skip to main content


Title: Synchronized eye movements predict test scores in online video education
Experienced teachers pay close attention to their students, adjusting their teaching when students seem lost. This dynamic interaction is missing in online education. We hypothesized that attentive students follow videos similarly with their eyes. Thus, attention to instructional videos could be assessed remotely by tracking eye movements. Here we show that intersubject correlation of eye movements during video presentation is substantially higher for attentive students and that synchronized eye movements are predictive of individual test scores on the material presented in the video. These findings replicate for videos in a variety of production styles, for incidental and intentional learning and for recall and comprehension questions alike. We reproduce the result using standard web cameras to capture eye movements in a classroom setting and with over 1,000 participants at home without the need to transmit user data. Our results suggest that online education could be made adaptive to a student’s level of attention in real time.  more » « less
Award ID(s):
1660548
NSF-PAR ID:
10211957
Author(s) / Creator(s):
; ; ; ;
Date Published:
Journal Name:
Proceedings of the National Academy of Sciences
Volume:
118
Issue:
5
ISSN:
0027-8424
Page Range / eLocation ID:
e2016980118
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. This research paper describes a study of elementary teacher learning in an online graduate program in engineering education for in-service teachers. While the existing research on teachers in engineering focuses on their disciplinary understandings and beliefs (Hsu, Cardella, & Purzer, 2011; Martin, et al., 2015; Nadelson, et al., 2015; Van Haneghan, et al., 2015), there is increasing attention to teachers' pedagogy in engineering (Capobianco, Delisi, & Radloff, 2018). In our work, we study teachers' pedagogical sense-making and reflection, which, we argue, is critical for teaching engineering design. This study takes place in [blinded] program, in which teachers take four graduate courses over fifteen months. The program was designed to help teachers not only learn engineering content, but also shift their thinking and practice to be more responsive to their students. Two courses focus on pedagogy, including what it means to learn engineering and instructional approaches to support this learning. These courses consist of four main elements, in which teachers: 1) Read data-rich engineering education articles to reflect on learning engineering; 2) Participate in online video clubs, looking at classroom videos of students’ engineering and commenting on what they notice; 3) Conduct interviews with learners about the mechanism of a pull-back car; and 4) Plan and teach engineering lessons, collecting and analyzing video from their classrooms. In the context of this program, we ask: what stances do teachers take toward learning and teaching engineering design? What shifts do we observe in their stances? We interviewed teachers at the start of the program and after each course. In addition to reflecting on their learning and teaching, teachers watched videos of students’ engineering and discussed what they saw as relevant for teaching engineering. We informally compared summaries from previous interviews to get a sense of changes in how participants talked about engineering, how they approached teaching engineering, and what they noticed in classroom videos. Through this process, we identified one teacher to focus on for this paper: Alma is a veteran 3rd-5th grade science teacher in a rural, racially-diverse public school in the southeastern region of the US. We then developed content logs of Alma's interviews and identified emergent themes. To refine these themes, we looked for confirming and disconfirming evidence in the interviews and in her coursework in the program. We coded each interview for these themes and developed analytic memos, highlighting where we saw variability and stability in her stances and comparing across interviews to describe shifts in Alma's reasoning. It was at this stage that we narrowed our focus to her stances toward the engineering design process (EDP). In this paper, we describe and illustrate shifts we observed in Alma's reasoning, arguing that she exhibited dramatic shifts in her stances toward teaching and learning the EDP. At the start of the program, she was stable in treating the EDP as a series of linear steps that students and engineers progress through. After engaging and reflecting on her own engineering in the first course, she started to express a more fluid stance when talking more abstractly about the EDP but continued to take it up as a linear process in her classroom teaching. By the end of the program, Alma exhibited a growing stability across contexts in her stance toward the EDP as a fluid set of overlapping practices that students and engineers could engage in. 
    more » « less
  2. Growing interest in “flipped” classrooms has made video lessons an increasingly prominent component of post-secondary mathematics curricula. However, relatively little is known about how students watch and learn from instructional videos. We describe and use an eye-tracking methodology to investigate attentive fidelity—the degree to which students attend to the visual imagery that is the subject of the video narration at each moment in time. Our preliminary study suggests that students’ attentive fidelity varies widely, but there was no evidence that this fidelity is connected to students’ ability to solve calculus problems. 
    more » « less
  3. Growing interest in “flipped” classrooms has made video lessons an increasingly prominent component of post-secondary mathematics curricula. However, relatively little is known about how students watch and learn from instructional videos. We describe and use an eye-tracking methodology to investigate attentive fidelity—the degree to which students attend to the visual imagery that is the subject of the video narration at each moment in time. Our preliminary study suggests that students’ attentive fidelity varies widely, but there was no evidence that this fidelity is connected to students’ ability to solve calculus problems. 
    more » « less
  4. Growing interest in “flipped” classrooms has made video lessons an increasingly prominent component of post-secondary mathematics curricula. However, relatively little is known about how students watch and learn from instructional videos. We describe and use an eye-tracking methodology to investigate attentive fidelity—the degree to which students attend to the visual imagery that is the subject of the video narration at each moment in time. Our preliminary study suggests that students’ attentive fidelity varies widely, but there was no evidence that this fidelity is connected to students’ ability to solve calculus problems. 
    more » « less
  5. Predicting where users will look inside head-mounted displays (HMDs) and fetching only the relevant content is an effective approach for streaming bulky 360 videos over bandwidth-constrained networks. Despite previous efforts, anticipating users’ fast and sudden head movements is still difficult because there is a lack of clear understanding of the unique visual attention in 360 videos that dictates the users’ head movement in HMDs. This in turn reduces the effectiveness of streaming systems and degrades the users’ Quality of Experience. To address this issue, we propose to extract salient cues unique in the 360 video content to capture the attentive behavior of HMD users. Empowered by the newly discovered saliency features, we devise a head-movement prediction algorithm to accurately predict users’ head orientations in the near future. A 360 video streaming framework that takes full advantage of the head movement predictor is proposed to enhance the quality of delivered 360 videos. Practical trace-driven results show that the proposed saliency-based 360 video streaming system reduces the stall duration by 65% and the stall count by 46%, while saving 31% more bandwidth than state-of-the-art approaches. 
    more » « less