The Challenges of Moderating User Content on the Internet (and a Bit of History)

Views: 856

I increasingly suspect that the days of large-scale public distribution of unmoderated UGC (User Generated Content) on the Internet may shortly begin drawing to a close in significant ways. The most likely path leading to this over time will be a combination of steps taken independently by social media firms and future legislative mandates.

Such moderation at scale may follow the model of AI-based first-level filtering, followed by layers of human moderators. It seems unlikely that today’s scale of postings could continue under such a moderation model, but future technological developments may well turn out to be highly capable in this realm.

Back in 1985 when I launched my “Stargate” experiment to broadcast Usenet Netnews over the broadcast television vertical blanking interval of national “Superstation WTBS,” I decided that the project would only carry moderated Usenet newsgroups. Even more than 35 years ago, I was concerned about some of the behavior and content already beginning to become common on Usenet. My main related concerns back then did not involve hate speech or violent speech — which were not significant problems on the Net at that point — but human nature being what it is I felt that the situation was likely to get much worse rather than better.

What I had largely forgotten in the decades since then though, until I did a Google search on the topic today (a great deal of original or later information on Stargate is still online, including various of my relevant messages in very early mailing list archives that will likely long outlive me), is the level of animosity about that decision that I received at the time. My determination for Stargate to only carry moderated groups triggered cries of “censorship,” but I did not feel that responsible moderation equated with censorship — and that is still my view today.

And now, all these many years later, it’s clear that we’ve made no real progress in these regards. In fact, the associated issues of abuse of unmoderated content in hateful and dangerous ways makes the content problems that I was mostly concerned about back then seem like a soap bubble popping, compared with a nuclear bomb detonating now.

We must solve this. We must begin serious and coordinated work in this vein immediately. And my extremely strong preference is that we deal with these issues together as firms, organizations, customers, and users — rather than depend on government actions that, if history is any guide, will likely do enormous negative collateral damage.

Time is of the essence.

–Lauren–

The Right’s (and Left’s) Insane Internet Content Power Grab (repost with new introduction)

Views: 610

The post below was originally published on 10 August 2019. In light of recent events, particularly the storming of the United States Capital by a violent mob — resulting in five deaths — and subsequent actions by major social media firms relating to the exiting President Donald Trump (terms of service enforcement actions by these firms that I do endorse under these extraordinary circumstances), I feel that the original post is again especially relevant. While the threats of moves by the Trump administration against  CDA Section 230 are now moot, it is clear that 230 will be a central focus of Congress going forward, and it’s crucial that we all understand the risks of tampering with this key legislation that is foundational to the availability of responsible speech and content on the Internet. –Lauren–

– – – – – – – – –  –

The Right’s (and Left’s) Insane Internet Content Power Grab
(10 August 2019)

Rumors are circulating widely — and some news sources claim to have seen actual drafts — of a possible Trump administration executive order aimed at giving the government control over content at large social media and other major Internet platforms. 

This effort is based on one of the biggest lies of our age — the continuing claims mostly from the conservative right (but also from some elements of the liberal left) that these firms are using politically biased decisions to determine which content is inappropriate for their platforms. That lie is largely based on the false premise that it’s impossible for employees of these firms to separate their personal political beliefs from content management decisions.

In fact, there is no evidence of political bias in these decisions at these firms. It is completely appropriate for these firms to remove hate speech and related attacks from their platforms — most of which does come from the right (though not exclusively so). Nazis, KKK, and a whole array of racist, antisemitic, anti-Muslim, misogynistic, and other violent hate groups are disproportionately creatures of the political right wing. 

So it is understandable that hate speech and related content takedowns would largely affect the right — because they’re the primary source of these postings and associated materials. 

At the scales that these firms operate, no decision-making ecosystem can be 100% accurate, and so errors will occur. But that does not change the underlying reality that the “political bias” arguments are false. 

The rumored draft Trump executive order would apparently give the FCC and FTC powers to determine if these firms were engaging in “inappropriate censorship” — the primary implied threat appears to be future changes to Section 230 of the Communications Decency Act, which broadly protects these (and other) firms and individuals from liability for materials that other parties post to their sites. In fact, 230 is effectively what makes social media possible in the first place, since without it the liability risks of allowing users to post anything publicly would almost certainly be overwhelming. 

But wait, it gets worse!

At the same time that these political forces are making the false claims that content is taken down inappropriately from these sites for political purposes, governments and politicians are also demanding — especially in the wake of recent mass shootings — that these firms immediately take down an array of violent postings and similar content. The reality that (for example) such materials may be posted only minutes before shootings occur, and may be widely re-uploaded by other users in an array of formats after the fact, doesn’t faze the politicians and others making these demands, who apparently either don’t understand the enormous scale on which these firms operate, or simply don’t care about such truths when they get in the way of politicians’ political pandering.

The upshot of all this is an insane situation — demands that offending material be taken down almost instantly, but also demands that no material be taken down inappropriately. Even with the best of AI algorithms and a vast human monitoring workforce, these dual demands are in fundamental conflict. Individually, neither are practical. Taken together, they are utterly impossible.

Of course, we know what’s actually going on. Many politicians on both the right and left are desperate to micromanage the Net, to control it for their own political and personal purposes. For them, it’s not actually about protecting users, it’s mostly about protecting themselves. 

Here in the U.S., the First Amendment guarantees that any efforts like Trump’s will trigger an orgy of court battles. For Trump himself, this probably doesn’t matter too much — he likely doesn’t really care how these battles turn out, so long as he’s managed to score points with his base along the way. 

But the broader risks of such strategies attacking the Internet are enormously dangerous, and Republicans who might smile today about such efforts would do well to imagine similar powers in the hands of a future Democratic administration. 

Such governmental powers over Internet content are far too dangerous to be permitted to the administrations of any party. They are anathema to the very principles that make the Internet great. They must not be permitted to take root under any circumstances.

–Lauren–