About this paper symposium

---

Panel information
Panel 24. Technology, Media & Child Development

Paper #1
You Do You[Tube]! The Multifaceted Roles of Online Video Viewing in the Lives of U.S. Children
Author information	Role
J. Alex Bonus, School of Communication, The Ohio State University, United States of America	Presenting author
Rebecca Dore, Crane Center for Early Childhood Research and Policy, The Ohio State University, United States of America	Non-presenting author
Brenna Hassinger-Das, Psychology Department, Pace University, United States of America	Non-presenting author
Julia M. Wilson, School of Communication, The Ohio State University, United States of America	Non-presenting author
Elena O’Hara, School of Communication, The Ohio State University, United States of America	Non-presenting author
C. Joseph Francemone, School of Communication, The Ohio State University, United States of America	Non-presenting author
Abstract
Introduction: Although parents are understandably concerned about children’s increased exposure to YouTube (Auxier et al., 2020; Rideout & Robb, 2020), many perceive digital media as a positive force in children’s lives (Silander et al., 2018). This optimism aligns with uses and gratifications theory (Katz et al., 1974), which states that individuals play an active role in selecting media to meet their needs. Although this perspective acknowledges that media use can produce harmful effects, it underscores the importance of individual agency and emphasizes potential benefits of media use that are often neglected in research with children, such as relaxation, diversion, and joy (de Leeuw & Buijzen, 2016; Nabi & Krcmar, 2016). Drawing on this perspective, we examined parents’ perceptions of the social and emotional gratifications their children derived from watching YouTube. Methods: Parents submitted three YouTube videos recently viewed by their children aged 0 to 8. For each video, parents reported the social context in which children viewed it (i.e., alone vs. with others) and rated its perceived impact across several dimensions (e.g., learning, bonding). Furthermore, our research team coded each video for desirable content (e.g., educational lessons) and problematic content (e.g., aggression). Study Population: The final sample consisted of 358 parents (62.3% female and three non-binary) with children aged 0 to 8 (M = 4.96 years; SD = 1.82). We recruited participants through Amazon’s Mechanical Turk (MTurk), which is a crowdsourcing website that allows users to participate in online studies in exchange for monetary compensation. Results: Results indicated that in-depth educational lessons were rare in these videos (6.9%), and most children (75.5%) viewed at least one video featuring materialistic messages or aggressive content. Despite these issues, parents reported that these videos often evoked children’s joy (M = 4.40 out of 5) and allowed children to explore niche interests (M = 3.45 out of 5). Some patterns varied by age, such that older (vs. younger) children viewed fewer educational videos (r = -.52), and they experienced fewer emotional gratifications from viewing (e.g., r = -.13 for enjoyment and r = -.14 for mood management). However, older children also viewed more videos with peers (r = .13), and parents reported that those experiences facilitated peer bonding (M = 3.84 out of 5). Discussion: Although we replicated some concerning patterns found in previous research, our results also revealed positive aspects of YouTube exposure that are neglected in popular discourse about children’s experiences online. Specifically, parents perceived YouTube as a resource for providing children with feelings of joy, strengthening children’s relationships with friends, and helping children explore niche interests. While it is imperative to acknowledge and investigate the adverse consequences of media use, it is also crucial to document possible benefits of exposure. This perspective is necessary for advancing the field toward a more nuanced understanding of popular media and its role in children’s development.

