PaperarXiv:2609.16077

Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups

Preprint2026Added Oct 30, 2026

Authors: Meisam Jamshidi Seikavandi, Tanya Ignatenko, Fabricio Batista Narcizo, Paolo Burelli, Jesper Bünsow Boldt, Andrew Burke Dittberner

arXiv preprint (cs.LG)

PDF arXiv2609.16077 Cite

TL;DR

This paper studies sparse self-report labels in GroupAffect-4, a dataset of 40 participants in 10 four-person teams doing collaborative tasks while wearing EmotiBit wrist sensors (PPG, EDA, skin temperature, IMU at about 25 Hz) and Tobii G3 eye trackers, with Big Five personality and post-task valence, arousal and dominance (VAD) ratings. Each 15-second window gets a 49-feature physiology plus gaze/pupil vector; VAD ratings are mapped to three classes with train-fitted thresholds and scored with macro-F1. Gaussian-process pseudo-labels lift a known-team RBF-SVM from 0.570 / 0.530 / 0.385 (V/A/D) to 0.608 valence and 0.435 dominance (0.546 arousal with a simpler variant).

Abstract

Physiological affect sensing in naturalistic group interaction is often limited by sparse labels rather than sensor data: wearable devices produce many time windows, while self-reports are collected only a few times per session. Using GroupAffect-4, a four-person collaborative dataset with wearable physiology, eye tracking, Big Five personality, and post-task VAD labels, we study pseudo-label augmentation for affect sensing under sparse supervision. We compare no augmentation, Gaussian Process pseudo-labelling, personality-aware trust weighting, and joint personality-plus-confidence weighting within a shared target-construction pipeline. Results show that pseudo-label augmentation improves over the labelled-only baseline in the known-team setting. However, the narrow range of Big Five cosine similarities (0.91-0.99) makes fine-grained personality weighting ineffective; personality similarity functions mainly as a same-team filter rather than a calibrated trust signal. With smoothing, augmented SVM variants are effectively tied on Valence and Arousal, while the joint personality-plus-confidence variant gives the highest Dominance score. Cross-subject LOSO transfer remains encouraging, especially for Arousal, whereas strict session-isolated LOGO removes the augmentation benefit. Given only 10 groups, LOGO should be interpreted as a conservative lower bound on unseen-group transfer. Overall, the results suggest that pseudo-label augmentation can make better use of sparsely labelled collaborative affect data, while personality information is most useful as a within-team selection mechanism.

Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)

Results on ChatPPG leaderboards · 1

All leaderboards

Leaderboard	Method	Result	Rank
GroupAffect-4 macro-F1Emotion recognition (valence/arousal)	Pseudo-Label Augmentation for Affect Sensing in…	0.608macro-F1, higher is better	#1 of 1

1 row is credited to this paper. Ranks apply within each leaderboard only, as reported by the source.

Tasks & signals

Keywords

EmotionValenceArousalDominanceGroup AffectPseudo LabelsEmotiBit

Datasets used

GroupAffect-4EmotiBit multimodal four-person collaboration40 subjects · PPG · IMU · EDA

How it works

Cite this paper

Cite this paper with the fields recorded here. Check it against the publisher before submitting.

Seikavandi, M. J., Ignatenko, T., Narcizo, F. B., Burelli, P., Boldt, J. B., & Dittberner, A. B. (2026). Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups. arXiv preprint. https://arxiv.org/abs/2609.16077

Seikavandi, Meisam Jamshidi, Tanya Ignatenko, Fabricio Batista Narcizo, Paolo Burelli, Jesper Bünsow Boldt, and Andrew Burke Dittberner. 2026. "Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups." arXiv preprint. https://arxiv.org/abs/2609.16077.

Seikavandi, M. J., Ignatenko, T., Narcizo, F. B., Burelli, P., Boldt, J. B. and Dittberner, A. B. (2026) 'Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups', arXiv preprint. Available at: https://arxiv.org/abs/2609.16077

Seikavandi, Meisam Jamshidi, et al. "Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups." arXiv preprint, 2026. https://arxiv.org/abs/2609.16077.

M. J. Seikavandi, T. Ignatenko, F. B. Narcizo, P. Burelli, J. B. Boldt, and A. B. Dittberner, "Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups," arXiv preprint, 2026, [Online]. Available: https://arxiv.org/abs/2609.16077.

@misc{seikavandi2026pseudolabel,
  title         = {{Pseudo-Label Augmentation for Affect Sensing in Small Collaborative Groups}},
  author        = {Seikavandi, Meisam Jamshidi and Ignatenko, Tanya and Narcizo, Fabricio Batista and Burelli, Paolo and Boldt, Jesper Bünsow and Dittberner, Andrew Burke},
  year          = {2026},
  howpublished  = {arXiv preprint},
  eprint        = {2609.16077},
  archivePrefix = {arXiv}
}

Oct 30, 2026Added to the ChatPPG research hub