meta³: Fact-checking fact-checkering of fact-checkers
Even the worst fact-checking cites sources and deals in facts
The movie “2000 Mules” has been thoroughly debunked. The most important issue is they refuse to make their “proof” public. They aren’t allowing anybody else to look at it, either to verify or debunk it. No matter how plausible their claims, the facts are that their claims are still unsubstantiated. The only way their claims can made believable is to show the proof they claim to have.
But in any case, their claims aren’t plausible. Geolocation data doesn’t work the way they claim. It’s accurate enough to find possible mules, but it’d be too inaccurate to avoid false positives. It’d would catch people like Uber drivers who merely drive nearby. Geolocation data isn’t accurate enough to prove they’ve visited multiple dropboxes, nor is it even accurate enough to provide probable cause for law enforcement to investigate.
Fact-checkers have pointed this out, like AP and Politifact.
A right-wing website RedState.com attempted a rebuttal, Factchecking the Factcheckers. What’s notable about their article is that it really cites no facts.
They do have a point. The AP and PolitiFact articles above weren’t very good. But here’s what makes fact-checking better than other crud: they cite facts. Whether they make a mistake, whether you agree with them, we can examine their citations and argue the point.
The crud from RedState rebuts the fact-checkers without citing any facts. Their rebuttal is full of rhetoric, but no facts. They aren’t fact-checking the fact-checkers.
In this post, I go through the points in that article, using facts.
Geotracking/Cellphone Location Data
The general claim of the movie “2000 Movies” is that activists called True-the-Vote purchases geolocation data from a data broker, and used it to find people who visited ballot dropboxes many times, as well as organizations affiliated with Democrats.
They refuse to let anybody else see their data, so it’s unverified. But the major point of fract-checking is that it’s implausible.
The AP fact-checker quotes university professor Aaron Stiegel saying geolocation data isn’t accurate enough. RedState counters with the following paragraph.
That’s simply not true. In the first sentence of the quote, the writer says that experts say that a smartphone can be reliably tracked within a few meters. Depending on what “a few” means in this case, that could be six or nine feet. That’s hardly leaving a healthy amount of uncertainty. Also, we’re not just talking about one visit to a ballot box.
There are no facts here, it’s just an argument, just rhetoric. They aren’t citing a qualified source. They aren’t asserting their own expertise. They aren’t citing any factual information.
I am an expert. Let me explain the truth.
The accuracy of mobile phone geolocation isn’t a matter of a theoretical debate but of practical measurements. Every ‘ping’ that includes latitude and longitude contains other parameters, such as accuracy. An example is shown below, showing the raw data for a location ‘ping’. We don’t need to debate what the accuracy might be, the data itself contains a precise measurement of how accurate it is.
There is a lot of interesting data here, such as estimating which direction you are moving and how fast. The one we care about is “horizontalAccuracy”. In this ‘ping’, it’s 21 meters or about 60 feet.
I used an app to record such ‘pings’ for a week, then analyzed them. No sample showed an accuracy better than 4.5 meters (15 feet), and most samples where around 20 meters (60 feet).
The ballot dropbox at my local library (Sandy Springs, GA) is located under a covered walkway. This interferes with GPS. While out in the open I might have 5 meters accuracy from multiple GPS satellites, when I walk under cover, the accuracy goes out to 20 meters.
Again, this isn’t theory, but what I actually measured.
Of course, not all phones are alike. My phone is an older iPhone XS. When I upgrade to an iPhone 15 later this year, I’ll repeat the experiment. It’s got enhanced “dual band” GPS that should give better accuracy.
This discussion is incomplete. You should treat this as definitive proof about accuracy. Apps and data brokers have algorithms that attempt to clean up the data. That’s why the accuracy you see on the phone appears greater than the raw data. It’s using tricks like noting pedestrian walk on sidewalks and cars drive on roads.
However, these algorithms are faulty when applied to unusual conditions, like somebody dropping off ballots.
The point of this conversation is merely to show the facts about accuracy. While sometimes a mobile phone might be accurate enough, generally it isn’t. No matter how many times you point to examples where the phone was accurate, you are ignoring all the cases when it really wasn’t accurate at all.
This is why asking the movie or Truth the Vote to show the raw data is so important. We can speculate about what the ‘horizontalAccuracy’ might’ve been in their dataset, but we really don’t know until we see the data. Most importantly, they haven’t actually made a claim about their ‘horizontalAccuracy’. They’ve instead just done a bunch of hand-waving about what it might’ve been.
The facts are that ‘horizontalAccuracy’ is generally going to be around 20 meters, which is more than enough to tell if somebody is nearby, but not enough to show somebody is precisely near a dropbox. It means that delivery drivers, election workers, Uber drivers, FedEx drivers, police, and a whole lot of other people will falsely get included in the list. Every expert cited by fact-checkers confirms this, the RedState article cites no fact or expert that would disprove this claim.
