In order to make transparent and clarify some of the criticisms that Google’s video platform, YouTube, faces about the type of content it recommends to its users, the company shared the way in which YouTube’s recommendation algorithm has evolved with the years and how users can delete their history to start over.
In a conversation with some international media in which I participate Forbes Mexico, Cristos Goodrow, Vice President of Search and Discovery Systems Engineering for YouTube, explained that without the algorithm, there would be a pretty boring platform.
“We know that if the recommendation algorithm didn’t exist, people wouldn’t be able to take advantage of the platform the way they do,” says Goodrow, who has worked on the video platform since 2008 and leads a team of “hundreds of people,” of which of which a third are focused on the process of refining and working on the recommendation algorithm.
Today the firm says that thanks to the development of artificial intelligence and machine learning, YouTube analyzes nearly 80 billion signals to build the video recommendation system for users.
While the number of views, watch time, and content authority are among those of the billions of signals that the algorithm takes into account to create the recommendation list, Goodrow explains that some of the most important elements are a series of surveys. that the platform does to the users in which it asks if in the video they have just seen they found added value.
Follow the technology information in our specialized section
“I cannot say that one factor has more value over another because it is an algorithm of machine learning, it is not an equation. But we take these surveys into account to improve the recommendation algorithm, ”he says.
Similarly, Goodrow adds that if YouTube users are seeing recommendations that are not to their liking or that do not deliver value, there is always the function to delete the history.
To do this, just go to the settings menu in the YouTube Android or iOS app or on the website, go to the “History and Privacy” option and the user will be able to see and choose the “Clear playback history” options. and “Clear search history”.
By doing so, the recommendation algorithm starts from scratch for the user.
Controversial algorithm
While Goodrow shares that Google has invested money, human capital, and outside expert advice to improve the recommendation algorithm, it does acknowledge that the system is far from perfect and has had to change “a lot” over the years.
“At first the algorithm worked based on the number of views or clicks to a video, but we realized that this number does not necessarily represent value for the user because there were people who clicked on a video but did not really see it or that. video was not what I expected and the seconds were out ”, he shares.
It was there when they chose to analyze other factors such as playback time, but he adds that in a short time they discovered that this factor did not always determine quality.
“There are times at night when you go to sleep and put a video on YouTube of something that doesn’t really interest you or doesn’t take much attention, that is, it doesn’t add value to your life and what we were looking for is that the time people spend on YouTube because it adds value to their lives, ”says Goodrow.
Starting in 2015, as YouTube gained ground as a platform not only for entertainment but also for information, the platform began to take specific actions, explains YouTube’s VP of Engineering.
Some were reducing the reach of tabloid videos, beginning to survey users, raising the visibility of verified and authoritative news outlets, as well as constantly reviewing the impact of the reporting system. machine learning when recommending content from minority group creators or new creators.
Despite the changes, the recommended video system has been one of the most criticized by content creators on the platform, particularly with new creators or members of niche communities or minority groups, who have raised their voice saying that YouTube only gives preference to the big names on the platform.
Follow us on Google News to keep you always informed
“We do not want people to spend time on YouTube that is not valuable, so we created control tools so that the user can determine how much time they can spend watching the app and we constantly work to measure the impact or operation of the algorithm”, explains Goodrow .