Abstract:

Teaching and learning research in general, and mathematics education research in particular, has increasingly focused on questions relating to the use of artificial intelligence (AI) in the context of learning (e.g. Engelbrecht et al., 2025). However, research projects that address the needs of a diverse student body within inclusive subject-specific teaching remain scarce. The need for such research is underscored by the emergence of an AI divide. This refers to an inequality among students — one intertwined with broader social inequality — encompassing differential access to AI technologies, varying competencies in understanding, applying, and critically evaluating these technologies, and the unequal benefits derived from their use (Carter et al., 2020). Thus, the unreflective integration of AI into subject-specific learning risks reproducing and entrenching social inequalities within the education system.

Realising a difference-aware, subject-specific education requires balancing three interconnected demands: providing all students with high-quality subject-specific learning processes; acknowledging each student's individual potential and tailoring learning opportunities accordingly; and critically reflecting on social difference within the respective learning settings.

This paper takes up this idea and presents the results of the project “AI-Enhanced Cooperative Learning in Mathematics Education: Student-AI Collaboration Through Specialised AI Agents (AICOOP-MATH)”. It thus focuses on mathematics learning as an exemplary subject-specific specification of learning in school. Adopting an interactionist perspective on mathematics learning, according to which mathematical knowledge is constructed through reciprocal meaning-making in social interactions (Salle & Schütte, 2023; Jung et al., 2022), the fundamental premise emerges that individualised learning opportunities should invariably incorporate cooperative learning formats (Johnson & Johnson, 1999).

Building on this, the project investigates the possibilities of AI-supported cooperative mathematics learning among pupils with heterogeneous learning prerequisites. To this end, a qualitative-interpretive research design is employed to examine how pupils with and without special educational needs (N = 50) collaboratively engage with an AI bot as a learning partner to solve substantive mathematical problems. The resulting interactive work phases are videotaped and transcribed. The text output of the AI bot is incorporated into the transcripts analogously to the contributions of a human interaction partner. The data are analysed by means of interaction analysis (Schütte et al., 2019). The findings are then critically examined in light of difference-theoretical considerations. On this basis, the limitations and potentials of AI-supported cooperative mathematics learning for a difference-aware mathematics pedagogy are discussed.

Preliminary analyses reveal that the tandem groups with heterogeneous performance levels differ with respect to:
(1) the distribution of responsibility among the pupils within the cooperative process;
(2) the integration of the chatbot into that process;
(3) the focus of the cooperative mathematical activity; and
(4) the influence of heterogeneous prior experience on the cooperative mathematical process.

Keywords:

Inclusive Education, Mathematics Education, AI-Enhanced Learning, Cooperative Learning.