With those comment counts, something dodgy is obviously happening.
The interesting question here is whether Facebook is somehow accidentally amplifying it. Certainly it is not in Facebook's interest to allow this kind of data harvesting. If it hurts you to think that Facebook somehow isn't maximally evil, at least consider that this is data that could be only Facebook's. Allowing somebody else to harvest it is money straight out of Facebook's pocket.
So, given FB should not be complicit, what mistake could they be making to allow the system to be haunted? The obvious guess is that they have a feedback loop in the ranking algorithm. It values comments very highly as a signal of good engagement, but they weren't prepared for "content" that is this good at eliciting low effort comments and have wide appeal demographically. As long as one of these reaches a critical mass, it'll be shown to tens or hundreds of millions of people just by any engagement feeding even more engagement.
> The interesting question here is whether Facebook is somehow accidentally amplifying it.
> Certainly it is not in Facebook's interest to allow this kind of data harvesting
Mark Zuckerberg has been quoted through leaked documents to be a strong purveyor of "engagement" at nearly all costs. I don't think giving Facebook the term "accidental" is appropriate anymore. Their desire for engagement trumps the health of their network. I'll dig through my favorited submissions for the WSJ article.
People are rarely motivated by evil, but they are motivated by opportunity to which an outcome can be perceived as pure evil by the people it affects most.
Engagement is a really dirty word to me nowadays - the attention economy comes with all sorts of really bad side effects. We’ve turned almost all conversations into an ad in order to sell more ads. It’s you vs 100 people with a doctorate in psychology at any given time.
Agreed. I've actively taken steps to combat it, only use my phone in black and white mode, and removed all apps. Stick to my PC for general browsing and my phone use as gone down to 30 minutes a day (mostly calls, messages with the wife, and email).
Here's a funny thing: I did most of the things you did, but I kept going back to twitter et al using my browser. The biggest change, though, was removing my ad blocker. Being confronted with tons of garbage ads forced me to stop using these services faster than the minor inconveniences like deleting the app.
Hey....now that's Adblock usage I somehow overlook.
I remember can't browse IG more than a couple minutes on a phone because there is an ads for every 5 pictures, but I can browse it mindlessly on my PC with Adblock installed, because the experience is so clean.
It's absolutely in Facebook's interest to allow this kind of viral garbage. It feels like engagement, gives people a way to engage socially in seemingly harmless questions. (I'm not at all convinced these are some dastardly data gathering scheme - many of the viral questions don't have meaningful answers.)
It would be so, so simple to stop these. Just de-prioritize posts with too many replies, or replies from people you don't know, or.. anything. The virality of these things sticks out like a sore thumb. Facebook is choosing to not stop them.
I’m sure some are but most are just trying to build a large following with this cheap viral garbage. If 1% of the respondents “Like” the page and share their stuff later then that’s quite a good audience builder
I've seen one that asks the last 3 digits of your phone number. Not sure how useful that one is but it could probably positively ID someone with a reasonably unique name.
My feed has become littered with these things because my friends are replying to them. Which encourages me (or, rather, the hypothetical FB user) to reply to my friend(s).
Added to correlational hints as to my location, it would clarify six of ten digits of my phone number.
There are probably other clever ways to skeeve the other four, and I cannot say for certain if some digits of those other four are situationally impossible. Certainly some combinations likely are, like 0911, and 0000.
I always thought that FB is somewhat complicit (not out of evilness necessarily).
I see a lot of "viral" posts - some like those mentioned in the article, but also a ton of odd woodworking, cooking, and "resin art" videos. The videos are quite repetitive and not really interesting so I wonder if they are maybe hidden ads, but they are not marked as such, and it is not clear what they are selling. (Well maybe they are trying to sell resin, which is really expensive.)
Anyway, it seems like they are different kinds of posts on FB. Some stay close to their point of origin, and only rarely get shown to other people who have not liked a page or are friends themselves. And other posts which, if somebody commented on or interacted in any way with them, get shown to their friends and friends-of-friends.
After running a charitable cause / political FB page for a while, I'm convinced that internally there are actually different categories of posts - ones that are shown to followers, and ones that are allowed to float or go viral. I really wonder what the mechanism is to get into the floating category. It doesn't seem to be based on quality, nor on money spent. Maybe it is some interaction metric that somebody learned to game?
> I see a lot of "viral" posts - some like those mentioned in the article, but also a ton of odd woodworking, cooking, and "resin art" videos. The videos are quite repetitive and not really interesting so I wonder if they are maybe hidden ads, but they are not marked as such, and it is not clear what they are selling.
As someone who got caught up in some of those videos when I was in complete "mindlessly browse facebook" mode, my guess would be they are optimized for "engagement", nothing more, nothing less. They are just interesting enough that you want to know how the end result looks while harmless enough to appeal to a maximally broad audience.
They grow a Facebook page they can then use to push sponsored content, sell it off to the highest bidder or use it to brag about how they’re a successful content creator and sell a course on how to be as “successful” as them.
Have you seen these videos? I agree they look like they have been "optimized for engagement"... I guess someone could be doing that just for fun, that's your theory?
A quick point I’d make is that it may not be a mistake (from facebooks perspective) to allow others to exploit a given system as long as they’re gaining enough value from it to outweigh that. If whatever is being exploited doubles how much they can charge for ads, they might accept some of their data being stolen until they could find a way to have their cake and eat it too.
They recently started showing friends’ response comments first, so you can easily see those of your friends out of the 116,000 replies. The rise of these in my feed seemed to correspond with this feature.
