From script to screen

An emergent view of AI-generated avatars

Abstract

In an age of ‘deepfakes’, designing learning media and video intentionally and ethically becomes more important than ever. Emerging educational research focuses on generative large language models, which use neural networks to produce human-like text based on prompts. Yet the educational design of synthetic media, particularly AI-generated avatars, which are also currently produced from text prompts, remains less explored. This paper share insights from an ongoing evaluation of realistic, AI-generated avatars as educational presenters at a large Australian metropolitan university. Using commercially available software, we created a suite of AI-generated avatars and accompanying interactives to explore the topic of ethics in business intelligence which was delivered to approximately 1200 students across two semesters in 2022 and 2023. The aim was to provoke critical discussion around ethical decision-making with AI by immersing students in text-to-video technology as part of their learning experience. To better understand students’ perceptions of synthetic media as a stimulus for learning, we conducted four focus groups over two iterations of the course. Our multidisciplinary team also documented the design, development, and implementation challenges of AI-generated avatars. Preliminary findings suggest that clarifying design intentions is key to the effective and ethical application of AI-generated avatars. Automating video content with synthetic media to simplify certain types of educational content could be a catalyst for increased sharing and collaboration. While traditional video formats may prove a simpler choice in the short-term, this study indicates the potential of more hybrid human and nonhuman representation in certain contexts. Students were comfortable with synthetic media use and blending humans and nonhuman elements, expressing a desire for greater interaction with both. We propose a posthuman perspective on AI-generated avatars, one which acknowledges the messy and entangled nature of learning and teaching with technology, that often surfaces issues of ethics and power. To assist educators, ‘VIEW’ is outlined, a brief educational design guide for AI-generated avatars that considers aspects such as the intentions or purpose of the video, the suitability of video as a medium, implementation, and ethical questions. Finally, we suggest more critical and ethical studies on the emergence of synthetic media applications in higher education are needed.