255 comments

[ 5.4 ms ] story [ 232 ms ] thread
Does anyone else think it is dangerously Orwellian to describe speech as people 'hurting themselves and others'?

We seem to have a serious problem resulting from people living in bubbles of information sources that only confirm their own viewpoint.

How can the solution to that problem be to have a single corporation design the bubble for everyone?

(Note: I know he's taking about actual videos of violence taking place. However my point is that violence is already happening, and hiding that from public view is 'out of sight, out of mind')

Nobody's forcing you to have a Facebook account, or participate in it. It's not "for everyone".
(comment deleted)
Unfortunately, it kind of is becoming that way. Facebook is becoming a de-facto public utility.

The phone and telegraph companies started out as "optional" services. If you didn't like the telegraph company arbitrarily blocking journalists that criticized them, you were free to not use telegraphs.

Facebook is being tied to credit worthiness, and job applications, whether we like it or not.

And in some parts of the developing world, it is more important than a phone number. It is recognized as being so vital, that providing Facebook access is subsidized.

We can no longer pretend that Facebook is some sort of toy, and dismiss criticism by saying that if you don't like it you shouldn't use it.

> Facebook is being tied to credit worthiness, and job applications, whether we like it or not.

Where, how? I've never heard of someone needing a Facebook account for credit and any employer asking for your Facebook profile is a huge red flag.

> Where, how? I've never heard of someone needing a Facebook account for credit and any employer asking for your Facebook profile is a huge red flag.

Here[1]. Saying "wasted" as a status update can affect a credit score. And even if it isn't disclosed, employers, and other entities, may use your Facebook activity. This is a totally valid issue. So, yes, Facebook does matter, whether you like it to or not. It isn't just a toy, and you can't just stop using it, without it potentially harming you. I think it's totally unfair, but people now have to consider their online presence when hunting for a job.

And they don't have to ask for your Facebook account specifically. Companies like ru4.com can tie your financial identity to your social media identity[2]. Just like how even though you didn't submit anything to the 3 credit bureau, they will find data for you, it is also becoming true with social media. I can't log into chase.com without allowing scripts from ru4.com. Luckily, I use a sandboxed Firefox profile, that has never touched Facebook.

[1] https://www.theguardian.com/media-network/media-network-blog...

[2] https://yro.slashdot.org/story/13/05/02/1521239/even-the-ad-...

Walled gardens are awful things in my book.
He may be referring to speech in part, but he's also referring to people killing themselves and others on camera on Facebook. I think that's what has triggered this.
Like the videos that sparked Black Lives Matter and United? Are you sure you want them silenced? And which hurt and kill is sharing worthy?

Any rule that limits speech can be twisted into censorship of worthy causes. And the restrictions on speech always hurt the weaker sides.

(not talking about first amendment which is about who has the right to limit it and in what conditions, just the effects of said limitations)

Pretty sure he is talking about physical violence.
Its unclear to me as well. I've been hearing speech can be violence from my younger coworkers.
Your younger coworkers are wrong. By definition, violence is physical.

There is, however, a push within educational institutions to redefine terms like "violence" as part of a wider "social justice" programme. The crux of it is that offending someone else is a very bad thing to do, regardless of whether it's intentional or your words are true.

Is it true that violence is physical, by definition?

It is one definition, but there are other more abstract definitions that mean "an unjust exertion of force or power", "rough or immoderate vehemence, as of feeling or language", "damage through distortion", etc.

For example, the phrase "violent scolding" is stuck in my mind from a fairy tale I read as a child, and indeed, if you search Google Books for that phrase, you'll see many examples dating back to at least the 1800s, examples that do not mean physical violence.

So, I don't think it's some new redefinition of violence to use it more abstractly. I think of it similar to the term abuse. Sure, abuse often means a physical act, but it often means verbal/emotional acts.

In "violent scolding", the violent is a figure of speech. It's not real violence.

Like a particularly showy car could be colored "violently red". Still, it's not violence that anyone should step in to stop.