Paper #2
What are toddlers watching on YouTube?
Author information	Role
Madalynn Woods, University of Michigan, United States of America	Presenting author
Alexandria Schaller, University of Michigan, United States of America	Non-presenting author
Heidi Weeks, University of Michigan, United States of America	Non-presenting author
Jenny Radesky, University of Michigan, United States of America	Non-presenting author
Abstract
Background: YouTube is the most popular video-sharing platform used by young children, with some children watching it for several hours per day (Radesky et al., 2022; Common Sense Census, 2020). In early childhood, content quality is particularly important for shaping media effects on child development (Madigan et al, 2020), but prior studies have shown that much of children’s content on YouTube is low-quality or contains negative themes such as violence, stereotypes, or commercialism (Radesky et al., 2020). Since then, YouTube has created guidelines for creators to make higher-quality “Made for Kids” (MFK) content and has restricted algorithmic recommendations to MFK videos when other MFK content is viewed. Updated research is needed to examine 1) to what extent young children are consuming MFK content and 2) whether the content quality of videos has improved compared to earlier research. Methods: Data were collected from 317 children in the baseline wave of a community-based cohort study of toddlers (24-26-month-olds) and their parents. At a home- or lab-based visit, research assistants collected demographic and media use surveys, as well as links to the last 10 videos viewed on YouTube by child participants. YouTube Kids histories were not collected. YouTube videos were coded with the Common Sense Media coding scheme, adapted for toddler content. We assessed: 1) number of ads per video, 2) MFK video status, 3) Bedazzling (extraneous attention-capturing design features), 4) Positive Role Modeling, 4) Negative Role Modeling, 5) Branded Content (e.g., marketing or name-brand placement), 6) Vicarious Pleasure (e.g., showing toys, luxury items, candy), and 7) Educational Content. Coders were trained to reliability of Kappa >= 0.7. Results: Toddlers were 49.2% female and 44.3% were an only child. Parents were 93% female, average age 36.4 years, 74.2% white/non-Hispanic race/ethnicity, and 54% had more than a 4-year college degree. Preliminary data analysis of 212 videos shows that participants are primarily watching MFK videos (79.7% of videos), averaging 3.3 (SD 2.2) ads per video (range 0-11). Frequency of content codes are shown in the Table, with Bedazzling and Vicarious Pleasure showing high prevalence. A smaller proportion of videos demonstrated strong educational quality (code = 2, 14%) or branded content (18%). Almost one quarter of videos contained negative role modeling in the form of rude or antisocial behavior (e.g. lying, cheating, pranking). Conclusions: Although most of the videos viewed in our sample are designated Made for Kids, significant levels of negative role modeling, vicarious pleasure, and low educational content (with high ad load) remain in YouTube content watched by toddlers. At this age, learning is often through parasocial relationships with media characters (Bond & Calvert, 2014), including influencers. Given the attentional load from ads and extraneous audiovisual features, as well as prevalence of negative role modeling, more research is needed on how YouTube content effects toddler learning and behavior.

Paper #3
A Content Analysis of Educational Alphabet Videos on YouTube
Author information	Role
Somin Park, The University of Hong Kong, Hong Kong	Presenting author
Rebecca A. Dore, Crane Center for Early Childhood Research and Policy, The Ohio State University, United States of America	Non-presenting author
J. Alex Bonus, School of Communication, The Ohio State University, United States of America	Non-presenting author
Brenna Hassinger-Das, Psychology Department, Pace University, United States of America	Non-presenting author
Wei Pu, Faculty of Education, The University of Hong Kong, Hong Kong	Non-presenting author
Abstract
Introduction: As digital media continues to expand, young children are exposed to YouTube more than ever before. One specific domain of knowledge that YouTube content can target is alphabet knowledge. Notably, alphabet knowledge plays a vital role in children’s future reading accomplishments. Alphabet knowledge is a prevalent topic on YouTube perhaps due to its concrete nature and the ease with which it can be portrayed in both visual and audio media. Many parents are amenable to using YouTube videos to support their children’s learning in language and literacy, including alphabet knowledge, yet only a few studies have examined the quality of educational content on YouTube. Therefore, we conducted content analysis to explore how alphabet content is included in YouTube videos viewed by young children, focusing on instructional strategies theorized to promote young children’s alphabet knowledge. We also investigated whether the prevalence of these strategies varied across YouTube videos viewed by children aged 0 to 7. Hypotheses: Based on the previous literature that suggested effective instructional strategies for teaching alphabets to children, we hypothesized these instructional strategies (see coding schemes in the Methods section) would be noted in the YouTube videos which included alphabet contents. We also hypothesized that there would be differences in instructional strategies across viewing children’s ages. Sample: We included a sample of 45 YouTube videos that included instructional alphabet content. We identified learning situations in each video, which we defined as a single, continuous span of time during which the video intentionally draws attention to a particular letter or set of letters (n = 1869). Methods: We coded learning situations with the following coding schemes: definition (whether a learning situation defined letter names, sounds, and forms), repetition (whether a learning situation mentioned a letter from a previous learning situation in the same video), association (whether the target letter was associated with a keyword, any modality, or reference to the printed letter), and meaningful context (whether learning situations included any narrative, prompted viewers to interact with others, directed to the child’s real-life experience). Results: We documented a variety of instructional strategies suggested in the literature that are used in these videos (e.g., defining letter names, repetition, and associating the target letter with the keyword). Nevertheless, there were also inaccuracies in defining letter sounds and missed opportunities for additional content, such as teaching letters in meaningful contexts. Also, instructors from the videos rarely labeled the uppercase or lowercase forms (e.g., uppercase and/or lowercase letters were displayed on the screen without explicitly defining the form). Age differences emerged for all instructional strategies except defining the uppercase form and the lowercase form. Generally, learning strategies were more prevalent in learning situations viewed by older children (e.g., defining letter names, repetition, and association). For learning situations of three instructional strategies - defining letter sounds, whether letter sounds are defined correctly, and prompting interaction with the viewing child - were viewed more frequently by young children aged 0 to 2 years old.

