2026 ASEE Annual Conference & Exposition

The use of videos in teaching robotics is known important to facilitate learning of robotics. Video-based learning has been known crucial for active learning and flipped classrooms-centered robotics instruction. Embodiment learning is known effective in helping students to better engage with the underlying learning subjects. Robotics instruction video involving embodiment learning features hands-on focused robotics instruction including gesture, eye contact, writing, manipulating, touch, hearing, and sensorimotor cues. However, the large number of online videos have made it challenging for instructors to select effective embodied learning videos to realize effective video-centered teaching. Furthermore, it is unclear for the correlation between the content of the videos and student's preference of the videos to determine whether the videos are suitable for robotics instruction.

Through the research, both conventional robotics instruction tutorial videos and embodied learning centered videos are used to understand their impacts on assisting learning of robotics. The research has studied the effectiveness of the videos in assisting learning of robotics with and without participants. Without participants, we have performed video user comments data analysis to understand the correlation between video user comments and video content. For the user study, we have conducted correlation analysis between users' comments and the video content.

In total, we have selected 20 Youtube videos with and without embodied learning components randomly, where the comments of the videos were also obtained. Overall, there are 3959 user comments for the 10 videos with embodied learning components. In contrast, there are only 2753 user comments for the videos without embodied learning components.

To understand the impacts of the video content and how it is correlated with the learning of robotics, the correlation between video frames and video user comments were obtained using large language models. Specifically, large language models including LLAVA and MiniCPM were used to extract the captions of the video frames, where Llama was used to perform the correlation analysis between user comments and the captions. The correlation between the video frame captions and user comments of the videos were subsequently obtained. In more details, we have also classified the video user comments to positive, negative, and neutral to understand the user perception and reaction toward the robotics tutorial videos.

We have involved 20 participants for learning robotics technology with the videos. In total, participants have watched 130 videos including 74 embodied videos and 56 conventional videos, Students were also involved in commenting on the videos for providing the reflection of the videos that students have learned. In total, we have gathered 130 comments for conventional videos and 167 comments for embodied videos.

Results showed that for the embodied learning videos, on average there are 295 user comments per embodiment learning video, whereas there are 275 user comments only per conventional video. For the embodied learning videos, the ratio between correlated user comments and overall video comments is 0.67, which is larger than the 0.46 of conventional videos. The positive comments of the embodied type videos are two times higher than the conventional videos. It therefore suggested that the use of embodied learning is effective in engaging user in learning robotics. For the user study, the correlation analysis between video content and user comments was found significant nearly through all the video frames. It therefore suggests that the video content regardless leveraging of either embodied or conventional instructional method is crucial for learning robotics technology. The result emphasizes the importance of the video content design in assisting learning of robotics.