Here are the inner workings of YouTube recommendation engine's accuracy

How many times have you found new or engaging content while clicking on recommended videos on YouTube? Recommendations may or may not have the majority user vote, but a lot of thought has gone into developing the algorithm that makes these suggestions to users, a top executive said on Wednesday.

Reiterating how recommendations play an important role in maintaining a responsible platform, YouTube today shared insights into how its recommendation engine works.

“Today, Recommendations drive a significant amount of the overall viewership on YouTube, even more than channel subscriptions or search. And, we are thinking about it in a responsible way. Our goal is to help connect viewers to high-quality information by minimising the chances of them seeing problematic content. Our goal is to have views of borderline content from recommendations below 0.5 per cent of overall views on YouTube.” said Cristos Goodrow, Vice President, Engineering, YouTube.

Unlike other platforms, YouTube does not connect viewers to content through their social network. This basically means that unlike Facebook or Instagram, you will not see videos in YouTube that your friends or larger social networks are watching.

YouTube’s recommendations' philosophy is based on accurately predicting the videos the viewer wants to watch - based on their own interest and preferences, and not based on the people they are connected with.

YouTube built out Recommendations on the simple principle of helping people find the videos they want to watch and that will give them value. Viewers find them at work in two places: the “homepage”, that appears when one first opens YouTube, displaying a mixture of personalised recommendations, subscriptions, and the latest news and information. and the “Up Next” panel, which appears when one is watching a video, providing subsequent suggestions for content based on the current video.

In addition, viewers have controls to manage what and how much they want to share to get a personalized experience on YouTube. For instance, viewers who do not want personalised recommendations can choose to delete watch history.

There has, however, been criticism of the way Recommendations on YouTube work. Earlier this year, research published by Mozilla found that YouTube’s artificial intelligence engine recommends content that users regret watching, in turn increasing views and serving more advertisements.

With people using YouTube not only for entertainment but also for news and information, the platform has used recommendations to reduce low-quality content from being widely viewed, said Goodrow. It has built out classifiers to identify and prevent racy/ violent videos from being recommended, started to demote sensationalist content, removed any video that showed minors in risky situations and further expanded the way the recommendation system is used to reduce problematic misinformation and borderline content.