Tackling recommandation biased algorithms with Tournesol
As free and open-source software advocater, this post may be the first of a long serie about projects that need help or deserve some attention.
A (quite) famous French videast (Lê from the channel Science4All) made a video some days ago about an open-source called Tournesol. It’s aim is to furnish to any Internet user a more sustainable recommandation engine when visiting websites like YouTube or Twitter. Why ? If you look at the business model of those companies, advertisement is always in first place (it’s just official 😄, check here). Therefore, the videos shown to you in the recommandation tab are precisely the one that will make you stay on the platform the most. The more you spent time on the website, the more you’ll have to see ads, and boom you have a business model.
You might say, well yeah, but if the videos are about subjects I like, what could go wrong ? Basically you’re accepting that these videos aren’t the one that are good, but the one that will please you the most. Let me ask you that question : when was the last time you’ve sawn a video that was entirely out of your scope ? Something you disagreed with, or that you take the time to find precisely. Chances are it isn’t usual. Welcome in your filter bubble, a world in which all materials you read, watch and learn conforts the opinions you already have.
An handful of PhD students and developers decided to tackle this. Basically, the purpose of that project is to build a recommandation algorithm based on peer-reviewed videos. As instance, anybody can judge a video on different criterias: are source provided, what are the subject, is it offensive etc.
More information on Tournesol can be found here : https://github.com/Tournesol-app/tournesol-app.