Algorithmic Content Recommendations on a Video-Sharing Platform Used by Children

Jenny Radesky,Enrica Bridgewater,Shira Black,August O’Neil,Yilin Sun,Alexandria Schaller,Heidi M. Weeks,Scott W. Campbell

DOI: https://doi.org/10.1001/jamanetworkopen.2024.13855

2024-05-30

JAMA Network Open

Abstract:This cross-sectional study examines how video-sharing platforms use attention-grabbing designs in video thumbnails to encourage engagement among children and whether recommendations offer more problematic content over time.

medicine, general & internal

What problem does this paper attempt to address?

This paper aims to explore potential issues with content recommendation algorithms for children on video-sharing platforms. Specifically, the researchers focus on how thumbnail designs on these platforms encourage user interaction through attention-grabbing designs and test whether algorithmic recommendations over time lead to more problematic thumbnail features.

In January 2022, the researchers conducted a cross-sectional study that mimicked children's search behavior on popular video-sharing platforms (such as YouTube) by randomly clicking on recommended videos to test whether thumbnail designs change after consecutively watching 20 videos. They created a data collection environment with no digital footprint and collected data through a browser using an unlogged user account.

The main results of the study include the coding of six aspects of features: visual impact, drama and suspense, luxury and wish fulfillment, creepy and disturbing, violence and pranks, and gender stereotypes. The study found a large number of problematic features in these thumbnails, such as creepy and disturbing features (44.6% of 1283 thumbnails), violence and pranks (40.6% of 1170 thumbnails), and gender stereotypes (18.2% of 525 thumbnails). Additionally, there were many attention-grabbing designs, such as visual impact (79.1% of 2278 thumbnails), drama and suspense (91.5% of 2636 thumbnails), and luxury and wish fulfillment (44.7% of 1286 thumbnails).

The study also found that while most problematic features did not significantly increase over time, gender stereotypes did increase with the recommended content (P-value <0.001). This suggests that video-sharing platforms may recommend problematic content to children, especially when they search for popular content. Therefore, further research is needed to understand how children cope with these thumbnail designs and whether these designs affect the quality of their content consumption.