Interaction Effects Between Learner Characteristics and Dialogue Format in TTS Dialogue-Based Lessons
This study of 222 high school students reveals that learner characteristics, specifically experiential learning styles and critical thinking dispositions, significantly interact with dialogue formats (teacher-student, student-student, and teacher-teacher) to influence motivation and learning outcomes in TTS-based lessons, suggesting that personalized dialogue selection is beneficial despite the preliminary nature of the small-to-medium effect sizes.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In the rapidly evolving landscape of education, a new question has emerged: does the way a lesson is delivered matter as much as the lesson itself? For decades, educators have understood that students learn in different ways. Some thrive when they can touch, see, and experience a concept directly, while others prefer to step back, reflect, and organize ideas before acting on them. This difference in how people process experience is known as experiential learning style. At the same time, the rise of artificial intelligence has made it possible to create educational videos where computers generate the scripts and voices, allowing for a variety of conversation styles that were previously too expensive or difficult to produce. The central puzzle for researchers is whether these new, computer-generated formats work equally well for everyone, or if a specific style of dialogue might spark more interest and understanding in some students while leaving others behind.
A team of researchers set out to solve this puzzle by testing how high school students reacted to three different types of computer-generated lessons. They created videos using advanced language models to write the scripts and text-to-speech technology to provide the voices. The first format featured a traditional teacher explaining concepts to a student. The second showed two students talking to each other, sharing what they knew. The third format was unique: two experts, or "teachers," discussing the subject matter with one another, without a student character present to ask questions. These lessons covered data science topics and were watched by 222 first-year high school students. After watching each video, the students answered questions about how motivated they felt, how well they thought they understood the material, and their overall opinion of the lesson. The researchers also measured each student's natural learning style and their tendency to think critically about information.
The results revealed a fascinating mismatch between what made students feel excited and what they thought was a good lesson. When the students watched the video featuring two experts talking to each other, those who preferred concrete, hands-on experiences reported feeling significantly more motivated than they did during the other formats. For these students, hearing experts debate and refine ideas seemed to act as a direct source of knowledge that sparked their interest. However, this same format did not work as well for students who preferred to reflect and conceptualize before acting; for them, the expert-to-expert dialogue actually felt less engaging than the traditional teacher-student lesson. This suggests that there is no single "best" way to present a computer-generated lesson; instead, the most effective format depends on the specific learning style of the viewer.
Despite the boost in motivation for certain students, the expert-to-expert format received a lower overall rating than the traditional teacher-student format. Students found the content of the expert conversation more difficult to follow and noted that the voices sounded stiff or unnatural. The researchers suggest that while the complex dialogue was stimulating for some, it created a mental burden that made the lesson feel harder to grasp. Interestingly, the students who watched the two-student conversation format also found the voices somewhat unnatural, likely because the computer voices did not quite match the casual tone of a real high school student.
The study concludes that while artificial intelligence can generate engaging educational content, the design of that content must be tailored to the learner. If a student learns best through direct experience, a dialogue between experts might be the most motivating choice. If a student learns best through reflection, a traditional teacher-student or peer-to-peer conversation is likely to be more effective. The researchers caution that these findings are preliminary, as the study was conducted in a single day with a specific group of students, and the lessons used different content for each format. Nevertheless, the work provides a clear path forward for personalized learning: by understanding how a student processes information, educators and software designers can choose the dialogue format that will keep that specific student engaged and ready to learn.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.