Video series on mathematics behind AI
Abstract/description and goals
Although many videos on AI exist, few explain how algorithms learn, search for patterns, and categorize data, while still being accessible for high school students. Most are either too advanced for Dutch high school students or only available in English. Together with Mathness, we are developing a video series on the mathematics behind artificial intelligence (AI) for Dutch secondary education. With support from 4TU.NIRICT, three videos are already in production:
- Introductory video for all levels (already produced)
- How to build a Neural Network (perceptron explained simply)
- How to train a Neural Network (backward propagation explained simply)
Our series goes beyond general descriptions by offering step-by-step explanations of convolutional image algorithms and LLMs, building from familiar mathematical concepts. This makes the videos uniquely suited for classroom use and aligned with the Dutch mathematics curriculum. Mathness creators Fréderique Kool and Jason Zijlstra are experienced high school teachers, ensuring alignment with the target audience.
This SEI started on 01-07-2026 and aims to reach its goals before 01-07-2027.
Goals
The goal of this project is to expand the abovementioned series with more advanced topics, taking the mathematics a step further (matrices, convolution, mathematics behind LLMs). This will be done by creating three additional videos to complete the series:
- How do computers see images?
- How do neural networks recognize patterns?
- How do computers read and generate words (LLM)?
These videos will be freely available and suitable for use in mathematics, ICT, science and NLT classes in Dutch high schools.
The topics addressed in these videos are highly relevant given their importance in everyday technologies. Short, engaging videos are an effective way to reach a broad audience of students, while requiring minimal preparation time for teachers. The videos will be disseminated via social media and the communication departments of the participating universities.
In the aforementioned 4TU.NIRICT initiative, we developed a syllabus, programming assignments, and teacher workshops on medical image analysis. Two of the three new videos directly connect to this material. We already have a network of around 40 mathematics teachers ready to pilot the videos and teaching materials. In addition, we will organize two workshops at the Nationale Wiskundedagen to spread the word to a wider community of teachers.
The videos can be used as standalone resources or integrated into a coherent lesson plan that connects students to STEM applications. Researchers from most 4TU institutions (Delft, Eindhoven, Twente, Groningen) have expressed interest in delivering guest lectures at nearby schools using our teaching material developed in the NIRICT grant. High school teachers can show our videos before such a visit, while guest lecturers use the material during their visit, minimizing preparation time for both.
Concrete outcomes:
- Three professional educational videos on AI and mathematics (images, patterns, words/LLMs).
- Integration into existing lesson packages via 4TU.Schools.
- Pilot testing with ~40 high school teachers.
- Two workshops at the Nationale Wiskundedagen to disseminate results.
- Guest lectures at secondary schools by 4TU researchers, supported by the videos and lesson material.
Participants
- Clara Stegehuis (University of Twente)
- Alexander Heinlein (Delft University of Technology)
- Fréderique Kool (Mathness)
- Jason Zijlstra (Mathness)