Maybe we need a word that is more powerful than harassment, that isn't physical violence.

I've heard people say things like "virtual rape". I think it's problematic to say that something that occurs in text can compare to having your body physically violated.

It's a delicate balance, but Facebook can't even remove overtly racist/sexist/etc. posts/accounts with any reliability right now.
How is this different than astroturfing?
It doesn't seem to fit the definition of astroturfing. [1]

>Astroturfing is the practice of masking the sponsors of a message or organization (e.g., political, advertising, religious or public relations) to make it appear as though it originates from and is supported by a grassroots participant(s).

Are they concealing the source of the messages? It seems like they are just adding more people who have the authority to remove content that violates the TOS.

[1] https://en.wikipedia.org/wiki/Astroturfing

Depends what they remove, and how they remove it.
Astroturfing is building opinions by faking or artificially mis-weighting some opinions, like using fake accounts on social networks to vote and contribute with a particular sentiment, making it look like more people feel that way than actually do.

What Facebook does is to allow and disallow certain types of content based on public guidelines. Even if those rules are nonsense, unfair or politically motivated, it's at least not done secretly unlike real astroturfing.

Since social media use itself contributes to lower self image and depression, how much are they going to look into their own product as contributing to a problem getting worse for an individual. It would seem that work being done to drive engagement is most problematic for those at risk.
(comment deleted)
Probably not at all. Facebook's collective desire for higher engagement and profit will likely outweigh any concern for individual or collective users.
I hear what you're saying, but from a slightly different angle. I wonder if the profit motive creates a conflict of interest the same way that corporate news tends towards bias to satisfy advertisers.

I am thinking of exiting Facebook for at least a couple of months because my posts/shares (which tend to have a political slant or at least broader perspective to them) don't seem to get any reaction or be shared anymore. Neither do music, alternative culture, or sustainability/environmental posts.

If Facebook is unable to give people the dignity to fail at debating one another and be challenged by new ideas, then that may not be compatible with democracy. I hope they fix whatever is going on with their feed algorithm, and maybe 3000 people training AIs will help, but I wonder if the problem isn't technology.

Your premise is overbroad. Some kinds of social media use patterns can certainly have such a result. For others it's a great help. I had catastrophic problems with self image and depression long before there was such a thing as the public internet and my online social networks are absolutely essential to my health and wellbeing.
I didn't claim it was the only cause/contributor. I only pointed out that by now, there is a growing body of research demonstrating that it contributes to such problems.
Fair enough, sorry for misinterpreting you.
this jew needs to be sent to the gas chamber

getting sick of him

Was just talking last week that the last thing any of these large internet companies wants to do is hire a large room filled with low paid workers to do anything, especially here in the US.

If they are making this move they must see some large liability looming on the horizon.

Politically that's a great idea. Put those large rooms in the right states and you can get favorable bills passed.
They don't want politics to get active to interfere with their business.
I didn't understand the "especially here in the US" part of your comment. Can you explain?
its not like low paid and probably part time workers have many employment rights in the USA
The minimum wage itself is a barrier here, along with overtime regulations and a host of other things designed to prevent employer abuse.
That must be a very small host when compared to other first world countries
Compare to Bangladesh, Vietnam, or the Phillipines. Low-wage American workers are not very competitive in a global marketplace. If the cost of shipping and importing goods and services < the cost of complying with U.S. regulations and paying American workers, there's every economic reason to have the work sent out to less regulated, lower-paid regimes.

The way to fix this is to even the playing field by ensuring that it's not cheaper to pay to have this done overseas. Otherwise, the "American-made" companies just have to hope that's enough to convince customers to help them make up the gap in profits ... and it's usually not.

That would require a thoughtful examination of our housing policies in the US that are designed first and foremost to drive up property as an asset class. It's probably the single biggest driver of inflation in the US that necessitates increases in wages.
Both part time and low paid workers are protected by the same labor laws as full time workers in the U.S.

Maybe I am not understanding, can you say what rights are you speaking of specifically?

They would have more protections and rights than an "offshore" employee would have I imagine.

