|
This is a really good question. I'm surprised how rarely we even entertain the possibility of these systems growing beyond our capacity for effective, low-suffering review and moderation. (As for your footnote - people will rightly point out that any media will be used for some awful content, but that doesn't mean every system has an acceptable rate. "Can we get the rate low enough to tolerate?" is still a legitimate question.) Every time Youtube makes the news for having unsavory content, their responses seem to have the same underlying tone: "we really are trying, but this is impossible." Every day, 24,000 days of video are uploaded to Youtube. Live human review for every new upload would take >100,000 full-time viewers, plus whatever is needed for comments, review appeals, and copyright notices. I can't even find good estimates on how much existing content there is. Some analysts have suggested Youtube might produce $15B/year in revenue. That review team would cost $3B a year at $15/hour, before payroll tax, benefits, counseling, etc; so we're talking about numbers in the ballpark of determining whether Youtube can be profitable. And yeah, you can do playback at triple speed, automate some basic content-filter removals, prioritize content posted under dubious keywords, de-prioritize major channels unlikely to upload something awful. That all lowers costs and makes it more plausible that troubling content will actually get caught. But you're right: even if we wanted to run this as a public good, all those cost-savings only serve to concentrate the human toll. Law enforcement has been dealing with this issue since photography became widespread, and hasn't found a way past traumatizing people who have to sort through abhorrent content. Now, the same digitalization that makes distributing media nearly free makes it possible to create that experience a thousand times as often. (Shock sites, after all, are just a version of the same pattern consciously centered on the unwanted views.) So what to do? Robust tagging systems can help reduce unwanted exposure to content without requiring moderator involvement, the same way it's helped fanfiction communities reach a detente over protecting readers while allowing mature content. But it's telling that the shouting matches over insufficient tagging and robust age controls continue, and of course this only works for content that's acceptable on the site - a "take this down immediately" flag doesn't get used. Federated approaches like Mastodon and WhatsApp groups might reduce legal liability and help people find spaces they're comfortable in and diminish unwelcome surprises. But that sacrifices both oversight and unity - even if we accept that some instances will be used to trade illegal content, plot violence, etc., we still lose the network effects of "searching Youtube" or "being on Facebook". At the other end, I suppose ever-more-aggressive centralization is possible; the ability to multiply harm depends on being able to act repeatedly. If Google demanded your SSN to leave a Youtube upload or comment and banned bad actors for years, or the government ran Facebook with warrantless access to every message, the frequency of this sort of content would decline and the moderation burden might shrink greatly. (After all, anyone can tack a grotesque picture to a physical bulletin board, but there's not much of an issue with obscene UGC there.) Of course, the issues with that are screamingly obvious. Identity and the risk of consequences have a chilling effect, legitimate behavior can still be embarrassing when such a store is predictably breached, usage creep for this sort information is not just a risk but an invariant, and governments are not only inclined but often required by law to act on all sorts of not-actually-bad content like "promoting drug legalization". Beyond that... I don't actually know. Allowing even minimal privacy and UGC while trying to maintain a palatable space seems like a genuinely unsolved problem. There might be a lot of unexplored value in trying to reduce psychological cost without taking humans out of the loop. Most 'digital natives' have seen some horrible shock content without lasting harm, so perhaps the winning approach is to reduce the time/frequency/vividness of exposure to a manageable level. At the very least, it seems like a less exhausted space than the technical side of things. |
It's a little discouraging (considering how influential it is in the evolution of tech) how quickly HN reaches for "rah rah decentralization" as a panacea for anything social media.