Election Security on Social Media Isn’t a Seasonal Business
Plus: How the border bill went down in flames before it even took off.
Get Press Pass every Tuesday and Thursday.
Meta, the parent company of Facebook and Instagram, has an independent oversight board that makes rulings and recommendations for the company’s platforms to implement. Their recommendations aren’t always followed, and Meta isn’t without its share of problems. Today, election security is a bigger issue than ever for the overseers of our online spaces—given the deliberate spread of false information, the organized coordination and incitement of violence, and related problems—and Meta’s oversight board is one of the most important organizations trying to do something about it.
To understand the issue better, I sat down with University of Notre Dame Law School professor Paolo Carozza, who sits on the board. We discussed election security, including lessons from past cycles, and the board’s most recent decision—that Facebook was right to leave up a video manipulated to falsely depict President Joe Biden as a pedophile. This interview has been edited and condensed for clarity.
What are the lessons learned from the Biden video case and how that plays into the coming election season?
We’re all talking about generative AI, and the effect it’s going to have on elections, even though this particular case doesn’t involve an instance of AI-manipulated media. It’s much more crude, you know—shallow, fake, if you will. Nevertheless, it still raises a lot of the same questions. And I think a lot of those questions are actually really a lot harder than I thought when I first began looking at the materials or thinking about the issue.
Part of the difficulty is we don’t have the technology to actually detect a whole lot of what’s manipulated, right? The technology of detection is not keeping up with the technology of generation, and the technology of generation is so rapidly evolving. I don’t come at it from [the] tech side, but from the briefings we’ve received and what I’ve heard, my understanding is that even months ago it would have been really easy to detect manipulated audio and really obvious [to] detect manipulated video. You know, [manipulated] photos were already kind of hard to detect a while ago, and already, months later, we’re in a completely different world.
I think another interesting issue is that a lot of the dangers that we associate with manipulated media—especially related to elections—are already prohibited under other kinds of rules, at least by Meta. So the [company’s] rules on electoral information already prohibit information that, for example, tries to commit fraud in deceiving people about when the election is held, or where the polling places are, or what the law is for who qualifies as a candidate, or who's running, right? That basically is already prohibited, [as are] certain kinds of incitement and bullying and so forth.
So the question that the Biden case poses, which is much narrower, is: What is the problem with manipulation as such, independently of whether it’s affecting the administration of elections or inciting violence? Because those things were prohibited anyway. And when you start asking that question, then it’s a lot harder for me to get to the conclusion that a platform like Meta actually ought to be actively intervening to prohibit that kind of information.
I think the much more effective way to approach that question is to ensure as much transparency as you possibly can [in order to] put the users, recipients, and consumers of that information in the [best] position to make their judgments, and then leave it open. Because otherwise it’s a really fraught situation where you have a technology that’s inadequate and a platform that’s saying, “Oh, we're gonna start taking down anything.”
How does Meta implement its rules and recommendations from the oversight board against bad-faith lawmakers and media personalities?
That’s super hard. We can pull back the curtain, but not everybody wants to look.
The scale of what’s going on is such that it is just physically impossible that this is not being done largely through automated systems. You can’t have human reviewers going through hundreds of millions of posts a day, right? That’s just not possible. And when you are moderating content at scale using algorithmic methods, even if they’re . . . as neutrally designed as possible, they would [still produce] substantial numbers of errors.
[For example, take] 100 million pieces of information. If you’ve got the best algorithm in the world—gold standard, better than anything that’s out there right now—it’s 95 percent accurate. Well, the raw numbers are still 5 million errors a day. So anybody is going to be able to find what they want to find. If you’re on the right and you want to find instances where [the algorithm has] made mistakes on the right and highlight those, you’re gonna find them. The same is equally true on the left . . . So that’s just a reality. And so a lot of the discourse has to be, you know: How do we reduce that error rate?
How is Meta prepared (or unprepared) for such a long election U.S. cycle?
I don’t have more information than what’s been in the public tech press about how they’re preparing for the United States [election]. I do think that some of the lessons that we’ve extracted from the cases that we’ve done on some other electoral situations can be helpful, though. So take something like the Brazil election, and the sort of January 6th parallel that happened in Brazil.
We found that Meta had acted pretty proactively to set up activated protocols and set up groups specifically dedicated to trying to watch what was going on and be really attentive to efforts to, for example, systematically coordinate disinformation or incite violence. The problem was that as soon as the election was over, they just dissolved, and it was gone. And a lot of the problems happened after the election. Part of the lesson to be drawn from that is that I think that Meta needs to put in the kind of vigilance that's required in electoral contexts [not just] the day that the first election starts and the day that the votes are cast, right? It precedes it, it builds up to it, and it tails off afterwards.
