Understanding Children’s Reading Patterns and their Associations with Learning Outcomes Using Machine Learning Techniques

• In Event: Implications of digital technologies for children's learning and development

Abstract

Electronic books (e-books) have become ubiquitous in children’s lives (Reich et al., 2016), and studies have shown that e-books can be an effective language learning tool in shared reading activities. Yet most of the studies of children’s interaction with e-books mainly focus on children’s verbal performance during the reading process, with a lack of emphasis on children’s nonverbal behaviors, such as gaze and posture, as well as the impact of these nonverbal behaviors on children’s learning outcomes (Sun et al., 2022). To our best knowledge, no studies mapping user behaviors have investigated how posture and distance-to-screen measure children’s learning outcomes in the e-book reading context, though they have been proven as reliable indicators of learning and engagement (Bibbo et al., 2019; Kaakinen et al., 2018). Meanwhile, previous studies mainly focus on the interaction between children and e-books, or between children and parents, but they haven't regarded those three dimensions together in the family setting.
To address these research gaps, this study explored children's reading behaviors in terms of gaze, posture, and distance-to-screen and their association with vocabulary learning outcomes. We conducted a secondary analysis using a dataset collected from a larger study. The current sample includes 121 parent-child pairs (children aged 3-7) from China, who were assigned to read the e-book twice. Reading behavior measures were coded based on their reading engagement. For example, gaze was coded on a 4-point scale (1= “child fixated the gaze on the e-book or their parent.”; 2= “the child was switching the gaze between e-book and parent.”; 3= “child’s gaze was unclear.”; 4= “the child was unfocused.”). Specifically, the higher the number was in each variable, the less attentive or engaged the variable represented.
Children were grouped into two clusters using k-means clustering based on the means of gaze, distance and posture: the High Engagement group (HEG) and the Low Engagement group (LEG). Then t-tests were used to compare the vocabulary outcomes between the two groups. The analysis showed that the HEG received 6.28 points higher than the LEG in the vocabulary test (t(114)=2.04, p=0.04).
Our results indicated that children who demonstrated higher levels of visual attention, leaned forward during reading, and seated within reachable distance to the screen tended to achieve higher vocabulary learning outcomes, compared with those who always looked elsewhere and laid back during reading. This might be because the HEG took more charge of the device and stayed more focused on the storyline. Also, children who looked at their parents more often tended to generate better learning outcomes. This might be because the parents of the HEG were more likely to interact with their child, e.g. repeating and explaining the key vocabulary, engaging in conversations. Our findings offered valuable implications for shared reading. Instead of clicking for the children, parents should ensure that their kids take control of the reading device. Parents should also interact more with their children and direct children’s gaze towards e-book or parents to keep children highly engaged.

Group Authors

Pu(Priscilla) Zhao; Feiwen Xiao; Huaming Zeng; Dandan Yang; Mark Warschauer; Penelope Collins

Authors

• Pu(Priscilla) Zhao, Stanford University
  Presenting Author

• Feiwen Xiao

• Huaming Zeng

• Dandan Yang, University of California, Irvine

• Mark Warschauer, University of California, Irvine

• Penelope Collins, University of California, Irvine