While it wouldn't solve the problem in and of itself, I think one mechanism that could help mitigate this phenomenon would be to compel companies like YouTube a

While it wouldn't solve the problem in and of itself, I think one mechanism that could help mitigate this phenomenon would be to compel companies like YouTube and Facebook to (somehow) disclose what their algorithms select for promotion (that is, for suggestion or automatic presentation). When the citizens don't and can't know what YouTube is playing for people, they can't really object. On the other hand, if the data was there and enough citizens were interested for somebody to analyze the data, then we could know: Oh, during the 6 weeks leading up to the Brexit vote, among the top 1000 videos X% strongly advocated for, Y% strongly advocated against, Z% didn't fit into strong advocacy. Oh, and, this false disproven pro-Brexit piece from Boris Johnson was played 18,960,000 times to an estimated 14,000,000 unique viewers. That would provide another vector of incentive/consequence that content promoters — algorithmic or otherwise — would have to factor into their decisions. I think this would also help with the other, extremely related problem of "algorithmic child abuse" described by the viral article "Something is wrong on the internet" from last year[1]. As we shift from a mass broadcast media landscape (TV, radio, print), increasing algorithmic content selection has huge implications, for society, democracy, children, etc. Because each person (or child) potentially receives their own custom-tailored feed of content (articles, videos) and ads, it is extremely difficult to, first of all, _know what is actually happening_. (That is, what is being shown to the public, or in the second case, what is being shown to individual kids.) This makes it effectively impossible to hold these companies accountable for what they are choosing to show people. I am sure that these companies would fight tooth and nail to prevent such legislation from being enacted, and I am not confident that the collapsing democratic institutions in my own country (USA) are even capable of doing so. It's a pretty hard thing to legislate. Even if the rule was something like each user's account had to make accessible, somewhere, the complete history of content and advertising selected for display for that user. Then, scientists and journalists could at least use random sampling of people, get their data, and do analysis that way. Although probably, with micro-targeting of individuals, we would also need the global totals for each content item or ad to also be available. I think algorithmic selection has the potential to so terrible because it is happening in the dark. There are very few consequences for bad behavior when it is known only to the "victims". Especially if they don't think they are victims; e.g., my young kids when they're watching the grotesque algorithmically generated videos of their favorite video game characters that YouTube suggests for them — my personal introduction to this entire topic. If you haven't seen it before, I recommend the below-linked article to anybody with children, or who just wants to think more about the implications of algorithmic/ML content selection. [1] https://medium.com/@jamesbridle/something-is-wrong-on-the-internet-c39c471271d2 https://medium.com/@jamesbridle/something-is-wrong-on-the-in...