The video removed was not close to any of the prohibited categories, so it's misdirection to say you can't get to 0% error rate. The claimed "error" is not an edge case that well-intended moderators might make an honest mistake on. Falling back to statistics just obscures that.

As a counterclaim I would say that your comment is misdirection of how moderation works in a big platform.

I have no insider knowledge, but for example last week a big Youtube channel had many of its followers banned because they were spamming (solicited) emojis in chat.

Somewhere deep in the moderation stack that tripped an edge case of spamming/abuse/who-knows-what (not youtube apparently, they had no idea how this happened

I expect them to work under the assumption of "innocent until proven guilty", not vice versa(which is easier: just do a huge banwave, and restore few complaining people as others won't care to fight for it)

> I don't understand what you are expecting of them; having so many moderators to have 0% error rate?

That sounds like a reasonable expectation, yes.

If a company doesn't achieve it or is not even making incremental progress towards it, then it should come under fire.

This position is either incomplete or untenable.

If you allow me a strawman, the only possible outcome of this is to only expect perfection from everyone, thus building a society based on hiding forever any and all your mistakes.

A satirical twitter account impersonating Satan once twitted "one sin or a thousand sin and you and in hell anyway. Why not go for a billion and come down like a hero?"

I am not saying your opinion is wrong, but it absolutely needs a concept of redemption for balance

Sure, I didn't mean that the company must be disbanded as soon as they make their first moderation error.

But perhaps they should pay a small fine for each account erroneously censored, so that they have a clear incentive to get better. (Who would levy that fine, in the context of multinational corporations, is a separate debate.)

I am more interested in who would design such fines. One of the great delusions we live in is that complex social problems have simple solutions.

As a practical example public train service in Germany was privatized some decades ago; obviously to guarantee quality the contract stipulate some quotas that need to be reached, one of these is that delays needs to be below a certain level; here comes a captcha, if a train is cancelled it does not count as a delay.

So the financial incentives are essentially to have little (but non zero) delays and cancel all trains with a above average delay.

You cannot game the system and expect the system is not going to game you.

Or better, are you going to be prepared for the consequence of a policy like the one you propose in practice?

What's your job? Do you do that with a 0% error rate? I certainly don't.

I'm sure you only use systems with 100% uptime too.

> If a company doesn't achieve it or is not even making incremental progress towards it, then it should come under fire.

Tiktok AFAICR doesn't have the poor systems that say various Google products do for example. I'm not aware of a pattern of behavior on their front, just suspicion.