Let's Talk About 'Neutrality' -- And How Math Works

from the you-don't-want-moderation-neutrality dept

So if the First Amendment protects site moderation and curation decisions, why are we even talking about “neutrality?”

It’s because some of the bigger tech companies -- I’m looking at you, Google and Facebook -- naively assumed good faith when asked about “neutrality” by congressional committees. They took the question as inquiring whether they apply neutral content moderation principles, rather than as Act I in a Kabuki play where bad-faith politicians and pundits would twist this as meaning that the tech companies promised “scrupulous adherence to political neutrality” (and that Act II, as described below, would involve cherry-picking anecdotes to try to show that Google and Facebook were lying, and are actually bastions of conversative-hating liberaldom).

And here’s the thing -- Google, Twitter, and Facebook probably ARE pretty damn scrupulously neutral when it comes to political content (not that it matters, because THE FIRST AMENDMENT, but bear with me for a little diversion here). These are big platforms, serving billions of people. They’ve got a vested interest in making their platforms as usable and attractive to as many people as possible. Nudging the world toward a particular political orthodoxy? Not so much.

But that doesn’t stop Act II of the bad faith play. Let’s look at how unmoored from reality it is.

Anecdotes Aren’t Data

Anecdotes -- even if they involve multiple examples -- are meaningless when talking about content moderation at scale. Google processes 3.5 billion searches per day. Facebook has over 1.5 billion people looking at its newsfeed daily. Twitter suspends as many as a million accounts a day.

In the face of those numbers, the fact that one user or piece of content was banned tells us absolutely nothing about content moderation practices. Every example offered up -- from Diamond & Silk to PragerU -- is but one little greasy, meaningless mote in the vastness of the content moderation universe.

“‘Neutrality?’ You keep using that word . . .”

One obvious reason that any individual content moderation decision is irrelevant is simple numbers: a decision representing 0.00000001 of all decisions made is of absolutely no statistical significance. Random mutations -- content moderation mistakes -- are going to cause exponentially more postings or deletions than even a compilation of hundreds of anecdotes can provide. And mistakes and edge cases are inevitable when dealing with decision-making at scale.

But there’s more. Cases of so-called “political bias” are, if it is even possible, even less determinative, given the amount of subjectivity involved. If you look at the right-wing whining and whinging about their “voices being censored” by the socialist techlords, don’t expect to see any numerosity or application of basic logic.

Is there any examination of whether those on “the other side” of the political divide are being treated similarly? That perhaps some sites know their audiences don’t want a bunch of over-the-top political content, and thus take it down with abandon, regardless of which political perspective it’s coming from?

Or how about acknowledging the possibility that sites might actually be applying their content moderation rules neutrally -- but that nutbaggery and offensive content isn’t evenly distributed across the political spectrum? And that there just might be, on balance, more of it coming from “the right?”

But of course there’s not going to be any such acknowledgement. It’s just one-way bitching and moaning all the way down, accompanied with mewling about “other side” content that remains posted.

Which is, of course, also merely anecdotal.

Reposted from the Socially Awkward blog.

Filed Under: content moderation, content moderation at scale, errors in moderation, neutrality
Companies: facebook, google, twitter