You have to work more than a certain number of hours for ObamaCare to cover you.

Its why Googles quality raters have been put on part time working Bad employers gaming the system its one of the drivers behind trump and brexit

Western workers tend be expensive.
$$$
Which FB has no shortage of though, no?
Capitalism operates on the basis of greed, not need. If a desired result can be accomplished at lower cost, it generally will be. Firms largely concern themselves only with first-order costs, leaving second-order costs to government, not least because the more abstract the nature of the cost, the fewer people are able to comprehend its existence.

If you're interested in this topic you may enjoy Ronald Coase's economic essay The Problem of Social Cost (which brought him a Nobel prize) and his body of work in general.

Seems to me that keeping the labor cost reasonable isn't really that difficult. The only real problem is concealing the process well enough that it doesn't provoke outrage. Understand as you read what follows and the hairs on your neck stand up that this is --- after enough layers of BS are applied --- very close to what will evolve.

Just yesterday we read about how Facebook 'helps' marketers 'track when people were feeling overwhelmed, worthless or insecure' [1]. Focusing on people that exhibit this frequently, particularly if they're being bullied, would catch a bunch of these online suicide cases without squandering labor on everyone else.

Then all you need to do is invent enough euphemisms to enable a degree of profiling; the rape streams that go unreported for days are almost always gangbangers. Should be a trivial job to factor that group out, by their grammar alone.

There you go; the chronically depressed types and the gangbangers account for probably 95% of the headline making cases Facebook has to worry about. Focus a small workforce on those two groups (and possibly a few others; extremists, etc.) and Bob's your uncle.

[1] http://www.moneyandmentalhealth.org/facebook-mental-health/

I think it's a fairly common recent strategy to hire a big pool of low cost workers with the intention of gradually phasing them out as ML tech improves.

See also: Uber drivers.

3000 persons on top of the current 4500 is a big addition. If all these persons are dedicated to the prompt elaboration of complains and violations of TOS, it might indeed make the difference.

I don't ask if it is economically viable, I guess he knows what he is doing. Facebook is not losing money anytime soon.

He is keenly aware (cf the privacy bugs fiascos of mid 2010) that trust in the system is hard to build up and easy to lose. It takes enormous sincere and public effort to pull out of loss-of-trust spirals.
You can count on the positions being low-paid in developing countries. 12 hours a day at $1 an hour maybe.
Isn't there a lot of cultural and language nuance in determining what's objectionable though? I can't see that working out well.
The majority of them should be from developing countries, since that's where most of FB users are.
Oh great, like I want my feed curated by someone from the other side of the world with radically different sensibilities. What could possibly go wrong?
(comment deleted)
OK, 7500 people dealing with (according to the article) millions of reports every week. So let's assume 2M reports, the smallest number that counts as plural "millions". Assuming a 40 hour week and no breaks, this give FB 300K man-hours to handle 2M reports, meaning each person has an average of at most 9 minutes to view the content, investigate, make a judgment and take an action. Anybody who's worked in such a role: Is that enough time?
The question is what '2M reports' means. If that means 2M reports are made by users, but they can be made on the same page/post/user, then they can be easily grouped. So, if you get 500 reports on a single post, the reports can be investigated in bulk and that would make the number of investigations smaller. I believe this is the case.
Nine minutes per piece of content would be _significantly_ more than any other company I've seen has given to reporting of user-generated content. I would also agree that unless it's a video, nine minutes is enough time.
This is a tremendous amount of time per post. A 6-7post/hr throughput is incredibly slow, unless their guidelines require they produce specific, rather than generated, feedback per post.

However, the actual number of people reviewing is unlikely to be 7500. A decent percentage of that headcount will be dedicated management, HR, analytics and administrative staff.

As another poster pointed out, you're assuming the reports are always unique, where in reality the number of unique instances/articles facebook has to investigate is going to be less than the number of times people report something.

Also 2 million isn't "the smallest number that counts as plural millions" - anything over 1 million is considered "millions".

