MathBuddy's multimodal emotion aggregation matched participant-reviewed emotions at 60% overall accuracy, with sharply reduced recall for the Neutral class¶
Claim · All claims
Evidence · 1 study · designr2·q2
Currency · newest study 2025 · not yet checked against its sources · not reviewed by a person · recheck by 2028: tool and model accuracy changes quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? Compared against participants' gold annotations of their own emotions, the aggregated text-plus-face emotion detection reached 60% overall accuracy, with Neutral recall falling to 0.15. → Debanjana Kar 2025
Evidence¶
Debanjana Kar 2025¶
Debanjana Kar, Leopold Böss, Dacia Braca, Sebastian Dennerlein, Nina Hubig, Philipp Wintersberger, and Yufang Hou. (2025). MathBuddy: A Multimodal System for Affective Math Tutoring. https://arxiv.org/abs/2508.19993 doi:10.48550/arxiv.2508.19993
q2 · i? · design · r2
User-study participants reviewed and corrected the system's predicted emotions in Emotion ON conversations; aggregated multimodal predictions were compared to these gold annotations (Table 5: Negative precision 0.57/recall 0.98; Neutral 0.71/0.15; Positive 0.79/0.41).
"the system achieves an overall accuracy of 60% in real-time usage. However, we observe a sharp fall in the recall scores for theNeutralclass"
Discussion¶
Related Claims¶
- In a within-subject user study, participants showed longer positive facial expressions under the emotion-aware (Emotion ON) tutoring condition — related
- Emotion features improve tutor performance only when grounded in education-theory-based pedagogical strategies; emotion information alone misguides a smaller model — related
- Text-based emotional expression in student responses showed no statistically significant differences between Emotion ON and OFF conditions — related
- Adding student-emotion information to LLM tutor prompting produces more desirable pedagogical qualities across most evaluated dimensions, with larger models benefiting most — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
multimodal-emotion-aggregation-60-percent-accuracy |
title |
MathBuddy's multimodal emotion aggregation matched participant-reviewed emotions at 60% overall accuracy, with sharply reduced recall for the Neutral class |
description |
MathBuddy's multimodal emotion aggregation matched participant-reviewed emotions at 60% overall accuracy, with sharply reduced recall for the Neutral class |
status |
review |
evidence_strength |
weak |
newest_evidence |
2025 |
recheck_by |
2028 |
generated |
by process:wiki-ingest on 2026-10-11 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — debanjana-kar-2025 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.