S04-07 13

ChatGPT and Mathematics Education: exploring its mathematical competencies


Deja tu comentario

Participa en esta ponencia enviádole tu pregunta o comentario a los autores

Añadir comentario


profile avatar
David Zatz CorreiaPontifícia Universidade Católica de São Paulo
profile avatar
Celina A A AbarPontificia Universidade Católica de São Paulo



In January 2023, after merely two months since its launch, ChatGPT reached a hundred million active users, a significant milestone that shows that it can not be overlooked. Therefore, it is known that society is actively engaging with it and thereby, it is necessary that its use in the field of education is explored.

ChatGPT is a large language model (LLM) that employs generative artificial intelligence. Its nomenclature refers to the way it interacts with its users, through text messages (chat) and to the methodology of its training as a Generative Pre-trained Transformer (GPT). Due to its nature as an LLM, ChatGPT may hallucinate, fabricating information that lacks any basis in reality.

It was created by OpenAI and gained popularity shortly after its launch in November 2022. In spite of its widespread adoption, the language model still causes disagreement among educators. On one hand, there are those who believe in its potential to be the futue of teaching, learning and research, and, on the other hand, there are those who feel concerned that it may diminish the analytical capacity of both teachers and students.


This study aims to explore the potentialities and limitations of ChatGPT regarding its use in Mathematics Education, specifically fucusing on assessing its mathematical competencies in solving exercises.


To achieve these objectives, this research adopts a qualitative approach focused on the description and characterization of phenomena, and has an exploratory nature.

The free version of ChatGPT was utilized, as it holds the potential to be used in educational institutions. Its potentialities and limitations were analyzed according to what was proposed by some authors.

Examples were selected due to their clarity to demonstrate the generative artificial intelligence’s potentialities and limitations in Mathematics Education and, specifically, in the resolution of mathematical exercises.


For this study, some examples of exercises proposed to ChatGPT and its respective responses are presented. These responses were analyzed in terms of the LLM’s mathematical competence and its potentialities and limitations.


ChatGPT demonstrated some strengths, such as language understanding, speed and a logical and well-organised approach, and limitations, such as the misinformation and ‘alternative facts’. It was possible to verify that there is a deficiency in this language model’s mathematical problem-solving capabilities, as it made errors during our investigations.


In face of the presented limitations, it is imperative to exercise caution regarding the uncritical use of the free version of ChatGPT in Mathematics Education. Nevertheless, it is believed that this language model could serve as an additional tool to facilitate learning.

Preguntas y comentarios al autor/es

Hay 13 comentarios en esta ponencia

    • profile avatar

      Fernando Raúl Alfredo Bordignon

      Comentó el 29/11/2023 a las 21:40:47

      Los firmantes de esta ponencia recuerden que el modelo de participación del Simposio indica que cada uno debe realizar dos preguntas a otros autores de trabajo. Adelante!.

    • profile avatar

      Fátima Rodríguez García

      Comentó el 29/11/2023 a las 09:39:13

      Congratulation for your presentation. It has been really interesting.

      I would like to know if you have evaluated the differences in outcomes based on the use of ChatGPT-3 versus ChatGPT-4 in educational settings, particularly in terms of effectiveness, accuracy, and student engagement?

      Thank you for your answer

      • profile avatar

        David Zatz Correia

        Comentó el 29/11/2023 a las 23:17:30

        Thank you for your question.
        At this moment, we haven't made that comparison yet. However, it is in our plans for future research.

    • profile avatar

      Javier de la Hoz Ruiz

      Comentó el 29/11/2023 a las 08:55:50

      Enhorabuena por la gran presentación realizada, muy productiva y con mucho sentido. ¿Podríais facilitarme acceso adicional a algunos recursos relacionados en el área?

      • profile avatar

        David Zatz Correia

        Comentó el 29/11/2023 a las 23:18:11

        No tengo no momento recursos disponibles que posa compartir.
        Lo sinto!!

    • profile avatar

      Javier Mula Falcón

      Comentó el 29/11/2023 a las 07:51:11

      Congratulations on a very interesting presentation. Best regards.

    • profile avatar

      Emanuele Fulvio Perri

      Comentó el 28/11/2023 a las 18:20:53

      First of all, I'd like to compliment with you, David and Celina, for your presentation. I'm into Ethics of AI and currently working as close as possible on the EU's AI act. In this regard, I agree with your conclusions—about the potential benefits of correctly employing GPTs, but also about the risks of their uncritical (mis)use; however, I do think that is important to define procedures for determining all those attempts of unfair uses from students, employees, etc. who are asked to complete certain tasks (like homeworks, assignments, research, etc.). This is important, considering that the majority of anti-plagiarism tools fail in detecting ChatGPT-generated texts: in the mathematical field this "detecting" can only be performed by a skilled mathematician, right? How can we improve the ethical side of deploying generative transformers (in this case for mathematics) in the educational filed? Thank you in advance for your kind reply.

      • profile avatar

        David Zatz Correia

        Comentó el 29/11/2023 a las 02:33:26

        Thank you for your question.
        I agree that it is important to define these procedures, given that ChatGPT has a human-like writing and it is not easy to detect when it has been used. This is not an easy question to answer and it has been a cause for concern among educators worldwide and we do not have a definitive answer for it. However, some authors recommend that teachers assign less homework and engage in more in-class activities, and others say it is important to develop activities in a way that ChatGPT could not be used to answer them. Another possible approach is to show students that generative transformers have limitations and can make mistakes in some cases, making it less tempting to use them for homework assignments.

    • profile avatar

      Fernando Raúl Alfredo Bordignon

      Comentó el 28/11/2023 a las 14:04:43

      Congratulations for your contributions to the Symposium. Colleagues David and Celina, I leave you with a question, from your experience, to what extent do students know the "behind the scene" of these tools? That is, how they work, their limitations in terms of handling and understanding the semantics, the biases, the business models that they try to impose from the EdTech sector. If so, how do they approach this issue?

      • profile avatar

        David Zatz Correia

        Comentó el 29/11/2023 a las 02:34:06

        Thank you for your question.
        Our research does not involve students at this point. However, based on my experience, I can say that I haven't met a high school student that has shown me they understood how ChatGPT works or that it has limitations and biases.

Deja tu comentario

Lo siento, debes estar conectado para publicar un comentario.


Egregius congresos


Egregius ediciones