You can’t address emergencies in an election year
It’s been a whirlwind couple of days on Capitol Hill. Sunday evening, a bipartisan group of senators released their long-awaited compromise package that pairs significant changes to border security and immigration policy with a supplemental foreign aid package to Israel, Taiwan, and, most controversially among Republicans, Ukraine.
Among other things, the compromise bill includes:
A complete shutdown of the border (not including appointments) if illegal crossings surpass 8,500 in a single day or average 5,000 on a daily basis. (Had this bill become law a year ago, this provision would have shut down the border for much of 2023.)
$7.6 billion in emergency funding for ICE.
$6.8 billion in emergency funding for U.S. Border Patrol.
$4 billion for Citizenship and Immigration Services.
Stricter requirements for asylum claims and resources for speedier processing.
The disposition of unused funds to build more of the border wall.
Broad pathways to citizenship, but in limited cases, such as that of Afghan interpreters and other partners who supported American soldiers in Afghanistan.
There are so many other significant changes proposed in the 370-page bill that, if implemented, they would amount to one of the biggest conservative policy achievements in a generation.
Naturally, most Republicans opposed it right off the bat.
Within hours of the full text of the legislation being publicized, the bill was swiftly chucked into the political fireplace. Top members of the Senate Republican leadership team were “undecided,” which is to say they were waiting to determine which way the political winds were going to blow. The ambivalent crowd included Republican Whip John Thune (R-S.D.) and Republican Policy Committee Chairwoman Joni Ernst (R-Iowa). Republican Conference Chairman John Barrasso (R-Wyo.) came out against the bill, and in a meeting Monday evening, Minority Leader Mitch McConnell recommended his colleagues vote against the procedural step to begin debate on the proposed legislation. McConnell’s expressed opposition can be read as a tactical move to give the bill more time, but imagining that it will fare better as a result is wishful thinking.
That’s partly because in the House, Speaker Mike Johnson declared it “dead on arrival.” House Republicans do not even want to entertain the amendments process to alter the bill to better suit them, and they are instead committed to advancing their partisan H.R. 2 bill—a complete nonstarter in the Democratic-controlled Senate and White House.
The sudden change in fortunes for the compromise package leaves a deep impression of bad faith. Congressional Republicans have a political interest in keeping border security a live, unsolved problem. Especially since Roe v. Wade was overturned, weakening their static appeal to pro-life voters, Republicans have made the border preeminent among the issues they campaign on every election season.
If you’ve followed the negotiations closely (or are simply a loyal Press Pass reader), you’ll understand that any potential foreign aid/border security compromise proposal has been on life support for months. The negotiation process was less about finding common ground upon which to build a consensus bill than it was about Republicans identifying their sacrificial lamb.
The last time Congress sought widespread immigration reforms, in 2013, the person who took the fall was Sen. Marco Rubio (R-Fla.). This time around, the victim appears to be someone with smaller political ambitions (and more thoroughly conservative credentials): Sen. James Lankford (R-Okla.). Luckily for Lankford, this process hasn’t yet given rise to a catchy nickname for him (remember the “Gang of Eight” a decade ago?), which could be a silver lining if it helps this episode to fade from memory faster.
But derisive moniker or no, Lankford is bummed. He stuck his neck out at his party’s behest only to catch the ax blade anyway. In a gaggle with reporters Monday, Lankford conceded that it might have been easier to achieve his goals for the bill if he had been able to put it forward in a non-election year.
What I think would have helped more than anything else is if this would have come to a conclusion in December rather than in February. I think that would have helped. But when we’re in February, we’re in the middle of the presidential primary season. . . . So I think it would have been very different to be able to have [last year]. As you remember in October, all Republicans were talking about how we got to have a change in law, we got to be able to get this resolved, and now that narrative seems to have shifted.
This is all par for the course for one of the least productive congresses in history. Things will only get slower and harder to accomplish as Election Day draws closer.
While he’s been out campaigning for president in the Democratic primary, Rep. Dean Phillips (D-Minn.) hasn’t voted in almost two months, but he’s vowed a triumphant return this evening to reject the House Republican impeachment articles against Homeland Security Secretary Alejandro Mayorkas.
Phillips last voted on December 14, not that he was doing much before that. Between the start of October and through December, he missed a whopping 83 percent of votes while dedicating himself fully to his longshot primary challenge against Biden. (According to ProPublica, Phillips has been the second-most absent member of the House during the 118th Congress.)
Despite his pledge to return to the Capitol, Phillips missed the first two votes this afternoon, one of which was the procedural rule vote on the Mayorkas impeachment. We’ll see if he makes it to the vote series later in the day.
In his most recent showing in the South Carolina Democratic primary, Phillips finished in third place with 1.7 percent of the vote, behind Biden (96.2 percent) and Marianne Williamson (2.1 percent). Maybe the return to Capitol Hill will help boost his notoriety for the next primary.