Abstract:

The last two decades of research in higher education have seen a major shift to learning that takes place online, for example as augmentation of classroom activities at a university, or as a new paradigm for remote education delivery (e.g., MOOCs). This shift has introduced many new forms of teaching and learning. Yet, a substantial component of education in higher education institutions continues to be delivered in analog form as lectures. The persistence of lecture-based instruction motivates our work that seeks to foster active student engagement with such content.

Studies have questioned the efficacy of lecture-based instruction, primarily because lectures are widely considered to be passive and teacher-centric. To ameliorate passivity in learning, teachers may interleave lectures with interactive activities like student group work, peer instruction or subscribe to engagement techniques such as student polling or inviting online discussions. We situate our research in this multifaceted context for lecture-based learning.

As a starting point, we probe the research landscape in three areas- education, digitization of lectures, and the application of artificial intelligence (AI) methods in education. First, we draw broadly from the field of education to delve into learning theories and active engagement for learning. Next, we explore multimedia work done in lecture recording, video player interfaces for access to recorded materials and spoken content retrieval systems. Third, we seek to broadly understand how intelligence has been infused into educational technologies.

Digitization of lectures seeks to capture what happens in a traditional classroom and create interaction affordances that are similar to what is possible in online learning. For example, digitized lectures enable students to revisit material covered in class at their own pace while also minimizing the need for real-time note taking. Thus, students can proactively take control and regulate their own learning. An early focus on such lecture capture technologies was to streamline storage and facilitate easy access to lecture recordings.

Concurrently, progress in natural language processing has expanded what and how course-related data might be available to students, making it easier to interact with lecturers’ speech (what has been termed spoken content retrieval). Systems for spoken content retrieval typically rely on transcription of recorded speech and subsequent analysis of the resulting text. A primary focus has been on algorithms that support accurate speech transcription of instructor audio, efficient indexing, search and access. However, studies of spoken content retrieval have largely been isolated from the field of education and research on learner-centered interface designs.

In contrast, the field of learning analytics has emerged, with the goal of ‘tracking’ student behavior during interaction with educational videos to infer learners’ engagement. Even so, research has paid little attention to how content analytics (i.e., tools that are tied to learning from specific pieces of lecture content) could be used to help learning.

In our work, we delve deeply into advances in these various streams of work, but also surface the siloed nature of the contributions. We articulate the need to bring these areas closer together while envisioning and designing new educational applications that are grounded in the learning theories, human-computer interaction, and AI.

Keywords:

Higher education, engagement, lecture recordings.