Blue Carrot, in collaboration with GenEd, embarked on a journey to explore emerging innovations, assess their real-world applicability, and develop a production process to seamlessly integrate AI tools at every stage of the development of an e-learning course, starting from instructional design to scripting and video creation. The goal was to accelerate production speed, reduce costs, and minimize the need for Subject Matter Experts (SMEs), thus enhancing efficiency, resilience, and scalability across the entire workflow.
Due to the novelty of these technologies, the lack of standards, and their lack of prior experience within the market, the main challenges were as follows:
We began with an in-depth exploration of new technologies, focusing on understanding their limitations and potential applications. Our first step was to examine the capabilities and constraints of OpenAI’s large language model (LLM).м
The lack of subject matter experts is a challenge in developing quality online learning content. Using LLMs in learning design can help address this issue, speeding up instructional design and reducing the time needed from experts.
To validate this approach, GenEd conducted a study for which we developed two courses—one using traditional methods and the other leveraging new AI-driven technologies.
Both courses were blindly reviewed by experts using predefined evaluation criteria, with no indication of which course was created by which method. Blind expert reviews showed both met the same quality standards, but the AI-driven approach was 25 times faster.
The next step was researching how learners perceive educational videos featuring AI avatars. With 83 adult participants, results showed that while learners could easily distinguish AI avatars from real instructors, their presence had no impact on learning outcomes—suggesting initial skepticism is more about perception than effectiveness.
After thoroughly analyzing all the most uncertain aspects, we created a demonstration course on Decision-Driven Decision Making (DDDM) to showcase the practical application of all the tools in a real-world case.
Integrating AI tools at every production stage enabled us to achieve a course development process that is 25 times faster than traditional methods, all while maintaining the same quality. AI-driven instructional design significantly lessened our dependency on SMEs; synthetic videos allowed us to bypass live filming, enhancing efficiency, scalability, and cost-effectiveness. Remarkably, the entire lesson was created in less than a week.
The research results and demo project were presented at AIED 2023, a leading conference on AI in education held in Japan, where they were well received by the audience. Attendees viewed the study as innovative and pioneering, particularly for its practical approach to applying generative AI in learning design. One professor from a Japanese university approached the team afterward, calling the DDDM prototype a “notable step” and asking about potential future studies to build on it. Another attendee appreciated the side-by-side comparison of traditional and AI-driven workflows, commenting that it was “refreshing to see real-world applications, not just theory.” The presentation sparked interest and discussion about further exploring AI in education. Since then, the paper has been cited multiple times a week in studies worldwide, reflecting its ongoing relevance.
We will reply within 24 hours during the business week
We will reply within 24 hours during the business week
