Improving YouTube’s Algorithm
These two links connect to the concept of web searches within the World Wide Web, specifically about how hyperlinks allow one to traverse the World Wide Web network to find their desired information. In the first link, a blog written as an official announcement by YouTube, reveals how the recommended videos on YouTube are similar to that of hyperlinks. The recommended videos serve as direct connections to other similar videos just like how hyperlinks connect similar information in the World Wide Web. Just as the Web is built on these hyperlinks, YouTube is built on their ability to effectively recommend videos that their consumers will enjoy as seen by the fact that their users spend roughly 700,000,000 hours each day watching videos recommended by the algorithm. In class, we learned about the concept of information abundances which is what the YouTube algorithm is trying to navigate through. There are many relevant videos for retrieval and the algorithm narrows those options for each individual. In the past, YouTube’s algorithm tracked and measured the previous viewing habits of the users, and similar users, to recommend other similar videos to them. The more data the algorithm has, the more efficient the recommendations to what the algorithm thinks is the desired outcome of every individual user will be. In other words, the algorithm will be able to recommend certain content to others and avoid recommending the same content to different users based on predictions from the different users’ history. Since then, YouTube has updated their algorithm so that the recommended videos that are suggested are now based more so on viewer satisfaction.
The first link is an official response from YouTube to the events that are described in the second link. In early 2019, it was revealed that YouTube’s algorithm made it “easier for pedophiles to connect and share child porn in the comments sections of certain videos”. Because of the nature of the algorithm, YouTube could not distinguish between the morally good or bad as the algorithm looked strictly at the numbers and specific key words or phrases of a user’s history in order to create a link to similar videos with those same specific keywords or phrases. The recent changes of the algorithm due to this scandal aimed to focus on viewer satisfaction, measuring likes and dislikes of videos, and providing the chance to do a survey. YouTube hopes that these changes will allow the algorithm to recommend the best videos to their consumers.
Link 1: https://youtube.googleblog.com/2019/01/continuing-our-work-to-improve.html
Link 2: https://www.wired.com/story/the-toxic-potential-of-youtubes-feedback-loop/