In short, using geolocation data this way is a good start, but you’d have to confirm each and every instance. You can’t say you’ve proven 2000 mules unless you’ve investigated each and every person and found other data confirming this.
The RedState article cites other stories hyping geolocation accuracy:
Pieces in the Washington Post and the New York Times also characterize cellphone location data as quite specific and reliable.
Well, yes, but nowhere do these stories refute the above claim. The WaPo story says only that it’s accurate enough to locate the building. Everyone agrees that its accurate to tell if you are nearby a building. We are disputing whether it can prove you visited a ballot box.
The RedState article isn’t dealing in any facts here, just innuendo. It’s using ill-defined terms like “precise”. Is 20 meter accuracy ‘precise’? If all you want is to know of somebody is nearby a building, it’s precise. If you want to show somebody near a dropbox, it’s not precise at all.
The RedState article attempts to refute the AP’s claims that the data is anonymized. It points out how it was used to identify those who invaded the Capitol building on Jan6.
In 2021 the Times published a piece titled, “They Stormed the Capitol. Their Apps Tracked Them.” about the January 6 protesters.
The difference here is that what police can get with court orders is very different from what you can get from data brokers. The police can get the identity of an owner of a phone in ways that data brokers cannot. They can also get more comprehensive location information than what data brokers can provide.
RedState claims:
The data purchased by True the Vote most likely contained the mobile advertising identifier information, which would allow True the Vote to discern the real names, addresses, etc.
This isn’t fact-checking but supposition.
We can still get a person’s identity most of the time. If ware actually getting a comprehensive list of all their locations, then we can figure out where they live and work. It doesn’t give you their identity within seconds, but can lead to their identity with a bit of work.
One of the arguments is whether True the Vote correctly filtered out false-positives, such as tracking the pattern of life before the election. The AP lists some plausible reasons why this won’t work, such as election workers who only visit all these locations during the election and not before.
Note that this is an inherent problem with fact-checking. The AP can’t disprove the claims by True-the-Vote that they filtered out false-positives, because we don’t have the data. It can only describe plausible reasons why it may not have worked. If one side makes implausible, unverified claims, the only real fact checking is that they are implausible and unverified. We can only show why they are implausible by coming up with plausible alternative. This doesn’t prove them wrong, only implausible.
The correct conclusion from the AP isn’t that it’s impossible to do what True-the-Vote claimed, but implausible enough that we can’t trust them without seeing the evidence for ourselves.
The RedState article implies that we can easily weed out false positives from “election workers” who pickup ballot boxes. This misses the point that there are a ton of people whose job it is to visit many locations who don’t pick up ballots. An example is in the AP article, quoting Sen. Sharif Street, how he and his staffers visit many locations.
Use of Gloves in Georgia Senate Runoff
The RedState argues other points than just geolocation information.
The movie “2000 Mules” claims not only geolocation data, but cites surveillance videos showing other suspicious activity, like people wearing gloves when dropping off ballots, which the movie claims was to hide fingerprints on the ballot. They vaguely cite a case in another state where people feared leaving prints on the paper ballot. The movie doesn’t claim it’s to avoid leaving fingerprints on the dropbox, probably because it wouldn’t make sense — too many people touch the dropbox that any fingerprints wouldn’t be recoverable anyway.
The AP rightly debunks this as being pure speculation, saying “This is pure speculation. It ignores far more likely reasons for glove-wearing in the fall and winter of 2020 — cold weather or COVID-19.”
The facts here aren’t that it’s definitely cold weather or covid fears, only that it’s pure speculation that gloves are to avoid fingerprints on ballots. The AP speculates about other possible explanations.
The RedState articles tries to debunk the “cold weather” theory but pointing out that during this election (the Senate runoff in December) that the weather wasn’t cold. This is disproven by videos form the movie. At timestamp 0:42:21 in the movie, we see a woman warmly wearing thick gloves. She takes off the glove of one hand to grab the ballot and insert it into the slot. It’s pretty clear here the weather was cold, and that since she took off the glove and touched the ballot, she wasn’t wearing gloves to hide fingerprints.
At timestamp 0:47:32, we see this guy drive up to a dropbox. He dons a surgical mask and latex gloves before getting out of the car. He’s wearing only one surgical glove — on the hand that touches the germ-ridden dropbox. The other hand leaving fingerprints on the ballots does not have a glove.
It’s obvious from this video that the intention was protecting against covid rather than to avoid leaving fingerprints on the ballot.
