YouTube dislike button and information cascade
Last week, YouTube announced a new update that will remove the ability to dislike videos. This was later clarified that users will still be able to dislike a video but the number of dislikes will be kept private from the viewer, available only to the content creator. The company’s reasoning for the change is to protect content creators, particularly small YouTubers, from organized “dislike mob”.This dislike attack, or vote brigading, has resulted in many high profile reports and even websites such as Wikipedia, dedication pages to listing most disliked videos on YouTube. For many content creators and users, the dislike button allows the community to police content. Many users look at the dislike ratio even before watching the video to decide whether the content is worth their time. But, does the dislike button promote negative vote brigading and harassment on YouTube?
Youtube algorithm ranks videos based on interaction and engagement. However, this also means that, in the eye of the algorithm, engagement is engagement regardless of likes or dislikes. As a result, heavily disliked videos can be trending and recommended simply due to the vast amount of dislike, which the algorithm simply sees as strong user engagement. An example of this was “YouTube Rewind 2018,” which remained the top trending video for almost a week after it was published, despite the overwhelmingly dislike ratio of 86%. With more than 19 million dislikes, the video remained the most disliked video to date. In the class, we know that the addition of review, such as like and dislike count, will cause the view count distribution to follow a power-law distribution more closely. This is because the feature leads users to rely on the like and dislike ratio as a way to determine the “goodness” for the video. More people will choose to watch a video with a higher number of likes. However, the combination of the YouTube algorithm and human mentality made the reverse possible too. As the dislike ratio arrives at a certain point, it will begin to follow a power-law distribution. The human element in this process follows the rich-get-richer process where the following view copies the decision of viewers before him by handing out dislike.
This can be represented in the form of an information cascade. If a viewer has ½ chance of giving out like and ½ of giving out dislike, the viewer will rely on the previous viewer’s decision. Given that there is a disproportionately high dislike ratio, the viewer will follow other views and give a dislike to the video.
But is there a better solution to preventing dislike attacks than simply removing the dislike number? While removing dislike numbers will stop the information cascade that allows “dislike attack”, the decision is also incredibly unpopular among YouTubers and viewers alike. Instead of providing viewers with a ratio of like and dislike, Reddit utilizes a very interesting system in which the dislike simply subtracts out the like. If a Reddit post gained 1000 likes and 200 dislikes, the post will simply show 800 likes to the viewers. However, upon reaching 0 like the system simply stops counting any additional dislike, preventing a negative dislike amount on the post. It should be noted that such a system is not necessarily the best in the context of Reddit, as the algorithm ranks content by amount of likes by default, resulting in content with lower likes being buried and forgotten. In the context of YouTube, however, such a system can alleviate dislike vote brigading by preventing users from seeing the amount of dislike, while also maintaining the ability for users to “police” low-quality content. Another option would be to force the viewer to provide feedback in order to dislike a video. This can help empower content creators by providing substantive criticism and act as a barrier of entry that reduces the impact of vote brigading.
https://blog.youtube/news-and-events/update-to-youtube/