” ChatGPT, indeed, seemed to mirror this instinct.

The Nature of Knowledge in AI

The researchers were careful to emphasize that while ChatGPT does not "think" like a human, its behavior suggests a blend of data retrieval and improvisational reasoning. Despite demonstrating a solid grasp of Plato’s work when quizzed directly, the AI’s inclination to choose algebra over geometry presents intriguing implications for educational methods. Stylianides remarked, “If it had only been recalling from memory, it would almost certainly have referenced the classical solution right away.”

The study further examined variants of Plato’s problem, including doubling the area of a rectangle. Unfazed by previous prompts, ChatGPT again gravitated toward algebra, even erroneously asserting that a geometrical solution was unavailable. Such missteps illuminate the potential space between what ChatGPT "knows" and how it applies that knowledge.

A Learning Opportunity

The findings hint at a metaphorical "Chat’s Zone of Proximal Development" (ZPD), a concept outlining the gap between what a learner knows and what they could achieve with support. Marco and Stylianides advocate for leveraging this zone by prompting and engaging with ChatGPT, transforming its limitations into opportunities for deeper mathematical understanding.

As Stylianides pointed out, understanding AI-generated proofs will increasingly become critical in mathematics education. By redefining how we instruct students—encouraging them to explore problems collaboratively rather than simply seeking direct answers—educators can foster a more dynamic learning environment.

Conclusion

This groundbreaking research invites educators to rethink the way AI tools like ChatGPT can be integrated into mathematics education. By treating these tools as partners in exploration rather than straightforward answer machines, students may develop essential analytical skills that are foundational to mathematical reasoning. As we advance into an era where technology and education intersect more profoundly, understanding the nuances of AI’s "thinking" may be just as important as mastering traditional mathematical concepts.