Lastly you seem to be assuming that this is 100% manual and that each report needs to be treated as seriously as each other report. I'd imagine that at the very least these reports will be prioritized and if something is reported once by a user who reports 100 things a week, it will probably be ignored - giving more time for higher priority things.

(comment deleted)
That's 7500 people whom will be "forced" as much as free employment can constitute force to do nothing but look at questionable often offensive content.

What a funny way our economy has evolved. 100 years ago many of these people might have worked in farms, 50 years ago they might have worked in factories. Today they sit in an office looking at ostensibly offensive material that facebook deems the rest of society should not be exposed to.

So? We have 1000s in the bay are building the next chat bot, cat pic ai engine etc.
Equating these two classes of activity is not even wrong.
I think it will soon enough be possible for machine-learning solutions in quality control of posts and checking of offensive content. We are not there yet, but the inroads are significant.
Well that's a big problem. Once you mechanize it, then there isn't even anyone to appeal to when you are subject to an unjust outcome. I wrote at length on this post above, but do we really want computers deciding who is or is not allowed to take their clothes off?
Fair point. But at least it will be objective. Applied to everyone with the same algorithm. And yes, the algorithm can get improved.
Makes me think of some semi-dystopian world like The Giver.
Reminds me of the Buckminster Fuller quote:

>We keep inventing jobs because of this false idea that everybody has to be employed at some kind of drudgery because, according to Malthusian Darwinian theory he must justify his right to exist. So we have inspectors of inspectors and people making instruments for inspectors to inspect inspectors.

It's a communication system used by billions, with a ratio of one moderator to hundreds of thousands.

Factory worker or a farmer would never produce anything for so many people.

What worries me is how these gatekeepers will be chosen. Will there be criteria that makes sure they aren't just applying their own biases? Will they come from all sides of the political spectrum, or will one political leaning be better represented? Will political content be reviewed by more than one reviewer?
I can't find it anymore, but there was a leaked pdf of their review guidelines. It was about 50 powerpoint slides and it's certainly more specific than "delete what you don't like". They had quite a few examples, usually in pairs (one to delete, one not to) to highlight the demarcation. I seem to remember stuff like "Sean Connery is stupid" (allowed) vs. "Sean Connery is stupid, like all scots" (delete, because it insults him based on his nationality).

(Not sure about the specific example–"stupid" may be too weak an insult)

They also devoted a surprising amount of pages for when to delete pictures of people urinating. Must be a much bigger problem than I would have thought.

Randomly. They'll just be the content janitors, and I doubt they will be subjected to anything more than the most rudimentary ethical screening or training.

I won't be a bit surprised if there's a systematic effort to encourage applications from within socially conservative entities such as religious groups, and going to be carefully monitoring their corner of the web for such.

Biased "junior level inquisitors" might not be all that bad as long as they represent a spectrum of biases and are reliably kept from specializing on opinion bubbles relevant to their personal bias. In any case, Facebook would be well advised to put some greying neckbeards in charge who remember everything about /. metamoderation.
> removing things we don't allow on Facebook like hate speech

or: How I Learned to Stop Worrying and Love Censorship

Not on Facebook for just over 2 years now. No regrets. Web forums and sites with individual focuses are such better communities in general.
This is sort of ridiculous, they can't seriously be expected to be held responsible for every video served on their platform. Nor should there be anything inherently worse about live-streaming of violence than that violence occurring in the first place.

I can only see this as a good thing if they manage to catch people before the act and intervene. Is this a primary goal of the program?

Yeah, random acts of violence getting exposure as they're live-streamed on Facebook has been interesting, to say the least. Without the criminal's willful publication of their act, the crimes may have gone unsolved and ignored for a very long time.

I'm not sure that we really need to stop people from publishing videos of themselves executing old men at random. It certainly seems to lead to a speedy resolution and really heightens the impact over the impersonal "Another shooting today..." on the evening news. It's virtually the same as walking yourself into the police station, with the added bonus that the whole country now hates you and is looking for you.

