PRIMA-KI: Pproblem-solving learning im Mathematic teaching in primary school with AI accompany.
Didactically sound AI learning support for problem-oriented mathematics learning in primary school
A research project to develop and investigate an adaptive tutoring system for individualized learning in mathematical discovery in primary school.
About the project
As part of a project at the Weingarten University of Teacher Education, we are investigating the potential and design options of AI-supported learning companions for primary school mathematics lessons. Artificial intelligence offers unique opportunities for individualized, adaptive feedback in problem-based learning—provided it is designed and implemented with a sound didactic approach.
The PRIMA-KI pilot project develops and researches prototype AI learning companions integrated into math apps that support children in exploring and solving mathematical problems. The focus is on cognitive activation and supporting independent thinking processes.
Project goals
- Development subject-specific didactic design principles for AI learning companions in primary school
- Development of optimally structured prompts and interaction designs for embedding in existing digital learning offerings for child-friendly, didactically meaningful adaptive support
- Identifying effective feedback strategies that reduce cognitive load and support children without replacing the learning process
- Evaluation of usage behavior and acceptance among primary school children and determination of the optimal degree of pedagogical pre-structuring of AI use for primary school children
- Creating an evidence-based foundation for the responsible and data protection-sensitive integration of AI in primary school mathematics education. Aspects of using AI with minimal or anonymized data are part of the research project.
Current status and planned approach
In the current pilot phase, we are focusing on developing and testing initial prototypes in controlled settings. Through systematic variations and optimization of the AI prompt designs, we are developing design guidelines and a framework for adaptive support functions based on the integration of AI learning companions into various apps by Christian Urff. Following individual trials, we aim to conduct the first more extensive field trials in 2026. We welcome schools or classes with iPads and Wi-Fi access who would like to try out the initial trials under scientific supervision. The apps will then be made available free of charge via TextFlight.
Examples and initial approaches
The Math stories uses AI to help children understand, work on, test and develop modeling tasks.
In the app number line An AI-powered learning companion provides individualized support on request or in case of errors, based on subject-specific pedagogical knowledge. This research investigates whether and how children use this detailed error feedback to learn from mistakes and improve their number-location tasks. Anonymized learning data can be shared within the app for scientific analysis.
In the app finger amounts The project investigates the support of AI in the analysis and processing of diagnostic data. AI analyzes (anonymously) strategies, flash memory times, etc., and generates short, helpful summaries for teachers to use in further support.
In the web apps Calculation Field An AI learning companion based on Elevenlabs accompanies the processing on the computer field and is being tested for this purpose.
…
Contact and collaboration
If you're interested in the project and would like to collaborate or participate as a school or researcher, I'd be delighted to hear from you!
christian.urff@ph-weingarten.de