Exploring and Analyzing the Effect of Avatar's Realism on Anxiety of English as Second Language (ESL) Speakers

TL;DR

The study addresses how avatar realism affects anxiety in ESL communication. It uses a mixed-methods laboratory design with Mandarin-speaking ESL participants exposed to cartoon-like, realistic-like, and live-vision representations, collecting self-reported anxiety and physiological signals (EDA, ECG, PPG). Key findings show that realism does not monotonically increase anxiety; live video yields the lowest self-reported anxiety, while realistic avatars induce the highest anxiety, and cartoon avatars can match live video in physiological arousal. Design implications argue for prioritizing clear social signaling and contextual appropriateness over realism, with live video as a robust default when privacy permits. The work provides practical guidance for avatar-mediated ESL interactions and demonstrates the value of combining subjective and physiological assessments to understand emotion in cross-cultural communication.

Abstract

Virtual avatars are increasingly used to support cross-cultural communication, yet their impact on communication anxiety among English as a Second Language (ESL) speakers remains underexplored. This study examines how avatar realism influences anxiety during English interactions between ESL speakers and native speakers. We conducted a controlled laboratory study in which Mandarin-speaking ESL participants engaged in guided one-on-one conversations under three visual representation conditions: live video, cartoon-like avatars, and realistic-like avatars. Anxiety was assessed using self-reported surveys and physiological signals, including electrodermal activity (EDA), electrocardiography (ECG), and photoplethysmography (PPG). The results show that increased visual realism does not correspond to a monotonic change in anxiety. Live video was the most preferred and was associated with the lowest self-reported anxiety. Cartoon-like avatars exhibited physiological anxiety levels comparable to live video and lower than realistic-like avatars, whereas realistic-like avatars elicited elevated anxiety across measures. These findings suggest that an effective avatar design for ESL communication should prioritize clarity of social signaling, reduced perceived social threat, and alignment between visual representation and interaction context, rather than visual realism alone.

Figures (5)

• Figure 1: Cartoon-like and realistic-like avatars used in the study. Both avatar styles were custom-designed to reflect the facial features of the corresponding native English speaker.
• Figure 2: Avatar control setup during the study. The left panels show the rendered avatars viewed by participants, including cartoon-like avatars (left column) and realistic-like avatars (right column). The top row illustrates a smiling expression, while the bottom row illustrates a mouth-opening expression. The corresponding live video feed used for motion and facial tracking is shown on the right for illustration purposes but was not visible to participants.
• Figure 3: Participant Flowchart for the Entire User Study.
• Figure 4: Detailed participant flow during the in-person intervention and post-intervention assessment phases.
• Figure 5: Distributions of NR, maxR, MH, and MA features when participants were in the control group or interacting with cartoonish avatars, realistic-like avatars, or live videos. Dotted lines marked with an asterisk (*) highlight significant differences between paired groups (p < 0.05).