Without an audience the crimes might never have happened in the first place.
I don't know, these types of crimes have happened for a long time sans audience. Perhaps some are motivated by the audience, but there are definitely some who would've committed the crime anyway, and their broadcast of their criminal activity only speeds up their arrest.

I can see that politicians and businesspeople would not like this development, because it makes a) their cities seem more violent, since we can actually see someone get killed in it; b) their businesses seem less positive and soft and fluffy, since we can actually see someone get killed on it.

However, I don't see why those concerns should override the public safety benefit of allowing criminals to broadcast their crimes to thousands of witnesses and simultaneously preserving an indisputable copy of the evidence for posterity.

Steve Stevens was caught only because the exposure from his broadcast allowed him to be recognized by a McDonald's clerk hundreds of miles away from the crime scene. Without his broadcast, assuming the execution would've occurred, it would've just been another routine story that wouldn't have made it past the local news.

There would probably have to be some real research conducted to find out if this is a net loss by spurring new crime or a net win by allowing us to easily capture the most oblivious criminals (who are not necessarily least dangerous).

that seems highly unlikely to me, given the abundant history of covert violence. Obviously it's hard to assess for individual cases but across a distribution people seem to avoid notoriety.

Perhaps we can make inferences by parallel; trolls engage in obnoxious behavior when granted anonymity or insulation from direct consequences. Is it likely that absent an audience for their trolling, they would be nice all the time?

Seems unlikely, and the historical record suggests that cruelty, criminality, and impunity are a common combination.

The primary goal of the program is to make it look like they are doing something about the problem. Whether it works or not is a different story entirely.
I predict that Facebook's current PR problem related to this is going to be replaced by a swatting problem before long. I.e. someone reports someone on Live as suicidal, SWAT/police show up, shoot their dog etc, public outrage against Facebook ensues.
I wonder if this will reduce the problem of fake accounts. I regularly get such friend requests, and it starts to get annoying.

Also, those seriously affect the attractiveness of ad campains. I dipped my toes into it once, but it looks like a large percentage of gained "users" are just fake ones...

My wife made an account for our dog, with an appropriate headshot as the profile photo, about 9 years ago. He posts regularly, and we tag him in photos.

My guess is that they only block fake accounts if they are manually reported by someone.

You would think this, but as a counter example Xbox (back in the 360 days) had a "report gamertag as offensive" (aka offensive username).

Users could pay $ to change their gamertag (~$10).

It was fairly common to see people publically asking "please report my name as offensive" as it was known that if you got enough reports, you would be "forced" to change your name for free. People who wanted to change their name anyway would solicit reports against themselves to avoid a $10 fee.

Clearly no human was reviewing these reports... and that was a paid service!

That's true. I know a lot of people with a lot of fake accounts for entirely legitimate reasons, as well as some trolls who abuse the ease with which they can be created. A good number of people in the first group maintain multiple sock puppets because their subject to frequent attacks by trolls who report accounts to under false pretences, knowing that FB's response is as likely to be wrong as right.

Really, what's needed is a decentralized approach where no central authority exists, only mutual reinforcements. I have found a marvellous solution for the next generation of social and informational networking but alas this comment field is too small to contain it.

In parts of Eastern Europe, Asia, South America, and Africa, Facebook could hire 3,000 college educated employees for less than $3M monthly, including taxes. So while this from a human point of view is a very decent thing to do, it's not necessarily as costly as one might think, especially not when considering the possible liability or regulatory backlash they may run into if they do nothing and Facebook becomes the place for suicides and violence.

Now, if they add all those jobs in SF, it's another thing entirely.

Dear mods, I enjoy TLAs as much as the next geek, but can we please change CO to Community Operations?

edit - that was quick, thanks!

What's weird is that Facebook cannot rely on their users to report blatantly criminal acts witnessed by thousands of people. It probably says more about Facebook users than the platform and makes me doubt that doubling this or that team size can make a meaningful difference.

Especially with this approach of manual monitoring which will probably just result in more questionable deletions Facebook is already known for.

