Home / SmartTech / Crowdsources from Mozilla add-ons examine YouTube’s opaque algorithm

Crowdsources from Mozilla add-ons examine YouTube’s opaque algorithm

YouTube has been criticized for years for recommending videos for people – especially the opacity of the algorithms on Google’s own platform. While we don’t Really To understand how this all works, we know that it relies on a real-time feedback loop to suggest new videos for viewing, with suggestions different for each viewer based on their individual viewing activity.

Studies have shown that YouTube plays a huge role in promoting conspiracy theories, spreading misinformation, and “radicalizing” society, and the company appears to have made some effort to reel in the monster it created. However, with the US presidential election approaching, this presents a unique opportunity to examine how ideas spread online. Mozilla today launched a new research initiative to better understand how YouTube algorithms actually work.


With RegretsReporter, a browser add-on for Firefox and Chrome, Mozilla invites users to send reports to Mozilla when “malicious”

; videos are recommended to them. After installation, a user can click on the RegretsReporter icon “frown” in their browser bar. This only works when there is an active YouTube video on the screen.

Above: RegretsReporter add-on

This will show the video viewing history of that session and allow users to click the “Report Video and History” button.

Above: RegretsReporter: Reporting a video

Before submitting, the user is asked to provide more information about what they found regrettable about the video.

Above: RegretsReporter form

According to Mozilla, it hopes to gain some insight into the patterns regarding the frequency or severity of malicious content and whether there are specific usage patterns that lead to such recommendations being made.


The real value of RegretsReporter, of course, lies in the quality and quantity of the data it collects. According to Mozilla, every report submitted includes YouTube browsing history for up to 5 hours prior to the report submission. This includes what type of YouTube pages were visited and how they were reached. Users can also manually delete from their history any specific video they do not want to send, although doing so would likely compromise the integrity of the data.

Additionally, it’s pretty clear that YouTube recommendations are based on display patterns that go back well over 5 hours. This raises questions about how insightful this research is likely to be – even if it does manage to scale it up in terms of installations. Mozilla recognizes this shortcoming, but believes the data is valuable enough to produce meaningful insights.

“Many of our research questions can be explored without knowing a user’s complete viewing information,” Brandi Geurkink, senior activist at Mozilla, told VentureBeat. “For example, are there any identifiable patterns in terms of frequency or severity of the reported content? And does the frequency of reports increase for users after they submit their first report? However, we know that the lessons we gain will not be comprehensive. We want to identify areas where more investigation is needed and then provide momentum to enable that deeper examination. “

And then there is the subject of data protection, a sensitive issue in today’s internet age. In addition to the data that is collected every time a form is submitted, Mozilla’s data protection regulations note that the add-on collects “regular, aggregated information” about a user’s YouTube usage, e.g. B. how often he visits YouTube and for how long – but Not what they see or are looking for on the platform.

So there’s a trade-off here between how willing the public is to participate in a program designed to shed light on YouTube’s dim AI smarts and the data they want to share with Mozilla.

From Mozilla’s point of view, RegretsReporter is part of an ongoing campaign against YouTube that included research and recommendations for YouTube to fix the problem based on Mozilla’s first YouTube regrets campaign last year. His suggestions at the time included that YouTube should open up more data and commit to working with independent researchers. Instead, RegretsReporter tries to get Google out of control, albeit to a limited extent.

“We were shocked by the number of responses to our first YouTube Regrets campaign, which suggests that this is a widespread problem,” added Geurkink. “We hope this will result in thousands of downloads of the extension.”

Based on its findings, Mozilla said it may work with other researchers, journalists, policy makers and “even engineers within YouTube” on solutions to the YouTube algorithm problem, and plans to make the results of its research available for public consumption publish – thought he stopped giving any time scales.

“It’s difficult to say [when Mozilla will publish results] without first understanding how users will be using the extension and without an average rate of regret, ”Geurkink continued. “However, we will run regular analyzes from the time of launch instead of waiting for a set period of time. We therefore plan to share the results soon after they are discovered. “

Source link