Does US federal law mandate a “kill switch” for alcohol-impaired drivers in cars made in the future? According to several fact checkers, no. However, an article by Jon Miltimore for FEE shows that it does.

The issue isn’t one of what the law contains, but of terminology. In claiming there is no kill-switch mandate, USA Today refers to the very text that mandates it:

Automobile experts told USA TODAY the bill does not direct a kill switch to be implemented in cars, nor does it give any third parties, including law enforcement or government officials, access to the in-vehicle technology. Rather, the bill in question directs a federal agency to require technology that would detect driver impairment and disable the vehicle in that scenario.
 
In 2021, Biden signed the Infrastructure Investment and Jobs Act, bipartisan legislation aimed at enhancing the country’s infrastructure. One provision in the bill seeks to prevent alcohol-related driving fatalities by making “drunk and impaired driving prevention technology” standard equipment in all new vehicles.
 
The claim, however, misrepresents the bill’s directive on the technology.
 
Specifically, Section 24220 of the bill directs the National Highway Traffic Safety Administration to develop rules that would require new cars to be equipped with technology that “passively monitors the performance of a driver,” identifies whether they may be impaired and prevents or limits motor vehicle operation “if an impairment is detected.”

In what sense is “technology that … prevents or limits motor vehicle operation” not a kill switch? USA Today’s criterion is human involvement. “[T]he legislation does not direct the agency to require a kill switch – a device that allows someone to shut off a vehicle remotely – that law enforcement or government officials can access.” That’s an odd definition for the situation, and it provides unwarranted reassurance to people who just skim the article.

The FEE article can also be faulted for not identifying the issue. It gives the impression the fact-checkers ignored or failed to notice the provision in the law. They didn’t overlook the text; they just addressed it in a rather misleading way. There are fake fact checkers, but I don’t think that’s what happened here.

In a way, a kill switch that operates automatically is more worrisome than one that requires human intervention. Critical failures of self-driving vehicles, though rare, always make the news. People are wary of decisions made by algorithms, to the point that “algorithm” has become a dirty word. Remote control of vehicles would raise serious concerns about privacy and abuse, but an automatic shutoff at the wrong time — like when you’re trying to move a car a few feet to a safer spot — might pose greater dangers than a remote switch requiring a judge’s warrant. Either way, false positives could interfere with normal operation.

Using the term “kill switch” to include fully automated systems is standard usage. For instance, a VPN kill switch shuts down a machine’s Internet connection if it loses access to the VPN, thus avoiding a privacy risk.

Giving the fact checkers the benefit of the doubt and assuming honest motives, why would they define “kill switch” in such a restricted way? Lack of familiarity with the term might be a reason. If you aren’t familiar with the concept, the term could look melodramatic. It might lead a commentator to think of images of an executioner with a hand poised on a button. A writer thinking this way might not think of an automated shutoff as a kill switch.

Another possibility is mutual cribbing. Writers, like everyone else, like to save effort when possible. Once an article had appeared declaring that the mandated technology wasn’t a kill switch, other fact checkers probably found it easiest to repeat the same arguments in their own words. I’m reminded of the time three major information outlets incorrectly said that the MP3 consortium had pronounced MP3 dead. I got taken in myself, though I subsequently posted a retraction. The false information came out just after the last patents on the MP3 format had expired. My best guess is that one of the sources badly misread the situation and the others based their stories on the first one.

Some writers may like putting politicians in a positive light, especially when they claim to promote “safety,” and so prefer to shade the meanings of terms to make them look good. A “kill switch” installed without your consent is disturbing; a device that “prevents or limits motor vehicle operation” sounds less frightening, even when it’s the same thing. Journalists like to stay on the good side of politicians.

In searching to see who else had used my unoriginal headline, I came upon an interesting article by Martin Gurri on the complexity of fact checking.

We can see that the obsession of Facebook and Twitter with fact-checking was based on the usual misunderstanding of the scientific method. They considered facts to be neutral and inert, simple to prove or disprove without much ideological fuss. Unfortunately, that only works if all of us agree on the framework of truth that integrates the facts—and it is precisely the disintegration of these frameworks in the digital age that triggered the epistemic crisis.

In this case, we have two competing facts. The law mandates technology that “prevents or limits motor vehicle operation”; it does not mandate remote activation of that tech. If you don’t know enough about the tech framework to understand what a kill switch is, you might let the second override the first when deciding what terminology applies.

This doesn’t answer another question, though. Why was the fact checkers’ focus just on whether the right term was used? When people in Congress decree that cars must incorporate a technology that decides whether or not we can drive them, that’s a serious issue by any name. Why should a bunch of power-hungry, corrupt liars have that control over something that I buy? The fact checkers didn’t care about that. They just came up with a bad definition of “kill switch,” decided it doesn’t apply, and considered their job finished.

Who will check the fact checkers? They’re human, have biases, and can make mistakes as bad as the ones they claim to find.