Logo image
Back
Designing for connection: how visual fidelity and animation quality in pedagogical conversational agents shape social presence, affinity, and the uncanny valley in virtual learning environments
Dissertation   Open access

Designing for connection: how visual fidelity and animation quality in pedagogical conversational agents shape social presence, affinity, and the uncanny valley in virtual learning environments

Eman Alharthi
Doctor of Philosophy (Ph.D.), Drexel University
Mar 2026
DOI:
https://doi.org/10.17918/00011302
pdf
Alharthi_Eman_20261.69 MBDownloadView
PDFOpen Access (License Unspecified),  Open Access

Abstract

Artificial intelligence Animation quality Social presence Uncanny valley Virtual agents Virtual learning Visual fidelity
Pedagogical, a large number of PCAs still use flat, two-dimensional models or static animations devoid of expressive lifelikeness, which limits their role as facilitators of instruction. This study analyses the impact of visual and animacy/frameworking on the dynamic social engagement of learners in virtual learning environments. It aims to determine the impact of social presence, social credibility, social affinity, and engagement, and their relationship with static and dynamic animations with cartoonish and realistic impressions. The analysis demonstrates that the PCA's movement and gesture behaviors, as well as lifelike, dynamic, and animated form, were the strongest predictors for social credibility and social presence perception. Social presence showed a moderate effect (p = 0.006, ges =0.087), and social credibility showed a large effect (p < 0.001, ges = 0.218) with a significant social presence effect. However, social presence showed a moderate effect and social credibility showed a large effect with significant cf. (p < 0.001, ges = 0.218). On the other hand, high visual fidelity did not affect social presence and social credibility unless paired with high animacy. Social facilitation, social affinity, and social engagement did not display significant effects across conditions (p>0.05). Social credibility pairwise comparisons showed that VF_High_Anim_High outperformed VF_High_Anim_Low (p = 0.000073), VF_Low_Anim_Low (p = 0.00082), and VF_High_Anim_Low versus VF_Low_Anim_High (p = 0.02164) significantly, indicating fluid, human-like behavior added to the perception of credibility more than visual realism. Likewise, for social presence, the only significant contrast was between VF_High_Anim_Low and VF_Low_Anim_High (p = 0.028), confirming the dominance of realism over fidelity. Results are consistent with the Media Richness Theory, the Social Response Theory, and CASA (Computers Are Social Actors), which all underline the impact social cues like time, motion, and gaze have on user interaction with technology. The data also supports the uncanny valley hypothesis, which suggests that comfort and immersion are negatively affected when behavior is highly realistic, but visuals are rendered in a mechanistic or robotic manner. Essentially, users rated PCAs--personally computerized assistants--higher when they acted in a natural manner, regardless of how cartoonish or low-fidelity their appearance was. On the other hand, agents that represented movement with great visual detail but acted rigidly were viewed as less credible and immersive. The conclusions drawn indicate that creators of virtual actors, social avatars, or educational social robots should focus on interaction realism, including human motion and smooth interactivity, rather than high fidelity visuals, to nourish and provoke meaningful educational trust and engagement. This research makes an important contribution to the literature aimed at improving the optimization of human-agent interaction in VLEs. The research results provide practical guidance for designing and improving the educational, vocational, and telepresence PCAs in educational, professional, and telepresence contexts. The study emphasizes critical engagement and credibility design issues by demonstrating the greater impact of behavioral realism compared to visual fidelity. Further research is needed on other aspects like voice quality, emotions and expressiveness, responsivity, and user retention over extended periods in different fields, demographics, and learning or work environments.

Metrics

1 Record Views

Details

Logo image