38
votes
Q&A with Yoel Roth, Twitter’s former head of trust and safety, on the whirlwind first two weeks under Elon Musk, Twitter’s content moderation approach, and more
Link information
This data is scraped automatically and may be incorrect.
- Title
- Banning Donald Trump and meeting Elon Musk: Former Twitter safety chief gives inside account - Poynter
- Authors
- Poynter Staff
- Published
- Jun 30 2023
- Word count
- 6615 words
You have to be careful with insider accounts since people aren't going to talk about their own screwups, but this is a solid, in-depth interview. There's a lot of detail and it rings true to me. Most outsider critiques of Twitter don't.
He does admit being wrong about this:
I think that’s due to the collective decision-making process he talks about. In some cases, without directly blaming his boss, he says he was overruled.
From the outside, it does look like Twitter making decisions, and they made a decision to “own” decisions in the sense of Twitter taking blame for them.
Also, I don’t know if this is related, but there’s a common practice at tech companies to write “blameless postmortems” where the goal is to describe an incident in detail (including who did what) and come up with all the ways of preventing it from happening again. Unless it’s a case of sabotage, the assumption is that people are well-meaning but often make mistakes, particularly when under pressure. All errors are assumed to be systemic errors and the system needs to be changed to catch mistakes before they become serious.
This is in reaction to blame cultures where the goal is to find a scapegoat and fire them. When that’s the process, people are not going to be honest about what really happened.
So, when I see the word “accountability” I stop and think about what it means. Is that blame culture? If not, how is it supposed to work?
But I don’t know how Trust & Safety can catch errors before they become serious. The stakes are high and available information limited because it comes from outside the company from unreliable sources. It depends on fuzzy judgement calls, and the results of a decision often better understood in retrospect.