One of the big problems that YouTube has faced, along with other social networks and communities, is the propagation and dissemination of false news, a fact whose consequences are often harmful in some ways. Therefore, the company has been promoting different measures for some time visibility and value to those channels and videos that really offer proven information.
Although part of the origin of the problem is, in part, on YouTube itself, as this year Bloomberg reported that YouTube executives gave orders to employees so that toxic videos would not stop being recommended, the reality is that in recent times YouTube months is doing to change the tables.
Proof of this is the fact that they announce today, the arrival in Spain of measures to reduce content that may include false or harmful information. It was something they have been doing for a long time in the United States, the United Kingdom, Ireland, South Africa and other English-speaking countries. Together with Spain, the set of measures reaches Brazil, France and Germany.
Evaluators who will give their opinion on the quality of the videos to prepare machine learning models
The first country in which the measures that we will now review were launched was the United States, in January 2019. In that and in the other countries mentioned, a team of external evaluators from around the world has been giving its opinion to the company on the quality of the videos, not in the sense of resolution or bitrate, but on the health or toxicity of its contents, as well as its veracity.
To do this, these evaluators are based on YouTube’s public guidelines, with up to nine opinions per video. In the event that the subject is more complicated than giving a simple opinion, YouTube states that Third parties specifically trained in the subject matter are also used to discuss whether the content is harmful or not.. The example they give of more complex videos are those who talk about medical treatments.
Once the opinion of the evaluators is collected, from Google they take this information to train machine learning systems tested to prepare models. It is with them, and not with human supervision, that the company claims "to review hundreds of thousands of hours of videos every day to find content that may include false information and limit its dissemination." In the statement they state that over time "the accuracy of these systems will continue to improve."
YouTube launches its plan in Spain to try to reduce content with false or harmful information
- Social networks and communities