11 ms·
Okay the speech issue aside because it's frankly been debated to death, what I find really crazy about the situation is how much this is being treated by Facebo
by Barrin92 9y ago
Okay the speech issue aside because it's frankly been debated to death, what I find really crazy about the situation is how much this is being treated by Facebook as a 'mechanical turk' like task.
The author describes having only a few seconds for a pretty complex task and is exposed to disturbing material for essentially their entire workday, without any adequate psychological training at all.
This is not how content moderation should be done at all, and I'm not even sure that this should be treated like a routine job.
- KSteffensen 9y agoAgreed, this sounds terrible. I can understand why the author quit after just 3 months.
- th0br0 9y agoNote that Facebook has outsourced the moderation to an external company (Arvato Bertelsmann) and isn't employing these people themselves.
- zeeed 9y agoHow exactly does that make the job any better?
- lrem 9y agoThis setup is to protect the company, not the workers.
- cleansy 9y agoIf anything, it makes the job worse by an order of magnitude. Arvato Bertelsmann is famous for bad working conditions. For example: circumventing minimum wage by (de facto) requiring unpaid over hours. They write into the contracts that you only need to work Mon-Fri but in reality they even ask you to show up for work sundays, christmas or whenever they feel like it. If you don't comply, they show you the door. Illegal yes, but with minimum wage you usually can't expense the time or money to sue them.
- twblalock 9y agoAt Facebook's scale, there is not a good way to do it. Even if they were able to scale it, it would not be good enough anyway. You can automate it, but clearly that doesn't catch everything and automated systems can result in just as much bias as humans doing the job. You can hire humans to do it and make them work fast, because the amount of posts per day is extremely high. You can use some automation to help them, but that will combine the problems of human moderation and automated moderation. You can hire lots of humans to work at a normal pace and give them psychological counseling, and spend so much on their salaries that you'll go out of business, and even then they won't catch everything and will be frequently accused of bias. It's a no-win situation for Facebook because even if they were able to objectively and fairly moderate every single post, they would be accused of bias by politically motivated people anyway. This is not a technology problem, it's a social and political problem and Facebook will never be able to please everyone. Twitter is in the same boat.
- tyingq 9y ago>At Facebook's scale, there is not a good way to do it. Even if they were able to scale it, it would not be good enough anyway Facebook is picking the quota for decisions per hour or whatever, and choosing whatever help is available to reviewers. Certainly they have some culpability for the outcome.
- twblalock 9y agoI wasn't trying to apportion blame, because I don't see how it's useful to do so when there is no possibility of a good outcome. If Facebook adjusts the quotas, they will just be doing a more effective job at pursuing an unworkable goal.
- mschuster91 9y ago> You can hire lots of humans to work at a normal pace and give them psychological counseling, and spend so much on their salaries that you'll go out of business, and even then they won't catch everything and will be frequently accused of bias. Then you should not fucking be in the business at all! What FB does is externalizing the cost onto societies and skimming the profit! Cost here being two things: first the cost of mentally fucking up the moderators (so an extensive amount of psychological healthcare has to be shouldered by society) and second the cost to society that arises from letting Nazis and jihadists spread their propaganda (including the cost from real-world violence enabled by said propaganda).
- Iv 9y agoHere is how I would do it: I would create an auto-moderation feature where users could report and classify inappropriate materials. I would have a team of moderators judging some contents in-depth and a system trying to match their judgment to a clique of auto-moderators that are not linked to each other. I do not understand the lack of research in automoderation systems. This has now become a crucial point of social media! The most advanced system I have seen is 20 years old (Slashdot's moderation system) but now we are back to much more primitive reddit-like systems. Why did people stop explore these? In a world where deep learning and blockchains are available through many easy-to-use packages I am surprised that there are not more experiments around that.
- twblalock 9y agoHow would blockchains be useful for moderating offensive content?
- briandear 9y agoWell, by using blockchain, you make whatever you’re doing infinitely cooler, thus attracting higher quality talent who can then do a better job moderating content. Alternatively, I bet we could find a way to use a middle out compression solution to make moderation problems smaller.
- Iv 9y agoI understand this is a buzzword that is overhyped these days and that by putting it next to deep learning I probably triggered many buzzword saturation thresholds. I usually point out that there is only one thing that the blockchain bring: it removes the need to trust a third-party. If you don't need that, then you don't need the blockchain. I would argue that political discussion in the current climate is one of the cases where the absence of a third party of trust would be an advantage.
- Iv 9y agoOut of the top of my head I could see : - Micro-transactions to give incentives/disincentives to some behavior. - Moderators team receiving compensation for their work - Parties escrowing funds in case of moderation appeal - A moderation implemented using smart-contract and not requiring third-party trust - A multi-dimensional currency based on the cliques of reader who approved your messages much.