Mozilla wants help from internet users to study YouTube’s algorithm and improve the video site’s recommendations. This week, the company launched an extension called “RegretsReporter”, which allows anonymous submissions of bad suggestions provided by the video platform to be analyzed by experts.
Mozilla’s goal is to end a major YouTube problem: the recommendations that lead to harmful or misinformation content. According to reports obtained by the company, a simple search for vikings on YouTube can end up in videos of white supremacy, if the user follows the suggestions provided by the platform.
The new extension does not currently block harmful video suggestions, but it can help develop tools for that purpose. In its current stage, the extension only provides ways for the user to send cases of wrong recommendations to Mozilla.
Knowing the algorithm
With the extension installed in Chrome or Firefox, the user can report the video he is watching and the suggestion he considers harmful. The tool also has options for the user to describe the reported content. The function allows you to classify the video as false, offensive or bizarre, in addition to bringing a text box for those who want to provide more information.
Mozilla wants to collect enough information to find patterns in YouTube’s recommendation system. According to the company, the help of users can answer questions such as what are the themes of harmful videos most recommended by the platform and how it is possible to avoid such content.
According to Ashley Boyd, vice president of Advocacy and Engagement at Mozilla, the objective of the initiative is to ascertain the evolutions made by Google in its video platform in relation to the spread of false and harmful content. According to the executive, YouTube usually says that it is taking measures and evolving its algorithm, but currently there are no ways to verify the information coming from the company.
YouTube commented on the matter in a note sent to The Verge. The platform said it has already made more than 30 changes to reduce the suggestion of “uncertain content”, but is open to further research on the subject.
The company points out, however, that the results obtained by Mozilla may contain non-conformities, since the YouTube algorithm is updated frequently. “It is difficult to draw broad conclusions from anecdotal examples and we continually update our recommendation systems to improve the user experience.”