Eye Tracking to Support eLearning

Abstract

Online eLearning environments to support student learning are of growing importance. Students are increasingly turning to online resources for education; sometimes in place of face-to-face tuition. Online eLearning extends teaching and learning from the classroom to a wider audience with different needs, backgrounds, and motivations. The one-size-fits-all approach predominately used is not effective for catering to the needs of all students. An area of the increasing diversity is the linguistic background of readers. More students are reading in their non-native language. It has previously been established that first English language (L1) students read differently to second English language (L2) students. One way of analysing this difference is by tracking the eyes of readers, which is an effective way of investigating the reading process. In this thesis we investigate the question of whether eye tracking can be used to make learning via reading more effective in eLearning environments. This question is approached from two directions; first by investigating how eye tracking can be used to adapt to individual student’s understanding and perceptions of text. The second approach is analysing a cohort’s reading behaviour to provide information to the author of the text and any related comprehension questions regarding their suitability and difficulty. To investigate these questions, two user studies were carried out to collect eye gaze data from both L1 and L2 readers. The first user study focussed on how different presentation methods of text and related questions affected not only comprehension performance but also reading behaviour and student perceptions of performance. The data from this study was used to make predictions of reading comprehension that can be used to make eLearning environments adaptive, in addition to providing implicit feedback about the difficulty of text and questions. In the second study we investigate the effects of text readability and conceptual difficulty on eye gaze, prediction of reading comprehension, and perceptions. This study showed that readability affected the eye gaze of L1 readers and conceptual difficulty affected the eye gaze of L2 readers. The prediction accuracy of comprehension was consequently increased for the L1 group by increased difficulty in readability, whereas increased difficulty in conceptual level corresponded to increased accuracy for the L2 group. Analysis of participants’ perceptions of complexity revealed that readability and conceptual difficulty interact making the two variables hard for the reader to disentangle. Further analysis of participants’ eye gaze revealed that both the predefined and perceived text complexity affected eye gaze. We therefore propose using eye gaze measures to provide feedback about the implicit reading difficulty of texts read. The results from both studies indicate that there is enormous potential in using eye tracking to make learning via reading more effective in eLearning environments. We conclude with a discussion of how these findings can be applied to improve reading within eLearning environments. We propose an adaptive eLearning architecture that dynamically presents text to students and provides information to authors to improve the quality of texts and questions.