An Opportunity for Twitter

How the revolution at Twitter may give them the chance to break their chains. Or not.

Like many people I’m curious about the ongoing train crash that is Elon Musk’s acquisition of Twitter. The first part of this is the human carnage being wrought as the company is dismantled and remade which horrifies and transfixes the person in my that has built organisations. It’s fair to say I’m amazed anyone is still there. I know I wouldn’t be. A different and interesting angle is also to think about how Twitter can dig themselves out of the mess they were in when Elon Musk entered the building. This is the same mess that almost all social media companies find themselves in right now. Content moderation and taking responsibility for the safety of their users.

I’ve written before about why Freedom of Speech is Hard on the Internet and with governments around the world finally starting to legislate for the harms of social media, hate speech and disinformation on the Internet most Social Media companies find themselves in a difficult position. How do they scale their content moderation so that they are able to meet their current and future legal duties?

Let’s put some numbers on the fundamental problem at play here. How do you scale content moderation? First up start with the headline number. There are 500 million tweets a day on Twitter. That’s 200 billion a year or 350,000 a minute. Whichever way you slice it that’s a huge number. The scale of this number is the first problem.

Taking the simplest case that you want someone to check every single tweet for hate speech, copyright, illegal or inappropriate content it quickly becomes apparent that you can’t. Or at least you can’t afford to. Even if you can check a tweet in 10 seconds on average (this assumes just text and not having to watch all of a video) then every day you have to spend 5 billion seconds a day checking tweets. If you assume a working day is 8 hours (or 28,800 seconds) then you’d need 173,000 people to do this. Clearly this is financially unfeasible so this is not the answer.

The next approach you might think about is to automate things completely. Search for phrases that are known to be offensive and ban them. This also doesn’t work and disproportionately affects people talking about Scunthorpe (a town in England). You also…