EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Psychometric Evaluation of Multimodal AI Assessment of University Students' Well-Being

Xingyu Wang, Dongbo Che, Xutang Tong, Wah Cheung Wong

International Conference on AI Research · 18.11.2026 · 7 (1) · 547-551

DOI: 10.34190/icair.7.1.5690 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Conference on AI Research, online erschienen am 18.11.2026. Verfasst von Xingyu Wang und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Universities need credible evidence to support student well-being before difficulties disrupt learning. Self-report questionnaires remain essential, but fixed responses may omit context that students express in their own words. A multimodal model can process language, voice and video from a short interview, although agreement with a benchmark alone cannot establish that its outputs measure well-being. This work in progress sets out an evaluation of My Well-being My Zone, a voluntary, non-diagnostic system intended for private student reflection. The central question is whether its model-generated scores and feedback can be interpreted reliably and usefully for that purpose. We plan to recruit 1,300 university students to answer five recorded questions on self-awareness, emotions, stress and behavioural tendencies and to complete matched self-report measures. Independent experts will score responses and provide formative feedback using predefined criteria. Data from 1,000 students will support parameter-efficient fine-tuning of an open-source video-language model. In a separate 300-student hold-out sample, we will compare the fine-tuned model with its base model on identical inputs. We will examine score stability, generalisability across questions and raters, agreement with independent experts, theoretically expected relationships with self-report measures, and added agreement with experts beyond self-report. Independent raters will assess feedback relevance and actionability; subgroup analyses will examine uneven errors. A related 101-person employee pilot and a working prototype establish technical feasibility only. We expect fine-tuning to improve expert agreement and feedback quality, but student results are not yet available. The evaluation will distinguish agreement with expert ratings from evidence about the broader meaning or benefits of an assessment. By separating expert agreement, score meaning, feedback quality and subgroup errors, the study contributes an assessment-oriented validation protocol for multimodal AI in higher education, rather than treating predictive accuracy alone as evidence of well-being measurement. It will identify which claims the data support, where uncertainty remains and what further evidence would be needed before the system is used beyond voluntary developmental reflection.

Zitieren

APA

Wang, X., Che, D., Tong, X., & Wong, W. C. (2026). Psychometric Evaluation of Multimodal AI Assessment of University Students' Well-Being. International Conference on AI Research, 7(1), 547-551. https://doi.org/10.34190/icair.7.1.5690

Vancouver

Wang X, Che D, Tong X, Wong WC. Psychometric Evaluation of Multimodal AI Assessment of University Students' Well-Being. International Conference on AI Research. 2026;7(1):547-551. doi:10.34190/icair.7.1.5690

BibTeX
@article{Wang2026,
  author = {Wang, Xingyu and Che, Dongbo and Tong, Xutang and Wong, Wah Cheung},
  title = {Psychometric Evaluation of Multimodal AI Assessment of University Students' Well-Being},
  journal = {International Conference on AI Research},
  year = {2026},
  volume = {7},
  number = {1},
  pages = {547--551},
  doi = {10.34190/icair.7.1.5690}
}
Kostenlos starten →