3 ms·
Looking at the example of their harmful content, half is videos about pro-republic videos talking about the US 2020 election. If that is representative of their
by belorn 5y ago
Looking at the example of their harmful content, half is videos about pro-republic videos talking about the US 2020 election. If that is representative of their study then that seems very much self-selected in both nationality and politics.
It should be noted that those that classified if content was harmful or not was a team of 41 research assistants employed by the University of Exeter.
- philwelch 5y agoYouTube can't win with these people. A few years ago there was a hit piece in the New York Times implying that if you watch YouTube videos of Milton Friedman or Ben Shapiro, the YouTube algorithm will decide you like right-wing content and gradually feed you more and more extreme right-wing content until you become a nazi. This was probably sensationalized and overstated, but there was this type of feedback loop in the algorithm. One obvious corrective to this is to make sure the algorithm also recommends videos that don't already agree with the viewer's inferred biases. As a result, if you're somewhere left of center and the videos you choose to watch on YouTube are influenced by that viewpoint, YouTube will occasionally recommend videos of "Ben Shapiro DESTROYING libs with FACTS and LOGIC" just so you're exposed to the other side.