MathBuddy: Affective Math Tutoring — EMNLP 2025 Demo. An emotionally aware LLM-powered mathematics tutor that dynamically models student emotions from both conversational text and facial expressions, aggregating multimodal affective signals to shape pedagogically appropriate LLM responses. Maps detected affective states to relevant pedagogical strategies, achieving a +23-point win rate advantage and +3-point improvement on the DAMR (Dimensions of Affective Math Response) score at the overall level. Evaluated across eight pedagogical dimensions in user studies, MathBuddy demonstrates that affective awareness substantially improves the quality of AI math tutoring.
Authors: Kar et al. (ITU-NLP) · Venue: EMNLP 2025 Demo Track · Code: github.com/ITU-NLP/MathBuddy
Key Findings
MathBuddy addresses a critical gap in llm-math-tutoring: most AI math tutors respond to the content of student answers but ignore the emotional state of the learner. This is problematic because mathematics learning is deeply affective — frustration, confusion, anxiety, and confidence all shape how students engage with mathematical problems and whether they persist through difficulty.
Multimodal emotion modeling. MathBuddy captures student affect through two complementary channels: (1) conversational text analysis, which detects emotional signals in the linguistic content of student responses (e.g., expressions of frustration, confusion, or confidence), and (2) facial expression analysis, which captures moment-to-moment emotional reactions through computer vision. These modalities are aggregated to produce a unified affective state assessment that is more robust than either modality alone — a design principle consistent with the broader multimodal-ai-tutoring literature.
Affect-to-pedagogy mapping. The system maps detected affective states to specific instructional-design pedagogical strategies. For example, detected frustration may trigger scaffolded hint sequences or encouraging messages; detected confidence may prompt the tutor to increase problem difficulty or reduce scaffolding; detected confusion may trigger a shift from procedural guidance to conceptual explanation. This dynamic mapping connects MathBuddy to the affective-tutoring paradigm and the student-modeling tradition in intelligent-tutoring-systems.
LLM-powered response generation. At its core, MathBuddy uses an LLM prompted with the aggregated affective state, the mathematical context, and the tutoring history to generate emotionally appropriate pedagogical responses. The LLM is not merely an answer engine but an affective-pedagogical reasoning system that must balance mathematical correctness with emotional attunement.
Rigorous evaluation. The system was evaluated along eight pedagogical dimensions — including scaffolding quality, emotional attunement, mathematical accuracy, and learner engagement — through controlled user studies. The +23-point win rate and +3-point DAMR improvement establish clear evidence that affective awareness adds measurable value beyond content-only tutoring approaches. This connects to the engagement-metrics literature and the growing recognition that engagement-intensity-learner-modeling is essential for effective AI tutoring.
Implications
For the intelligent-tutoring field, MathBuddy strengthens the case that effective tutoring is fundamentally relational, not merely informational. The substantial performance gains from adding affective awareness suggest that affective-computing should be considered a core component of next-generation tutoring systems, not an optional enhancement. This aligns with findings from the multimodal-affective-its-presentation literature showing that affect-aware tutors outperform their affect-blind counterparts.
For k-12 mathematics education specifically, MathBuddy's multimodal approach to emotion detection offers a pathway toward AI tutors that can respond to the full range of student experiences — including math anxiety, which is a well-documented barrier to math-education achievement. The ability to detect and respond to negative affective states in real time could make AI tutoring particularly valuable for students who might otherwise disengage.
The dual-modality aggregation strategy (text + facial expression) also has implications for multimodal-learning-genai system design more broadly. MathBuddy demonstrates that combining complementary signal types produces more reliable affective assessments than either modality alone, a finding that generalizes beyond math tutoring to any educational context where learner emotions matter.
Methodologically, the DAMR evaluation framework and the eight-dimension pedagogical assessment provide a template for rigorous evaluation of affect-aware tutoring systems — addressing concerns about ai-tutor-behavioral-evaluation and the need for multi-dimensional quality metrics beyond simple accuracy measures.
Finally, the open-source release of the codebase (github.com/ITU-NLP/MathBuddy) contributes to the open-source educational AI ecosystem and enables further research into affect-aware pedagogical-llm-training.
Related Pages
- intelligent-tutoring-systems — The broader ITS context
- affective-computing — Affective computing in human-computer interaction
- affective-tutoring — Affect-aware tutoring systems
- llm-math-tutoring — LLMs applied to mathematics education
- k-12 — K-12 education context
- multimodal-ai-tutoring — Multimodal approaches to AI tutoring
- multimodal-affective-its-presentation — Multimodal affective ITS presentations
- student-modeling — Student modeling in ITS
- engagement-metrics — Measuring learner engagement
- pedagogical-llm-training — Training LLMs for pedagogical purposes
- math-education — Mathematics education