Disclaimer, am infrequent FB user and this may have been around for longer than I realize.
It has been this way for years, so it's definitely no accident. Posts that elicit more comments absolutely end up getting ranked higher by The Algorithm™.
> The interesting question here is whether Facebook is somehow accidentally amplifying it.
Someone somewhere found a way to exploit what FB's engagement metrics do. Is it 'accidental' that FB amplifies things if their system is designed to do exactly what it does when gamed?
One thing that's nearly impossible to appreciate about social media platforms, even small ones, is their scale. Both in terms of users and content.
FB see about 5 billion items posted daily. Or about 58,000 per second Most of those simply die unread. Our Internet has become a write once, read never medium...
If an item is to be picked up by amplification algorithms, it needs some indicia of relevance or significance. Broadly, that comes from one of 3 propeties:
- Content itself. Keywords, hashtags, URLs, other profiles linked.
- Social graph. Followers and readers of the submitting account, and their own followers.
- Engagements and interactions. Any likes, comments, or re-shares, with their own attributes as well (content, social graph, engagement.
Given that signal for a naked submission is so thin, any indication of additional relevance is likely glommed on to, and a set of rapid initial engagements might be a sign of high-value content ... or of a bad-faith mutual-admiration-society cabal (MASC). Even on sites such as HN, a little early engagement on a submission goes a long way.
Note that a group of freinds engaging with one another's content will look a lot like a MASC, though in most cases the significant distinction will be posting volume. It's rare for a person to consistently post more than 10--30 items a day, and much above that tends to get seen as annoyingly verbose by others. Promotion and amplification accounts can post many tens, hundreds, or thousands of items, hoping one will take off. Their goal isn't engagement but manipulation, they have cheap content-creation processes (stock bits, redistributed content from other sites, randomly-generated crud), and can afford to be profligate. Until the system actively penalises based on submissions without significant uptake, that's going to be the case.
Why this is happening is anyone's guess, though in general, cultivating a capitive audience has value, whether for conventional advertising or propagandistic purposes. It may be that the accounts are intended for direct use or will be farmed off to other buyers or uses later.
Gaining profiling information on follower demographics and influence points would also be part of this.
Given political cycles, odds that this is prepatory for the US 2020 campaign seems a plausible explanation.
It's useful for FB to let an outside agency to getting backdoor data because it keeps them distanced from whatever outcome that group comes up with, with plausible deniability that FB "didn't allow" that group to mine their data.
>Certainly it is not in Facebook's interest to allow this kind of data harvesting.
How so? They benefit directly from it without having to do any of the hard work and they can put the blame on someone else if the whole thing blows up. Seems like the perfect crime.
"Allowing somebody else to harvest it is money straight out of Facebook's pocket."
In that metaphor, Facebook owns the soil and they sell the harvest by charging for promoted posts and other forms of engagement purchasing. If this engagement truly drives election outcomes the way the post hypothesizes, the demand side will come back every time there's an election anywhere. Let's hope Zuck and co understand how closely this piece of their revenue is tied to democratic engagement ;)
Isn't the obvious answer to stop scraping (or, at least try)? The author states that the value here is collecting data (not money from ads or something) and Facebook's APIs don't allow for the kind of analysis they'd need to build profiles. Article specifically talks about using Python to scrape profiles (presumably using a logged in account).
How does some tiny org harvesting some data harm FB? Will FB sell fewer ads because some third party extracts mother's maiden name, make of first car, home state, and whatever else is revealed in response text? Will the third party org serve ads to those people outside of FB? Or, will they use that information to hone a FB strategy, potentially meaning a net increase in the number of ads bought AND giving FB enough distance that it wouldn't be an actionable scandal when it was inevitably discovered (like with Cambridge Analytics)?
I really don't see much actual downside for FB in allowing this kind of data collection. The people (like myself) who are disgusted with FB and think FB is a profoundly negative force that is strangling local journalism, toxifying the discourse, and stroking rage are already against FB. But most people just don't care (and may actively appreciate those effects) and won't change their behavior in response to FB allowing another likely election-outcome-altering disinformation campaign to do recon work. And FB knows that. As long as Zuck feels safe from regulation, he's not going to stop, address, or reduce any non-CSAM thing that would threaten growth or time-in-app.
> It values comments very highly as a signal of good engagement, but they weren't prepared for "content" that is this good at eliciting low effort comments and have wide appeal demographically.
Fecebutt was built on content like this. On LiveJournal before it, this kind of personality-quiz stuff was also rampant, but personality-quiz apps and this kind of meme-question stuff dominated Fecebutt from the time it launched the "Facebook Platform" in 02007 for many years, maybe even until they banned personality-quiz apps last year.
The interesting question here is whether Facebook is somehow accidentally amplifying it. Certainly it is not in Facebook's interest to allow this kind of data harvesting. If it hurts you to think that Facebook somehow isn't maximally evil, at least consider that this is data that could be only Facebook's. Allowing somebody else to harvest it is money straight out of Facebook's pocket.
So, given FB should not be complicit, what mistake could they be making to allow the system to be haunted? The obvious guess is that they have a feedback loop in the ranking algorithm. It values comments very highly as a signal of good engagement, but they weren't prepared for "content" that is this good at eliciting low effort comments and have wide appeal demographically. As long as one of these reaches a critical mass, it'll be shown to tens or hundreds of millions of people just by any engagement feeding even more engagement.
Is there anything less obvious?