When Caesar was summoned as a witness against Publius Clodius, his [second] wife Pompeia’s lover, who was prosecuted for profanation of religious ceremonies, he declared he knew nothing of the affair, although his mother Aurelia, and his sister Julia, gave the court an exact and full account of the circumstances. And being asked why then he had divorced his wife? “Because,” he said, “my family should not only be free from guilt, but even from the suspicion of it.”
Life of Julius Caesar, by Suetonius
Legitimacy is a second-order phenomenon, an ideal example of social construction. “Perceived legitimacy” is oxymoron. There is no meta-standard against which things can best tested as legitimate; we create categories and then evaluate cases against those categories.
Often, need to defer this task to a third party, who performs a legitimating function — but this merely kicks the can down the road. Who legitimates the legitimators?
The recent salience of electoral legitimacy in the US demonstrates the reality of second-order legitimacy. Materially, what happened is the same as what happened in 2012: millions of Americans legally cast ballots, which were duly counted by poll workers and deemed legitimate by the larger electoral apparatus. The facts on the ground have not changed.
Yet electoral legitimacy is in a fragile place. In 2016, the specter of Russian misinformation spurred many Democrats to claim that Trump’s election — though not literally fraudulent — was still not fully legitimate because it had been tainted by foreign interference. And then, of course, Trump went much further in 2020 by claiming that results produced by the electoral apparatus were not legitimate because they were literally fraudulent.
My belief is that Russian misinformation likely had a negligible effect on the outcome of the 2016 election. We’ll never be able to know for sure, and for present purposes it doesn’t actually matter. I can still think that these tweets — combined with the extensive discussion of these tweets in the media — decreased electoral legitimacy.
Because it turns out that “electoral legitimacy,” as a social construct, cannot be ease neatly defined as we might like. Again, the reality on the ground was the same as it was in 2012: people went and cast those ballots, they were correctly counted, no one explicitly cheated at the process by which we literally legitimate elections.
But it turns out that 2016 made salient another feature of electoral legitimacy that we weren’t used to thinking about. Legitimate elections are ones in which citizens have a fair chance to consider the evidence for and against each candidate and to cast an informed ballot, absent any attempt to manipulate their information environment.
This is, of course, ridiculous: by global standards, US institutions allow for an unbelievable amount of “manipulation” of the information environment. Our yearslong campaign cycle and laissez faire campaign finance laws make “manipulating the information environment” central to the practice of electoral democracy.
The combination of our aging population and digital media technology makes the problem even worse, as I argue in Generation Gap: Why the Baby Boomers Still Dominate American Politics and Culture. Just one gripping example. A shockingly high percentage of donations during the 2020 election were literally fraudulent, to such an obvious extent that they were actually returned.
Given that these are our “legitimate” or at least normal electoral institutions, maybe the real crime of the Russian bots in 2016 is that they did it all for cheap.
But these are our institutions. The UK, Australia, Canada, every country in the EU — by their standards, corporate PACs financing television ads 10 months ahead of the election would be disastrous for electoral legitimacy.
For better or for worse, we have doubled down on the liberal principle of unrestricted, no-holds-barred “free speech” as the guiding principle of our information ecosystem. But as the 2016 election/Russiagate illustrates, we still care that it’s our information ecosystem, to be manipulated by us—not by foreign state actors.
TikTok therefore represents an unacceptable vulnerability for electoral legitimacy. To date, there is zero evidence that the CCP has pressured TikTok into putting their thumb on the algorithmic scales. There does seem to be systematic differences in the popularity of different political topics on TikTok and Instagram based on whether those topics are sensitive to the Chinese government, but this doesn’t necessarily imply explicit pressure to change the algorithm. Remember, the “recommendation algorithm” is ultimately dependent on the content that people choose to create; it could easily be the case that people who want to talk about these issues systematically prefer to do so on Instagram.
But that doesn’t mean they won’t do it in the future! At the risk of playing Hume, there’s no way to inductively demonstrate that something won’t happen just because it hasn’t happened yet.
By the standards of Russiagate, this would be a disaster for American electoral integrity. Imagine, for the weekend before Election Tuesday, the TikTok algorithm up-ranking videos attacking Biden. Or, for that matter, Trump. Or RFK Jr.
Journalists, NGOs and (presumably) government agencies would catch on pretty quick. But not within those three pivotal days before the election!
Sure, if this happened, we’d definitely ban TikTok. But we’re close to doing that anyway. If anything, the high-profile discussion of banning TikTok (or forcing it to fully divest) makes such an attack more likely in the upcoming election; they’re not likely to get another shot at it.
The manipulation could be much subtler and still be very effective. Consider the classic paper “A 61-million-person experiment in social influence and political mobilization”, in which Facebook sent out platform messages encouraging people to vote by showing them pictures of friends who had reported their intention to vote. This single intervention — which was essentially free for Facebook to deploy, on the scale of campaign spending — increased voter turnout by 0.4% among people exposed to the message.
That is a massive effect, for the cost. And 0.4% was lower than Trump’s margin of victory in Michigan in 2016, or Biden’s margin of victory in Arizona in 2020.
The Facebook experiment randomized who saw this “social pressure” message across everyone on the platform, so the net electoral effect was likely negligible. But if they had specifically targeted only Republican-leaning people in swing states, they might have swung at least some of those states.
And TikTok has demonstrated that they are willing to use these kinds of platform messages to influence American politics. On March 7, they sent out the following message en masse.
The flex of power seems to have backfired; it was immediately obvious to Members of Congress and their staff that these calls were the result of a coordinated pressure campaign rather than a grassroots campaign.
The subtle, get-out-the-vote manipulation might be more directly impactful on electoral outcomes. But this kind of highly visible intervention is worse for the electoral legitimacy. Our electoral institutions implicitly tolerate the former as part of the open battlefield of “free speech,” but as we have seen, the latter move of brazenly undermining elections is more damaging for how people perceive those elections.
Caesar’s family must be beyond even the suspicion of guilt. If anything, the facticity of the guilt is immaterial; when it comes to legitimacy, perceptions are what matter. Allowing TikTok this kind of power over our perceptions is an unacceptable vulnerability in our most important epistemic institution.
An objection is that TikTok is not unique in this respect — that other big tech platforms have the power to significantly de-legitimize US elections. Fair.
So there are two coherent positions: either all of these companies should be free to exercise that power, or none of them should.
I know which of these I prefer — and tomorrow, when Ban TikTok Week continues here at NMS, you will too!