17 Sep 2021 Facebook Whitelists Reveal Need For Real Viewpoint Diversity In Tech
Well, what do you know? Facebook has been caught lying again, and again so that it could apply its vaunted safety and protection rules in biased ways.
The details follow the now-familiar pattern, and further underscore the central truth: There are only two ways for Facebook and other social-media companies to avoid rampant and crippling discrimination. Either they need to get out of “content moderation” entirely, or they must ensure that their moderators think like America as well as looking like it. There is no third way.
The Wall Street Journal broke a blockbuster story earlier this week (with further installments apparently on the way). It seems that despite protestations to the contrary, Facebook does not apply the same censorship rules to all users. Rather, it has developed a “cross check,” or, inevitably, “XCheck” system under which favored users are “whitelisted” for special treatment. When the company’s (already biased) algorithms flag a post by one of those privileged users, the normal removals, suspensions or other reactions are belayed until a living human can review the flagged post. But, it turns out, only about 10 percent of the VIP-user posts ever get reviewed. And even when they are reviewed, the reviewers are instructed to err in favor of leaving the post up. As a manager of the XCheck program wrote in an internal memo, “We … index to assuming good intent in our review flows and lean into ‘innocent until proven guilty.’” (Seriously: who teaches these people to write like that? And can they be flogged?)
A Facebook lackey was scooted out to obfuscate, but in fact confirmed that Facebook has no intention of ending XCheck whitelisting, which has grown to include nearly 6 million favored users. And he further confirmed that Facebook recognizes – and doesn’t care – that the methods applied to the rest of its hoi poli users are junk. For the special people, you see, Facebook had “to create an additional step so we can accurately enforce policies on content that could require more understanding.” But for the peons, any old result will do.
The most troubling thing about this program is that a great many employees have whitelisting authority, with lists of VIPs scattered throughout the company, and without clear or consistent rules about who can be whitelisted and on what grounds. What this means is that it’s, in effect, up to employee biases. If the decision-making employees like the users, or the users’ worldview or message, then those users can be whitelisted, so that their posts aren’t reviewed for “misinformation” by bots the way everyone else’s are. And so unless Facebook works hard to have a workforce that thinks like America – that is, that has about the same range of worldviews in about the same proportion as the population – then Facebook’s censorship (or non-censorship) programs will be inherently biased.
But not only does Facebook not work hard to achieve that viewpoint diversity and proportionality, it has actively opposed every effort by shareholders to even get it to affirm that it doesn’t actively discriminate on the basis of viewpoint in hiring, or to include a prohibition against such discrimination in its non-discrimination policy.
And there is no doubt whatsoever that it does discriminate – and that the effects of that discrimination are profound. Recall that Nick Clegg, Facebook’s chief censor, declaredduring last year’s shareholder meeting that he knew that Facebook’s censorship was hitting the unbiased sweet spot because conservatives were opposed to it while liberals were calling for even more.
Now, a reasonably well-educated middle schooler would have seen the flaw in that logic. Unbiased censorship would have everyone across the spectrum responding in broadly proportionate ways. And if Facebook really wished to be unbiased, it would have enough conservatives and libertarians amongst its directors and executives that someone could have set poor Cleggy straight.
But Facebook doesn’t wish. Instead, bias is built into everything. It’s endemic. I’d stake my life on the fact that these whitelists, other factors held equal, are well-stocked with lefties, but show a noticeable dearth of we fine folk of the center and right.
The company has suggested that it’s working toward capping the members of the Bookerati on the whitelist, but even that would only lock in the already extant bias. And even if the company were to dissolve the lists and genuinely subject everyone to the same dumb algorithms, those algorithms will also be biased if they were developed and “trained” by a cadre of lefties. If you write the algorithms so that calling Republicans racist is fine, but calling BLM racist is verbotene rede, then a neutral application of the algorithm will create biased results.
Facebook knows this, of course. Old Dead-Eyes Zuck and his little dog Cleggy are not confused about what their real “moderation” goals are, as Clegg’s Kinsey Gaffe revealed. They know that if they wanted to fix their comprehensive bias – shall we call it systemic discrimination, perhaps? – they would have to go on a center-right hiring binge of world-historical proportions. They won’t. They’ll continue to discriminate against the center/right of this country at every juncture.
Who will moderate the biased “moderators” of this monopolist, friends, if they won’t moderate themselves?
Oh: and who wants odds on whether Facebook keeps blacklists, too?