A policy change by YouTube to limit the spread of harmful videos across the wider internet appears to have had the desired effect.
In January 2019, YouTube said it would no longer recommend what it deemed to be harmful videos through its recommendation algorithm, although it still allowed those videos to be stored on its website. at the New Jersey Institute of Technology and his co-authors were sceptical such an approach would have much effect because links to YouTube are so ubiquitous across the broader social media landscape.
鈥淚 was actually surprised at the results we found,鈥� he says. He and his colleagues monitored how 920,000 videos were shared on Twitter and Reddit over an eight-month period before and after YouTube made its policy change.
Advertisement
The researchers looked at three types of video content: two that are considered harmful by YouTube鈥檚 policies 鈥� conspiracy videos, and videos posted by a group of alt-right YouTubers in New York in 2018 called the Alternative Influence Network (AIN) 鈥� and a control group of videos from mainstream news channels. They analysed 827 million tweets and 79 million Reddit posts to see how many included the videos monitored. Facebook wasn鈥檛 included because of issues accessing the data 鈥� a limitation the authors acknowledge.
Conspiracy videos known to have been subject to YouTube鈥檚 move saw a drop in how many times they were shared on Twitter and Reddit, as did videos from the AIN. Mainstream news channel videos saw an increase in shares outside YouTube 鈥� indicating that the policy change worked. However, less overtly harmful conspiracy videos not targeted by YouTube saw an increase in reach off the platform.
The authors suggest that this could be because the YouTube algorithm isn鈥檛 recommending overtly conspiratorial videos, but recognises that users have interest in videos from the same creator 鈥� so recommends more innocuous or borderline ones.
鈥淭he solution to the perceived issue did work, and then as a result of that had some unintended consequences,鈥� says Buntain.
at Data & Society believes that the results show how YouTube can act effectively in a way between doing nothing or completely banning certain channels or users. 鈥淭hey make a case that some of those midway options do strike a pretty effective balance in terms of limiting the spread of potentially offensive content, or content that might be spreading harmful misinformation,鈥� she says.
Reference: Proceedings of the ACM on Human-Computer Interaction, DOI:
