Social Media's Opaque Response to Conservative-Bias Charges
The question of whether Silicon Valley is biased against conservatives was back in the news recently, along with the usual denials of ideological bias and apologies that any perceived bias was merely a human or algorithmic accident. The problem is that for all their assurances of impartiality and bias-free moderation, the social media platforms refuse to release any data that would prove or disprove their neutrality. They also make it nearly impossible to understand why they remove a particular piece of content.
Take last month’s actions by Facebook when it removed links to an article by Salena Zito. A veteran political reporter who covered the 2016 campaign with distinction for The Atlantic and New York Post, Zito wrote a piece explaining why Donald Trump voters have not changed their views despite the convictions of Paul Manafort and Michael Cohen. Zito posted a link to her story on her personal Facebook page at 7:55 a.m. on Thursday, Aug. 23. Within two hours, she heard from friends who couldn’t access the post and whose own posts linking to her work had also been removed. In each case Facebook said it had removed the post either because it was considered “spam” or because it violated the company’s content standards.
As is typical in such cases, Zito tried exhaustively to contact Facebook to ask why it had censored her writing and to get it to correct its mistake, to no avail. Only when the ban itself provoked an outcry did the company restore the posts and offer its boilerplate statement: “Yesterday we fixed an issue affecting our automated systems that caused some posts to be incorrectly marked as spam. We have restored the posts that were affected and we apologize for the inconvenience.” When asked why some of the posts were allegedly flagged as violating its content standards instead of being listed as spam, the company declined to comment.
Unfortunately, Zito’s experience is merely par for the course in today’s online dictatorships, in which a handful of private companies now control what we are permitted to read and write, subject only to the whims of their founders. Even though social media platforms such as Facebook and Twitter are where we increasingly consume news, as private companies they are not subject to First Amendment protections of free speech and can legally ban anyone they dislike and delete any viewpoints they disagree with.
Even the criteria for what constitutes banned speech is largely secret and constantly changing. After Twitter mentioned last week that a set of accounts it had deleted were “sharing divisive social commentary,” it declined to address how it defined “divisive social commentary.” It clarified that accounts were not removed based on their content, but rather on secret metrics related to their behavior that the company will not divulge. When asked then why it mentioned “divisive social commentary” in its official announcement if those deletions had nothing to do with the content of the tweets, the company also declined to comment.
Social media companies are not accountable to anyone but their shareholders, meaning they have little incentive to make it easy for users to complain when their posts are deleted or their accounts suspended. They typically decline to take any action or even respond in a timely fashion when they mistakenly delete a post or account. Ordinary users who run afoul of Facebook’s opaque rules have little hope of the company responding to their complaints and little recourse. It usually takes a public outcry and a growing media storm before Twitter and Facebook will reverse themselves and issue the predictable statement about human or machine “error” and apologize for the “inconvenience.”
Given the growing number of suspensions, deletions and other actions against leading conservative voices and the lack of recourse to reverse those actions, it is unsurprising that Silicon Valley has faced increasing concern over whether it exhibits ideological bias. The companies freely acknowledge that their staffs tend to be overwhelmingly liberal but argue that their written policies explicitly prohibit bias and thus there cannot be bias.
“Some critics have described the sum of all of this work as a banning of conservative voices,” Twitter executive Nick Pickles told a Senate committee earlier this year. “Let me make clear to the committee today that these claims are unfounded and false.”
Similarly, in response to concerns over “shadow banning,” the company acknowledged “impacting” Republican accounts, but argued that it had also impacted Democratic ones and thus there was no bias. However, Twitter declined to offer a breakdown of those accounts. Nor did it provide the list of Democratic accounts it claimed to have affected, so it is impossible to evaluate its claims. Facebook has similarly argued that it is ideologically neutral.
These are companies that live and breathe data and employ armies of top data scientists who understand their platforms. It seems obvious that they have this information, so why don’t they provide it?
When asked whether Facebook could offer any actual statistics to support its claims that it does not exhibit bias against conservatives, the company declined to comment. It also declined to comment on whether it had actually attempted to measure whether it had bias. Similarly, a Twitter spokesperson forcefully denied that the company had ever exhibited bias against conservatives, but when asked whether it had ever attempted to measure how often it took action against conservatives versus liberals, the company declined to comment.
If Twitter and Facebook were so confident in their ideological neutrality, one would assume both companies would readily release reams of statistics proving it. Instead, they stonewall reporters’ questions while disseminating opaque statements that blame any appearances of bias on secretive behavioral indicators about how accounts “interact” with each other and that perhaps Republican accounts “behave” differently, without offering further details.
Those on the left laud Silicon Valley for suppressing what they see as harmful content, while those on the right condemn it for censorship. All the while, we have precious actual data to know whether any of these claims are true.
Why does this matter?
It matters because of the deeper questions it raises with regards to the biases of the online world. Could Facebook’s and Twitter’s content moderators be taking action more often against conservatives because of subconscious “implicit biases”? If so, that has grave implications towards their impartiality. Could the issue be algorithmic, where machine learning algorithms have inadvertently learned to suppress conservative speech? Or have the suspended conservative voices simply engaged more forcefully in behaviors like bullying or using bots? If the latter, are the companies applying those same standards evenly to all and are there lessons we can learn in how to improve our online discourse?
In the end, the real story is not whether there is ideological bias. It is about the absolute power a handful of companies in California now wield over what Americans are allowed to say and see online -- as well as the companies’ total lack of accountability. As Twitter’s CEO testifies before Congress on Wednesday, perhaps he’ll finally offer some hard numbers to prove the company’s neutrality, but more likely it will merely be a day of more empty assurances.