The appropriateness of media for children is traditionally evaluated based on factors such as language, explicit content, product placement, and the presence of positive role models. An increasingly important and overlooked dimension of digital content quality for kids is the level of visual stimulation. Features like rapid scene cuts, high-contrast colors, and extraneous visual effects are being leveraged to sustain viewers’ attention, especially on engagement-driven platforms like YouTube and TikTok. These features are likely overstimulating, especially to young viewers, and are therefore important when evaluating the quality of digital media for viewers of all ages.
We use computer vision to quantify these overstimulating visual features that are linked to adverse cognitive and behavioral outcomes (e.g., Lillard & Peterson, 2011; Shepherd & Kidd 2024). One measure, “flicker,” captures the dynamic visual properties of media to assess its potential to be overstimulating (Essesx et al., 2022). Flicker quantifies how often and to what degree the visual scene is changing in a video, making it a valuable indicator of content quality.
We created an open-source computer vision toolkit to quantify content quality across platforms, comparing how content consumed by low-and high-income children differs. We discovered that ad-supported media — the kind that is consumed more often by low-income children — contains significantly higher flicker and thus offers less learning potential. We applied our model to quantify visual salience in three categories of children’s media: ad-supported, paid, and public PBS television. Our work represents the first evidence that platform design perpetuates disparities in the quality of media available to families from different socioeconomic backgrounds. We have developed our model into a scalable tool for quantifying features linked to adverse outcomes. Currently, Common Sense Media is developing an AI product integrating our model to make these ‘overstimulation’ ratings accessible to families and educators.
Figure
Free content (YouTube, in blue) is significantly higher in the “bright and shiny” feature of flicker that can disrupt learning, compared to paid streaming or PBS media.
March 30, 2026
This project revisits the net neutrality debate in light of contemporary technological realities. Distributed caching infrastructure is no longer incidental; it is functionally necessary for high-performance internet service. The authors propose democratizing caching by requiring that first-hop ISPs offer standardized caching services as a component of baseline broadband access.
March 30, 2026
This project looks at how AI is being developed and adopted within religious organizations. This is a project that spans disciplinary boundaries, bringing digital sociology and the sociology of culture into conversation with theology and religious studies.
June 10, 2025
The digital economy has penetrated nearly every aspect of society. This research project seeks to examine the effects of this prolific phenomenon on our social and personal development, our ability to access basic needs and social services, and our political and legal institutions and culture.