MEGA: Mathematics Explanations through Games by AI LLMs
Background: Many students struggle to understand Mathematics (Math) because existing methods for explaining some concepts fall short. This discourages some students from pursuing Math-related disciplines at higher education institutions (HEIs). Particularly, 13% of 15-year-old Swedish students scored above level 4 and, while girls scored similarly to boys, boys across the OECD outperformed girls by five score points, in a relatively recent PISA survey.
Aim:
MEGA aims at improving the learning of students in Math through a novel combination of interactive fun games and alternate explanations by application of large language models (LLMs), as 80% of students play computer games.[1]
Scope: MEGA applies to students of both HEIs and High Schools.
Methodology:
This will involve prompt engineering through instructions and Langchain for LLMs. The strategic actions involve:
- Literature review and experimenting with Math explanation-oriented instructions for state-of-the-art (SotA) LLMs, including ChatGPT, LLaMA-2 and Mixtral 8X7B.
- Experimenting with interactive game instructions for the same models.
- Merging the 1st and 2nd actions into the comprehensive set of LLM instructions for students to utilize for a fun, learning experience, as guided by their teachers.
An example problem is the simultaneous equations: 2x – 3y = 13; 6x + y = 7; What are x and y? When a student instructs an LLM to explain the solution, the model guides the student and awards scores to the student which may be put on a leaderboard. If the student fails to understand then an alternate explanation is offered using the earlier process. The possible risk that may arise from this method is addressed in ProCot.[2] We commit to sharing insights from the research with the rest of the department through pedagogical fika sessions and 2 publications (1 each in an educational venue and a journal on LLM).
Impact and Future Projects
With this project, we will enhance our expertise and portfolio in AI education for various subjects. We plan integrating the outcomes into other courses. We also plan collaborating with the already existing AI4Edu project and lay some ground work for future EU applications, such as the AI4Math, which was submitted in February, and planned projects with our international partners from Belgium, Greece, and France (already 2 successful projects and 4 project proposals). We will also connect with the joint educational effort between Örebro and LTU, where the researchers at TVM (Stefan Ericsson) are creating Math education.
Schedule:
Action 1: March 1st - April 30th
Action 2: March 1st - April 30th
Action 3: May 2nd - August 30th
Publishing: May 2nd - December
Contact
Oluwatosin Adewumi
- Postdoctoral researcher
- 0920-49
- oluwatosin.adewumi@ltu.se
- Oluwatosin Adewumi
Updated: