Birdwatch/Community Notes

In September, I was invited to participate in “Birdwatch,” now known as “Twitter Community Notes.” The concept is a misleading tweet can be annotated with additional information or context by the community.

XKCD 386 - someone is *wrong* the Internet. If I don't correct them, they'll remain wrong.
Image source: XKCD 386: Duty Calls

For example, consider this headline:

Polk County Sheriff Grady Judd holds press conference after agency says it seized enough fentanyl to kill 2.7 million people.

WFLA, October 21, 2022

The first part, “the Polk County Sheriff’s office said it helped confiscated approximately five kilos (11 pounds) of fentanyl” is pretty straightforward. But they sensationalize that their work with “enough fentanyl to kill 2.7 million people.

That number intrigued me because it’s presented as a scary fact… and is close to four times the population of Polk County, Florida. Intrigued, I went down the rabbit hole of information.

To get to 2.7 million, they assume 5kg fentanyl seized / 2mg fentanyl / death = 2,500,000 deaths. When I try to work out details:

From this, I’d say the claim of “enough fentanyl to kill 2.7 million people” is bullshit.

Since Twitter only provides a limited amount of space to write and include sources (and I don’t have the benefit of editing, like I did here), this is what was added:

It’s languished since. The original tweet was getting a fuck-ton of views and amplification with the buildup towards the election. Enough so, people will regard it as self-evident.

I spent time on this because I know that every decade has Halloween fearmongering. Indeed, there was a brouhaha about kids’ rainbow-colored Halloween candy being laced with fentanyl that, not shockingly, has completely abated now that the election is over. (The belief that drug cartels are growing their customer base by giving away rainbow colored fentanyl to kids is Underpants Gnomes stupid.)

Election Disinformation pollutes.

The election produced a lot of opportunity for misleading and needing context. For example, there was a tweet by a political party accusing another of being the sole cause of inflation. For context, I noted that inflation is currently a problem in other countries and has multiple causes. The note went live, but in the time it took me to walk over to my wife’s den and show her the link, it was brigaded back into “needs more votes” status.

That happened a lot. During my brief time on Birdwatch (ooh, I’m foreshadowing my next blog entry), I was involved with approximately 350 tweets. Only 60 ever made it to the live site, and most of those were stupid memes (“this 3D forensic rendering of Mary Magdalene by Stanford/Harvard/Crypto Bros looks strikingly like Lady Gaga/Zendaya/Rihanna”).

Please refrain from licking our van!

In short, Birdwatch is depressing and has some severe problems:

  • There are more bullshit tweets than community note makers’ time. Tweets must be flagged misleading then require at least five folks to weigh in on notes, the same way, to appear.
  • While Birdwatch’s intent is to value contributions from different views, and nuance, Agreement from a diversity of perspectives results in a lot of paralysis because …
  • Sources of bullshit proclivity are asymmetric. Of the political-focused tweets I commented on, most were supporting the conservative candidates. My sense is the right-wing echo chamber is much larger and more easily called to blindly claim without evidence/retweet than the left-wing echo chamber.
  • Certain tweeters are “regulars” in disinformation. This, to me, is the biggest area of opportunity, and may be a large cause of the asymmetry. And also the one least likely to be addressed under the new management.