In British slang, calling someone a muppet is a way of saying they are foolish and have about as much intelligence as a cloth puppet. When it comes to twitter’s artificial intelligence for detecting “potentially sensitive content” the label seems to fit.

Here is an example of something Twitter felt needed to be hidden as sensitive content. The reference to the end of the world and “Don’t panic” comes from Douglas Adam’s Hitchhiker’s Guide to the Galaxy. The figure next to it is Beaker, who is literally one of Jim Henson’s muppets. The tweet, from an Australian user, was about the censure motion in Parliament against former Prime Minister Scott Morrison and Morrison’s response.

Mean time, at around the same time, here is an example of a tweet with an image which was not marked as sensitive content, nor removed. This tweet was in reply to a tweet showing Nick Fuentes denying the Holocaust. The post condemns Holocaust denial, then engages in Holocaust distortion, disinformation, and antisemitism.

The first problem with this image is the conflation of the abortion debate in the United States with the Holocaust. Experts are careful not to use the term Holocaust even in the case of other well documented genocides. There are characteristics that make the Holocaust quite unique. The the term Holocaust is used in other settings it erases understanding about these unique factors and distorts understanding about the Holocaust. Using the term Holocaust in other contexts degrades understanding of the Holocaust even further (and that is true in this case regardless of whether a person in the US context is pro-life or pro-choice).

The second problem is that there have been mass killings as a result of incitement against abortion providers. This tweet might be considered incitement to violence.

The third problem, is that this tweet seeks to immunise those on the right from self reflection if they do something that spark a negative reaction and leads to some calling them “racist, fascist, or Nazi” (to quote the post). Such a posts undermines thinking and moderation in favour of encouraging people to ignore criticism and opposing views, stay in the echo chamber, and drift to the margins where the politics are more extreme.

The fourth problem, is that the person in the image is not George Soros. It is Oskar Groening, an enthusiastic Nazi and bookkeeper/accountant at the Auschwitz-Birkenau concentration camp. Misinformation claiming the image is of George Soros, and that Soros was a Nazi, has been circulating for years and the claim was fact checked and proven to be false by the Reuters in June 2020. George Soros is in fact not only Jewish, but a Holocaust survivor. Misrepresenting him as a Nazi is not only disinformation, but is antisemitic and contributes to Holocaust distortion.

Given this particular meme is not new, the Twitter bot should be able to identify it and remove it, mark it as disinformation, or at a bare minimum hide it as potentially sensitive information. Instead it is the muppet that gets hidden. This AI needs to do a lot better.

UPDATE 1/12/2022:

The above tweet was reported to Twitter at the time we saw it. The report was rejected about 24 hours later. Here is the reply:

So not only is the AI failing, but the review process is failing as well. While reporting this we found the image was posted types more by this user within the last 24 hours. By not taking action, Twitter allows it to stay online, but also fails to provide the signal saying it is inappropriate so the person (if it is a result of ignorance) stops posting it.

Here are the other copies: