Context & Scale
This pattern addresses the generation and implementation of realistic AI avatars for educational video content. It is particularly relevant in higher education contexts where video scripts or content have multiple authors and present core disciplinary concepts and case studies that may be easily edited and updated. Avatar-led videos can facilitate sharing in multiple accents and languages for large, diverse cohorts (Vallis et al., 2024).
Problem
The use of realistic AI-generated avatars for educational videos is an emerging area in higher education (Vallis & Britton, 2024). Educational institutions may seek ways to present content in various accents and languages to cater to diverse student populations and increase accessibility. Streamlining video content production, especially for materials that require frequent updates or multiple versions, may also be a motivator.
Hence educators need guidance on when and how to implement AI-generated avatars effectively and ethically, as such video content may be misleading and deceptive (Vaccari & Chadwick, 2020). Such AI-generated content needs to be considered in the context of the overall learning design. Long periods of viewing AI-generated video presenters is less engaging so is unsuitable for recorded lectures (Vallis et al., 2024). Also, educational videos and materials are best generated in an intentional and ethical manner that is transparent and trustworthy (Khosravi et al., 2022).
Solution
To address these challenges, this pattern proposes an approach that includes a decision-making framework for educators to determine when AI-generated avatars for educational videos are appropriate. Then a process for creating and implementing such content is outlined to assist the production process and ongoing evaluation. Students and educators are encouraged to critically consider this technology and their ethical decision-making by using generative multimodal AI technology into learning and teaching activities.
Implementation
Assess the purpose for using AI-generated avatars. If the purpose of the video is to share personal stories, give a sense of teacher presence online, or to create improvised, temporary content, then traditional video formats may be better suited. Contexts where AI-generated video may be more effectively produced include:
- Concept videos that are generic and benefit from updated examples
- Short, scripted presentations of general content (less than 10 minutes)
- Content presented in multiple languages or accents, or where multiple voices and a team of teachers are involved
- To complement the curriculum at strategic points (e.g., depending on context, once per week? Or once per semester?)
Prepare the content. Write a short script for the avatars. Make the sentences as short, clear and conversational as possible. Long sentences with jargon sound unnatural and unclear.
Select and set up the AI avatar software. Individual users may trial free software. Obtaining a licence and establishing a process may have long-term benefits for scale and allow more features such as choosing diverse avatars and accents to deliver the same content, to appeal to different student cohorts.
Generate the video content by entering the script into the software. Where feasible, create multiple versions of the same content with different avatars and accents. Provide links or present content to student on a webpage so they may easily choose the version that suits them best.
Review the video to see how it flows. Tweak the script and regenerate the AI presentation. Iteration will improve the output. Adding relevant diagrams and supporting text to reinforce key points is helpful.
Be transparent when embedding the video in a learning activity or website. Explain why AI-generated content is being used and clearly indicate its use within the content and surrounding context. Use labels to identify AI-generated content.
Ask students to think critically about technology. Ideally, create activities where students evaluate and reflect on technology and AI use in education and society. For example, discuss the experience with students, their perceptions and its ethical implications. Alternatively, students might co-create their own avatars in groups, where appropriate.
Evaluate and iterate. Gather feedback from students and teaching staff to assess the effectiveness of the AI-generated content. Refine future iterations based on these insights.
Examples of pattern in use
Presenting business ethics in a Business Information Systems subject
Acknowledgments
This pattern was tested in a second-year postgraduate core subject, INFS6018 Managing with Information and Data. Since 2018, the cohort has grown from 60 to more than 1000 students in 2021.
We acknowledge the coordinators and lecturers involved from the Business Information discipline, including Associate Professor Daniel Gozman and Professor John Buchanan in conjunction with a committed teaching team. The approach was iteratively developed and implemented with these business academics, the media team led by Boyd Britton, and with learning design support from Benedicte Rokvic.
Description
In this subject, students learn a business rather than technology perspective of information systems, in the context of strategic and operational analytics and decision making. Higher order, critical thinking and reflective practice skills are encouraged (Gibbs, 1988).
In 2022 AI-generated avatars, along with interactive materials, were designed and developed in partnership with the Business Co-design team to immerse students in the ethics of business intelligence. AI-generated videos were used as both content delivery tools and examples of technology to be critically examined. Recorded lectures were redesigned into shorter chunks and diagrams were developed to enrich the content. The videos were embedded into a self-paced online module on Canvas LMS, with relevant activities and links to ethical decision-making in AI.
Student Feedback
Four student focus groups were conducted over two semesters in 2022 and 2023. Student feedback indicated that the AI-generated content was acceptable, even preferred, in specific contexts. However, students feedback indicated that initial awareness that the video was generated with AI was key. Some students found it highly engaging, relevant and effective in increasing their awareness of potential ethical issues in future business and society. Also, students did not want it to completely replace human interaction. didn’t want it to replace all human interaction.
Student focus groups conducted over two semesters in 2022 and 2023 indicated that AI-generated content was acceptable and sometimes preferred in specific contexts. However, students emphasised the importance of transparency about the AI nature of the content and did not want it to completely replace human interaction.
In summary, we found the following design principles essential for student participation in ethical AI avatar use in education:
- Avatar(s) fit learning design and purpose
- Students are aware of the use of AI
- Both human and nonhuman interactions
See figures below from Vallis et al., 2024.
Leave a Reply