>What's weird is that Facebook cannot rely on their users to report blatantly criminal acts witnessed by thousands of people...and makes me doubt that doubling this or that team size can make a meaningful difference.

Then wouldn't it make sense to add more people whose job specifically is to report criminal acts instead of relying on the users to do it?

Are the reviewers going to browse random content? Somewhat possible with trend detection (make a human look at quickly rising stuff) but very wasteful since most of that will be either benign or incomprehensible to the outsider.

Will they watch private streams and read private messages? That sounds like a privacy disaster even Facebook would prefer to avoid.

Which leaves them with reports from users. Basically, reviewing millions of "hate speech" reports and either effectively instituting a very strict speech code or ignoring vast majority of them leading to further complaints.

Maybe I'm missing something but I don't quite see the point of rebutting issues you raise through your own speculation.
"report blatantly criminal acts"

The marketing campaign will be against murderers, the implementation will be against people who voted for Trump.

Why would they do that? I'm sure Trump voters click on adverts as much as anyone else.
Easier to sell ads if users don't have to deal with differing opinions. Arguments are hard to avoid and disrupt the casual consumption experience.

Conservative voters probably represent a much smaller portion of facebook's user base than liberal voters, given facebook's age demographics.

There is definitely an assymetry in what kind of hard-line political pages are likely to get banned. They have relaxed a bit in the last few months, but many political pages constantly have to worry about "getting zucced".

I imagine that plenty of people 'report' people after disagreeing with them politically.
Zuckerburg has political ambitions, a little help from the Facebook moderation team might earn him a few favors. He certainly wants that infrastructure in place once he runs.
[warning, mild politicking]

Don't worry; now that Trump has turned out to be just another shill for Globalism, Facebook will have no problem with him or anybody who continues to support him.

I come from significant privilege, so I won't be hurt by people dividing (and fighting among) themselves by ambiguous lines drawn in the sand. But, I don't understand how so many people that don't come from such a background think of globalism as a bad thing.
Because Globalism means the extremely poor elsewhere in the world get less poor, while the relatively rich in the US get relatively less rich, and also the extremely rich in the US and elsewhere get more rich. This does not jive with a lot of people.
Observation shows the first part is promised but does not happen. All the money goes in one pocket, and as a hint to which, its not the lower or middle classes of any country.
Users do report posts, from what I understand quiet regularly. It still requires manual moderation tho otherwise the reporting process can be abused (think anti-competitive purposes)

I've yet to hear of a way that this is solved without hiring thousands of people in what are horrible jobs (see Adrian Chen's excellent reporting on the issue[0][1])

I have no doubt that this will eventually be solves or assisted with ML, and that the solution is likely to come out of FB or G

[0] https://www.wired.com/2014/10/content-moderation/

[1] http://www.newyorker.com/tech/elements/the-human-toll-of-pro...

Those Chen articles hit close to home when I read them, because I've had to verify and act on child pornography complaints at a hosting service. Almost eight years later, I still have nightmares about the (thankfully) little I've seen. When I got into hosting I didn't realize dealing with stuff like that is table stakes; that was my first, and last, hosting employment, but I'll carry that aspect until I die.

That experience really makes me feel bad for the 3,000 new hires, honestly. I couldn't imagine moderating the human condition, which one could argue Facebook basically is. They'd better not clean out Adecco, and actually pay those people with the long-term damage of the job in mind, but that won't happen. Would actually make for a good union...

Completely sympathize. I had an experience in my black hat days - broke into a server and found folder upon folder of JPG's. Stupidly downloaded some of them and opened the first to find an image so disturbing that I can't even begin to describe it.

We were a bit conflicted about what to do (more how to do it), and ended up reporting it to both the US and Australian feds (which I suspect may have given me a free pass on one of the crazier things I later did).

I really didn't take it well, but one of the guys in our group was inspired to start a vigilante group that would hunt these distribution networks down and it achieved some success in the 90s.

