|
|
|
|
|
by grzm
3010 days ago
|
|
There's been evidence, admittedly anecdotal AFAIK, that YouTube recommendation algorithms can tend towards increasingly extremist content from quite innocuous starting points. For example: * https://twitter.com/zeynep/status/973716995521748992 * https://twitter.com/Limerick1914/status/972837152940855296 This doesn't mean YouTube staff are purposefully designing the algorithm to specifically trend towards extremism: this can fall out naturally from human psychology and ML algorithms that note which videos increase engagement and recommend videos based on user interest and recommending videos that will increase the user's time on site. I think this is worth looking into in a more methodical manner. You can dismiss it as cherry-picking, but I think it's something reasonable people can be concerned about. And these services and products should be looked at separately, determining which parts are good and which parts are bad. |
|