The facts here support the AP’s speculation. From these screenshots we can see that at least some wore thick gloves because of weather, and some wore thin gloves because of covid. In both pictures, since they touched the ballot with their ungloved hand, the glove’s purpose was not to avoid fingerprints.
The RedState article debunks something that wasn’t in the AP article, namely:
AP has no comment as to why one would be out in the middle of the night placing numerous ballots in the drop box
Again, despite being a “fact-checking” article, there’s no fact-checking here.
First, the facets. During that runoff election, sunrise was 7:39 AM and sunset at 5:32 PM. It would likely be dark when people were dropping off ballots on the way to/from work. You would expect most surveillance videos to show darkness.
Second of all, there’s no reason evildoers would do the deed at night. The entire point of visiting many dropboxes is to avoid suspicion, because dropping off 5 ballots for a household is neither illegal nor suspicious. You’d want to do it during the day with lots of people around. If you do it at midnight, in a well-lit location with a surveillance camera watching, you are just asking for suspicion.
The movie implies that these things were at midnight (rather than shortly before/after work), but provides no evidence to backup this assertion. Most of the videos shown in the movie are in daylight.
In any case, there are plenty of police and Uber drivers active around midnight.
Photographing Ballots/Ballot Boxes
The movie claimed people photographed themselves dropping off ballots to prove their actions so they get paid.
The AP rightly points out this is just speculation, that there could be lots of reasons why people might photograph themselves at dropboxes, that “across the U.S., voters frequently take photos of their ballot envelopes before submitting them”.
RedState correctly points out that the AP doesn’t substantiate this claim, that they assert it with no evidence.
But it’s not to hard for us to substantiate the AP’s claim. If you search social media, you find lots of people taking pictures of themselves, their ballots, and dropboxes. An example is this tweet:
This person just takes a picture of the building from her car:
This person takes a video of them inserting the ballot into the dropbox. This person has a friend take a picture shortly before dropping off the ballot, as did this person. This person wonders how many ballots might be dropped off in the pictured dropbox. This person drops off her ballot at night, what is she up to??
Finally, this is a picture taken by me during the 2020 Presidential election of the dropbox at my local library. Was I a mule and not even know it??
People post pictures of their activities all the time on social media. That’s the Occam’s Razor explanation, not that they needed to take pictures to get paid.
Dropping Off Ballots for Family Members
The movie shows clips of people dropping off multiple ballots, saying they are mules “stuffing” the boxes with an illegal number of ballots.
The AP rightly points out that none of these videos show anybody stuffing an illegal number of ballots, as it’s legal to drop off ballots for a household. No video from the movie shows anybody visiting a dropbox more than once. The AP points out the case of a man who was investigated, cleared, and who is now suing the movie makers for defamation for calling him a criminal with no evidence.
The AP makes the incorrect claim “there was no way to tell whether those voters were the same people as the ones whose cellphones were anonymously tracked”. Obviously, they can match up the timestamps from the geolocation ‘pings’ with surveillance videos.
But curiously, the movie never does that. It’s the entire premise of the movie, matching geolocation data with surveillance videos, but never ever in the movie do they even do this. The only imply they’ve done this.
The RedState article takes their vague claims as fact that they’ve done this. It’s not “fact”, it’s “claims”. That’s the entire point of fact-checking, to distinguish claims from objective facts.
Some “Mules” Were Also at Antifa Rallies
This claim has been so widely debunked that D’Souza removed the claim from this book version of the movie.
The claim was specifically about details, that True-the-Vote used a database from a group called ACLED to matching BLM rioters with ballot mules. ACLED pointed out that this is impossible, as their data is anonymized. It also doesn’t make sense given the technology.
It’s certainly possible that True-the-Vote could’ve also requested data from the BLM protests/riots, but they don’t claim to have done so. They only claim to have matched with ACLED data.
Though to be fair, the RedState post has a good point about bias. The AP describe the Jan6 Capitol siege as a “riot” but refers to BLM riots as merely “protests”.
Conclusion
I agree that the AP and PolitiFact both did a poor job “fact-checking” the movie “2000 Mules”. They rightly point out that the geolocation claims are implausible, but they could’ve done a better job.
But still, even bad fact-checking is fact-checking. They cite sources for their claims which we can double-check ourselves.
In contrast, while the RedState piece claims to be fact-checking the fact-checkers, it contains virtually no facts. It’s all rhetoric, inuendo, and argument.
To improve this, they need to either cite hard facts, cite experts that agree with their claims, or assert they have personal expertise in the subject.
I admit this post isn’t as robust as it could be. I am an expert in how mobile phones track geolocation and the algorithms data brokers use. So I’m not citing other experts here. But at least, I’ve provided enough details so that other experts can refute my claims.