Do Student Emotions Influence Learning During Virtual Field Trips? A Computer Vision Analysis of Emotion States in Desktop-based Virtual Reality
Accepted manuscript for the ASCE International Conference on Computing in Civil Engineering (i3CE 2026). Conference proceedings forthcoming.
Ryan T. Woo, Foysal Ahmed, Parth Bhadaniya, Ricardo Eiris, Masoud Gheisari
Abstract
Student emotions shape learning by influencing attention, engagement, and self-regulation. Virtual field trips (VFTs) in desktop virtual reality can elicit affect, yet the emotion-learning relationship in construction VFTs remains understudied. This paper tests whether facial-expression-derived emotion metrics during a drywall-focused, instructor-guided VFT relate to learning gains. Webcam video was analyzed with a computer-vision pipeline (S3FD face detection and EmoNet classification) to label five expressions (neutral, happy, surprise, fear, sad). We computed expression-duration proportions and an emotion switch rate and related them to pre-/post-test improvement. After a face-visibility quality filter, 16 participants were analyzed. Negative expression proportion (fear + sad) correlated with learning gains (r = 0.532, p = 0.034, R² = 0.283), as did switch rate (r = 0.506, p = 0.046, R² = 0.256), while neutral proportion showed a nearly significant marginal negative association (r = −0.484, p = 0.058, f² = 0.306).
Keywords: Virtual field trips; construction education; affective computing; facial expression recognition; computer vision