Is anybody watching: A multi-factor motivational framework for educational video engagement

Videos are one of the most widely used teaching and learning modalities and form the backbone of many online courses. Therefore, understanding what influences video-watching behavior has been an area of intense interest to educators and course designers. Exploration of video engagement in online courses has focused primarily on factors intrinsic to each video, such as video length and production style, with less exploration of the effects of the course context. Furthermore, the prevailing narrative that emphasizes keeping videos relatively short (under six minutes in length) stems from studies looking at engagement during individual viewing sessions for each video, rather than total engagement measured over all of a student’s viewing sessions for a given video. Using detailed viewing data from 3.1M video views in a variety of online biomedical science courses, we demonstrate that students regularly engage with educational videos over multiple viewing sessions. This behavior causes single-viewing metrics to significantly underestimate true engagement, and indicates that the commonly accepted wisdom of limiting videos to only 6 minutes in length is not well-grounded in learners’ actual behavior. We also show that elements of course context are strongly associated with video engagement. Such elements include the number of graded assessments associated with each video, a video’s position within a course, and requiring video viewing in order to obtain desirable outcomes, such as certificates. Based on the data, we create linear models that are able to predict whether a video will garner high or low engagement across single and multiple viewing sessions. These models are significantly more accurate than models that rely on video length alone, further substantiating the importance of course context. These results suggest several factors that educators and course designers can utilize in video design and course development to increase video engagement. • This study investigated educational video engagement using 3.1M video views. • Analysis shows students regularly engage with individual educational videos over multiple viewing sessions. • Single-viewing session metrics alone significantly underestimate true engagement. • Elements of video design and course context are strongly associated with video engagement. • We create accurate interpretable models for prediction of key aspects of video engagement.