Recent Developments in Mathematics and Education
On 8th September, two significant announcements emerged in the realm of mathematics, albeit from vastly different contexts. Firstly, OpenAI unveiled a potential solution generated through artificial intelligence (AI) for one of the seven Millennium Prize Problems. Concurrently, the Organisation for Economic Co-operation and Development (OECD) released the findings from the 2025 Programme for International Student Assessment (PISA), which revealed a decline in mathematics and reading skills among participating countries.
The Millennium Prize Problems encompass seven formidable challenges, meticulously identified by the Clay Mathematics Institute as some of the most perplexing unsolved mathematical conundrums. To date, only one of these problems has been resolved—the Poincaré Conjecture, solved in 2002. OpenAI now claims to have discovered a potential answer to the Navier-Stokes problem, which aims to elucidate the natural behaviours of fluids such as water and air. This proposed solution is still pending independent verification. The science community has engaged in a discussion regarding the credit for this announcement, as detailed by WIRED.
In contrast, the PISA 2025 assessment evaluated over 760,000 15-year-old students across 91 countries and economies. Alarmingly, one in five students from OECD countries scored poorly in mathematics, reading, and science, with the average mathematics score dropping to its lowest level since 2003, the year when mathematics was first assessed as a core subject in PISA.
These two announcements converge at a pivotal moment: AI is beginning to tackle some of the most challenging mathematical problems known to humanity, while simultaneously, it has infiltrated classrooms where a segment of the student population continues to grapple with fundamental skills.
The Integration of AI in Learning Environments
For the first time, the PISA assessment inquired into how students utilise AI chatbots in their learning processes. The responses indicated a widespread adoption of these tools, although the outcomes in mathematics varied significantly between countries. In Latin America, Colombia reported one of the highest weekly usage rates at 56%, with 29% of students reaching at least level 2 in mathematics. In Argentina, over half of the students, specifically 52%, reported using AI tools weekly, while 24% achieved the same level in the assessment.
Mexico’s usage of AI aligns closely with the OECD average at 47%, with 30% of students attaining level 2. Spain showcased a different scenario, with 67% of students reaching this benchmark and 54% indicating regular use of chatbots for learning purposes.
PISA categorises level 2 as a foundational reference, encompassing skills such as representing simple situations mathematically, comparing two routes, or converting prices between currencies. These capabilities are essential for progressing towards more complex problem-solving.
AI has also become prevalent in various aspects of academic work. Participating adolescents reported using AI to gather information on topics, summarise readings, or prepare texts, with the mentioned countries frequently surpassing the OECD average in these activities.
Evaluating AI Responses in Education
PISA 2025 introduced an assessment focused on digital learning. The evaluation required students to utilise computational tools to seek solutions and break down complex problems into manageable steps. Nearly two-thirds of participants achieved the expected level, although only half demonstrated basic proficiency in mathematics, reading, and science.
Among those who frequently employed AI for learning, those who had practiced evaluating information generated by these tools tended to perform better in science assessments. The OECD clarifies that this does not imply a direct cause-and-effect relationship but suggests that teaching students to critically assess AI-generated responses may enhance their utilisation of these resources.
The organisation acknowledges that AI tools can provide explanations, feedback, and personalised support. However, for these tools to facilitate learning effectively, students must engage with the information: comprehending it, connecting it to other knowledge, asking questions, and identifying errors. When AI alleviates part of this effort, it may inadvertently diminish the depth of learning.
As OpenAI employs AI systems to tackle highly complex mathematical problems, students from various countries are increasingly relying on these technologies to research, summarise, write, or solve problems. While PISA does not assert that AI usage leads to poorer performance, the findings underscore that critical thinking—a crucial skill in the age of AI—remains an area where most students are not sufficiently developing their capabilities.