Hopefully these employees can be eventually protected with some basic level of ML that would filter out the worst of the worst (apparently Microsoft Research have a well-developed fingerprinting system for child exploitation images) - because i'd really hate to imagine the scenario you and Chen describe, and what I briefly experienced, as becoming more common.

>Users do report posts, from what I understand quiet regularly.

Yes, they do - even when the post has nothing wrong.

Perfectly legitimate accounts writing reasonable posts, without any "fake news" content, or any "hate", are reported, and accounts are blocked.

This is a result of political activism by users. A very hard thing to solve for someone like Facebook.

Based on my experience reporting security issues to them and their condescending and irrelevant responses, I would just call the police. I wouldn't bother with Facebook.
It's a design problem too. If it's not obvious that you can/should report something then you're never going to do it. Plus if there isn't enough feedback about your help then you won't feel like you've made a difference.
(comment deleted)
In many places in the western world, facebook users are some 90 percent of everybody in a certain age group. So if it says anything about Facebook users, it says something about everybody.

Facebook is approaching 2B users. Let's say that every 1/1,000 users per day posts something that somebody else flags or which the system flags because of the words used. Then it's some 2M posts or comments that are flagged.

Facebook currently has 4,500 in the community operations departments ("moderators"). Each moderator then has to screen around 450 posts, which is difficult in an 8 hour shift. So obviously, that's not the way it works. They surely have algorithms that sort flagged posts with higher priority; more people flagging means it's more urgent, certain profiles are more urgent, for example people who have used violent language before, users with certain friends are more urgent, users at a certain age are more urgent, certain hours of the day are more urgent, post with certain words associated with violence or suicide are more urgent than posts with racist or sexist slurs or nude material etc.

But even if they correctly identify a threatening suicide or hate crime, how do they prevent it? Shutting down a live video is one way, but contacting authorities would probably also be necessary. How do you do that when your users are spread over 100,000 different jurisdictions? It's a big task.

> Facebook is approaching 2B users.

If you count also their sister properties including FB Chat and WhatsApp and Instagram. But Facebook the social network hasn't 2B monthly active user, as many left or rarely return.

It's not something I have thought about before, but I wonder how prevalent the bystander effect [0] is in online media streams. Say, if there are 1000 people watching a live stream on Facebook of a horrendous crime, will no one report it because everyone thinks someone else will report it? If that's the case, Facebook is up against human psychology in hoping people will point out these acts.

[0]: https://en.wikipedia.org/wiki/Bystander_effect

I'd expect the lack of actually seeing the bystanders to weaken the bystander effect.

Lack of physical proximity to what is happen might lessen the urge to act. Pressing 'report abuse' is really easy, but calling the police based on online material is a lot harder.

Machine Learning and AI are dropping the ball here?
Slashdot’s moderation and meta-moderation system (https://slashdot.org/moderation.shtml) always comes to mind. Could something like that work for Facebook self-moderation?
yes, i always think of this. Why is this system not in use at Reddit or HN? Is it patented? I would pay 1000 euro's if they could implement it for my facebook group, with 2000 members it takes an hour a day which would halved with such as system, 14 hours a month.
Even without the meta-moderation aspect, just having multiple dimensions for evaluating any given post is so useful and seemingly unique.

"Upvote/Downvote" is not nearly as useful for ranking as the range of, say "+1 Funny, +1 Insightful, +1 Agree, -1 Disagree, -1 Spam, -1 Flamebait/Troll" (I've forgotten exactly what set /. uses)

We laughed when China added x thousands to their great firewall content moderation team.
Mr zuckerberg's involvement signals it is a huge problem, one unable to spin away..

People always figure out a way around the latest attempt to control - and these measures will be overcome by bad intenders.

So if FB admits it has a community safety and fake news problem - by hiring 3000 additional enforcement agents, why would one stay with FB if there is an alternative. Which there isn't.

I would like a FB lite. Photo sharing and comments and discovery of old friends. No news. No menu of features.

