Mozilla researchers found that user interactions with YouTube videos have little effect on the behavior of the recommendation algorithms. This is reported by The Verge.
Researchers analysed the accounts of more than 20,000 users of the service who installed the RegretsReporter extension. The extension allows users to filter content and hide unwanted videos.
According to Mozilla representatives, the “Dislike”, “Not interested” and “Don’t recommend videos from this channel” buttons have little to no impact on the recommendation system. Users still encountered similar content.
The authors of the study claim that, at best, the algorithms fail to do the job and in the feed offer more than half of the unnecessary videos. At worst, the effect of the buttons is negligible.
On average, the “Dislike” option affected recommendations in 12% of cases, “Not interested” in 11%. The buttons “Don’t recommend videos from this channel” and “Remove from watch history” proved more effective — 43% and 29%, respectively.
The researchers argue that the video service should address the issue and show fewer unwanted videos in the feed.
“YouTube should respect user feedback about their experience, treating it as meaningful signals about how people want to spend their time on the platform,” they said.
YouTube criticised the study’s results and said Mozilla did not take into account many factors affecting how the algorithms operate.
According to the company’s spokesperson Elena Hernandez, the system deliberately promotes a portion of unwanted videos so that users do not find themselves in an information bubble.
“Crucially, our controls do not filter out entire topics or viewpoints, as that could have negative consequences for viewers,” she said.
Hernandez noted that video and channel interaction buttons affect specific videos and creators, not topics as a whole. The service will not stop recommending all content related to a topic, viewpoint, or speaker, she added.
Earlier in September 2022, researchers reported a rise in user trust in AI moderators.
In October 2021, a former Meta employee accused Facebook of deliberately using algorithms to incite hatred for profit.
Subscribe to ForkLog’s Telegram news: ForkLog AI — all the news from the world of AI!