In this month’s research spotlight, we highlight recent research from COSMOS that focuses on using YouTube videos to discover meaningful trends—specifically, the studies titled,
 

• “Unveiling Bias in YouTube Shorts: Analyzing Thumbnail Recommendations and Topic Dynamics,” and
• “Investigating YouTube Narratives and User Resonance in the South China Sea Dispute.”
  

Each studied different trends in YouTube data, such as popular narratives or algorithmic recommendation bias. These studies were published and presented recently at this year’s annual International Conference on Social Computing, Behavioral-Cultural Modeling & Prediction and Behavior Representation in Modeling and Simulation (SBP-BRiMS 2024), which took place from 18 to 20 September in Pittsburgh, PA.

“Unveiling Bias in YouTube Shorts: Analyzing Thumbnail Recommendations and Topic Dynamics” investigates bias in YouTube Shorts’ recommendation algorithms by analyzing thumbnail recommendations using advanced topic modeling techniques. Focusing on the South China Sea Dispute as a case study, the researchers collected over 100,000 videos and used generative AI and BERTopic to track topic shifts across recommendation depths. Their study revealed a significant algorithmic bias where initial serious or geopolitical topics quickly drifted towards more entertaining, high-engagement content. As recommendation cycles progressed, substantial topics were replaced by lifestyle, entertainment, and miscellaneous videos. This bias potentially marginalizes important subjects by prioritizing popular, less serious content and underscores the need for more transparent and equitable recommendation practices.

“Investigating YouTube Narratives and User Resonance in the South China Sea Dispute” analyzes YouTube narratives about the 2022 South China Sea Dispute using advanced AI techniques. The researchers examined 4,225 videos, extracting insights on sentiment, emotions, morality, and toxicity. Key findings included a prevailing negative sentiment, particularly around the Russia-Ukraine conflict, and high levels of fear in the narratives. The research highlighted narratives about NATO’s support for Ukraine, China and Philippines’ diplomatic interactions, and emerging geopolitical tensions. Notable trends included surges in harm-related narratives, toxic content featuring political figures, and emotional variations between video content and user comments. This study provides a nuanced understanding of the online discourse surrounding complex international disputes.