I wish there could be an alternative. ("Its just like. fB but without the news and crap)

I'm not quite sure I follow. I only see news posted by my friends and the pages I follow. What would FB lite look like?
I think some alternatives do exist, but I do not believe an near exact clone of Facebook with less of the problems of Facebook exists.

Maybe it's just infeasible to properly monitor such a massive social network? I don't know for sure, but I left Facebook long ago and have not regretted it at all.

I suspect the long term plan is to create a training dataset labelled by the 3000 people and, when they have sufficient training data, let machine learning / AI take over.
Do things that don't scale... Until you can scale. I think you are absolutely correct on this point.
I assumed this too, those datasets would worth too many $$$
Imagine the horror that awaits the 3000.
100% automated ML doing this well seems far off to me.

Google's ML for rich snippets thinks a quarter is worth 50 cents. http://imgur.com/1nNreR2

Really parsing language for meaning is hard. Sentiment analysis, which seems simpler than this problem, still has relatively poor accuracy.

I could see it perhaps thinning out how much has to be done manually, but I don't see it removing that need entirely.

That information is so bad for such an easy question, I can only wonder if google is more interested in using their userbase as a training ground than actually providing a useful service.
It's not an outlier either. I can find about 1 obviously bad answer a minute just randomly typing fact type questions that pop into my head.
That's amazing. Nearly every fact presented there is completely wrong.

Plus, the picture is a standing liberty quarter that is nearly 90 years old and pretty much completely out of general circulation.

Could you specifically address his concerns for those not in-the-know about ML?
I believe "Nearly every fact presented there is completely wrong" was referring to Google's snippet in the screenshot, not my comment.
Sorry, my comment was ambiguous -- I was referring to the Google result. Thanks for pointing it out. The commenter that I was responding to made a great point that I agree with and was highlighting as a particularly good example of ML gone awry in the wild.

Other than the subjective statement that quarters are "useful", every fact in the google response is wrong.

Quarters are:

- Worth $0.25 (not $0.50)

- Not the largest. (Dollar coins and half-dollars are larger)

- Made of a blend of copper and nickel. (No silver content for circulating coins)

The picture depicts a standing liberty quarter that is no longer a circulating coin.

It's amazing that Google would put that in production.

I thought this screenshot may have been from an early and outdated rollout of the Rich Snippets, but no, I just searched and Google still says a quarter is worth 50 cents.
Easily explained as an artifact of just how accurate Google's ML systems really are.

What you're seeing is a hyper-accurate snapshot of the exchange rate between "a shave and a hair cut" and "two bits"[1]. Typically you hear about the value over a much longer period of time and the average value of "a U.S. Quarter Dollar" == $0.25.

/snark ;)

[1] "Two bits" -- Noun. Slang term used predominantly by Neo-Victorian communities living in California's Bay Area megalopolis.

Edit: Fixin' of the grammar

Last time facebook tried that they hired a bunch of lefties (collage kids are cheap labour) and they got pilloried for partisanship.
Are they kids that make artwork from newspaper and magazine clippings?
You are describing the past. That work is already done. All of these "community moderation" jobs at big companies are just checking the AI's work. If you have a billion posts an hour you're not hand triaging that with a team of 4500 people.
> These reviewers will also help us get better at removing things we don't allow on Facebook like hate speech and child exploitation. And we'll keep working with local community groups and law enforcement who are in the best position to help someone if they need it -- either because they're about to harm themselves, or because they're in danger from someone else.

The workers aren't there to be a basis for machine learning (though that is a nice benefit). They are there to communicate with local community groups / law enforcement on an ongoing basis.

This way you will have even less way to find a solution when you are hurt because of a false positive.

I mean, the GAFAs are already not known for their stellar customer service, but if we get rid of all humans in the process, it's going to be Brazil on steroid pretty soon.

Not optimistic but hopefully things will improve. As of now Facebook can't even remove an account for a hate blog local to my area.
Hopefully things will improve and we will see better things in our feed
I wonder if Facebook made much money on the ads displayed alongside this type of content (or perhaps pre- or mid- roll ads). Do they have a responsibility to treat this income differently?