Paper #4
Developing AI to Classify Educational Content on YouTube
Author information	Role
Dr. Claire Christensen, Ph.D., SRI Education, United States	Presenting author
Anirban Roy, SRI, United States	Non-presenting author
Madeline Cincebeaux, SRI Education, United States	Non-presenting author
Ramneet Kaur, SRI, United States	Non-presenting author
Abstract
Online videos have potential to support young children’s learning and development. Children spend more time watching videos online than in any other format (Rideout & Robb, 2020). Unfortunately, 95% of the videos children watch online lack strong educational content (Radesky et al., 2020). We developed Assisting Parents to Review Online Videos for Education (APPROVE) to automatically detect early literacy and mathematics content in videos. It is intended to help caregivers increase children’s exposure to high-quality educational videos, and to help researchers measure content exposure. Hypotheses: APPROVE will: 1. Detect educational videos with greater-than-chance accuracy: 1a. In a curated dataset. 1b. In a naturalistic dataset. 2. Enable estimation of educational video exposure on YouTube among young children. Population: We addressed H1b and H2 using the 2019 Naturalistic Viewing (2019-NV) dataset, drawn from a 2019 randomized controlled trial of the impact of PBS KIDS science and engineering resources (Grindal et al., 2019). Participants were 55 children, ages 4 and 5, in the control group who had YouTube viewing data. Researchers instructed them to use educational media as usual for 1 hour per week over 8 weeks, and gave them data-enabled tablets that blocked PBS KIDS resources. Software tracked YouTube videos viewed in a Web browser. Of 6,258 unique videos collected, the 2019-NV dataset includes 4,613 of them (75%) that were downloadable. Methods: We built and tested APPROVE in 4 steps: 1. Creating a codebook, based on the Common Core State Standards and Head Start Early Learning Outcomes Framework (National Governors Association, Council of Chief State School Officers, 2010; Office of Head Start, 2015). 2. Annotating videos to train the model. Trained annotators curated the APPROVE video dataset by searching keywords for each code. They recorded videos’ content in Qualtrics. 3. Developing a machine learning model: We used a multimodal content understanding framework to combine videos’ visual and audio cues. It extracts visual and audio information, separately classifies videos based on each modality, and combines these classifications into multimodal classifications (Figure 1). 4. Training and testing the model. We randomly assigned 75% of the APPROVE dataset for model training and 25% for testing. We tested the model as compared to human annotations in the APPROVE and 2019-NV datasets. Results: APPROVE is more accurate in a curated dataset of educational videos than in a dataset of videos watched by children (Exhibit 2). Its accuracy in both datasets is greater than chance. APPROVE adapted well when given 10 annotations in the novel dataset. Within the 2019-NV dataset, APPROVE estimated that 16% of videos children watched were educational, meaning they included early literacy or mathematics content. Further analyses will explore the range of exposure by child and indicators of pedagogical quality. APPROVE facilitates efficient research on the impacts of children’s video content exposure. It may also enable development of an educational video recommender.

⇦ Back to session