The use of avatars in education is becoming increasingly common, but research on their impact on learning and trust remains limited. This study investigates the perceived trust in a highly anthropomorphic AI-generated avatar in educational contexts and how it affects learning. We examined how awareness (informed group, n = 29) of the presenter in an educational video being an avatar affected learning compared to an uninformed group (n = 29), employing a mixed methods design (quantitative and qualitative). The groups were compared on (1) memory performance of the educational content (multiple-choice questions), (2) perceived trust of the avatar (Likert scale questions), and (3) emotional reactions (open-ended questions). The results revealed a significant group difference, the informed group performed lower in the memory test (U = 278.000 p = 0.026), but no significant group difference in perceived trust was shown, with both groups' responses being neutral or positive. However, the informed group exhibited a significantly less positive attitude toward receiving future educational content from avatars. The findings are discussed in relation to theories of implicit biases and divided attention, which may explain the lower scores within the informed group. The qualitative phase further enriched these results, indicating that awareness of the AI nature of the avatar led to more critical engagement with the content, potentially impeding learning. Future research should explore the roles of divided attention and implicit biases to better understand the conditions under which avatars impede learning.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:umu-225981 |
| Date | January 2024 |
| Creators | Westin, Fanny, Öberg, Amanda |
| Publisher | Umeå universitet, Institutionen för psykologi |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0028 seconds