More gruel
 
Something Weird Is Happening on Facebook

Something Weird Is Happening on Facebook

There’s nothing new about clickbait posts on Facebook. Social media is a relentless contest for attention, like junior high. We’re used to seeing a lot of harmless, banal mommy-chum like this:

Aw, how funny! Click, share.

This, however is a little different:

That first post generated 21 comments and 59 shares. The second one on the same hyper-twee recipe page generated 1.4 million comments and 35,000 shares. Yes, a question-post invites more engagement than a simple comment, but there’s something else at work here. We’re seeing a rash of posts like this soliciting a kind of engagement that would reveal valuable personal insights. In many cases they’re coming from pages purporting to be “blogs” which in fact are nodes in an “affiliate network.”

These affiliate networks are the new Tupperware or LuLaRoe, where housewives or hobbyists sell clicks instead of Amway. The humans in this network provide a veneer of authenticity. Network owners give them access to troves of thin content, usually recipes or vapid “lifestyle” tips, while cramming their pages to the gills with ads. Bloggers get a cut of the ad revenue, but here’s where it gets interesting. They aren’t doing a lot of selling. These social media and blog ads don’t appear to generate much revenue. Rarely are the blogs specifically selling anything on any scale. For that matter, they rarely write much either. This multi-billion dollar industry has to be getting revenue somewhere else.

It’s tough to tell where the money’s coming from because no one in this business would respond to inquiries, but there’s a likely answer. In the social media industry, data is the product. Facebook’s APIs allow me to pull some aggregate data but other methods let me scrape the profiles of people who interact with my posts. I can then collect and analyze their responses to my posts. If I have access to dozens or even hundreds of “affiliated” pages the potential reach becomes vast. Someone appears to be pouring enormous energy and effort into a data collection project aimed at building personality profiles from social media interaction…again. This has been tried before, most prominently by Cambridge Analytica back in 2016.

Cambridge Analytica was a short-lived political consulting firm famous for their role in Russian meddling in the US & British 2016 Elections. CA used data gathered in violation of Facebook’s terms of service to build complex voter profiles. They used that data to target political messages, often containing phony information, to users based on their “psychographic” profile. Cambridge helped the Trump campaign target misleading messages based on material stolen in the Russian hack of Democratic National Committee servers toward voters most likely to believe this fake news.

To build these profiles, CA promoted an online personality quiz through an app that required a Facebook login. They used that login to steal data via Facebook’s API which they tied to the respondents’ answers. Getting only 270,000 responses to their personality test, they were able to assemble profiles on a purported 87m+ social media accounts.

Facebook has shut down that particular avenue of data harvesting. There are disputes about whether CA’s voter profiles even worked. However, the dream of building psychological profiles of voters from patterns of social media engagement remains alive. Such an engine could power potent networks of political disinformation. Major companies have long had far more powerful targeting tools at their disposal, but those tools are too expensive for politicians and aren’t aimed squarely at the traits that drive political engagement. Political operatives need tools that are cheap and dirty. That brings us back to our affiliate networks.

Let’s take a look at another pair of posts, this time exactly the same meme on the same page, posted a couple of weeks apart.

Same post, with the same meme, on the same page, a couple of weeks apart produces results that differ by orders of magnitude. It could be that the second one is just newer, but based on the second post’s engagement trajectory that doesn’t seem to explain the difference. Scrolling through other memes on the page another pattern emerges. The difference appears to be the “with” accounts, other social media entities “tagged” to assist with promotion.

What are those accounts and why do they matter? It isn’t clear and there’s no definite pattern. A lot them are, for lack of a better term, squirrelly. They seem to be “promoter accounts,” generating dozens of posts of day consisting mostly of spam. Some are offshore. A few appear to be ordinary people. Why these tagged accounts deliver such huge impact is unclear, but the results are impressive. How this works is murky, but it’s clear that someone has found a way to gain absolutely stellar reach for these apparent spam posts. What they have in common is help from affiliate networks.

If you spend any time on Facebook you’ve probably noticed a blizzard of question memes coming from clickbait accounts. You’ve likely either commented on them yourself or seen comments from close friends. Many of these posts look like they’re probing for answers to security/verification questions, but the ugly reality is that your passwords are nearly worthless. Chances are your passwords are already circulating on the dark web, sold in batches of millions for as little as a few thousand dollars. Unless you hold the password to something wildly valuable, like major corporate or government assets, nobody cares except kids playing around.

By contrast, what could I learn about someone by knowing their answers to these questions?

How valuable would it be to build up profiles of millions of social media accounts based on these answers, while also learning the contours of their social networks? Feed that kind of sentiment data through a machine learning algorithm and I could build a dataset on which to build a powerful disinformation campaign. Someone seems to think there’s value here, investing significant time and money by recruiting affiliate networks to distribute this content and gather results.

Some of these posts seem more obviously aimed at personality data:

Going a little further, there’s a tranche of questions circulating that look more clearly political:

How old are you? Where do you live? What are your entertainment tastes and how far do you live from your hometown? Properly scrubbed, these answers could probably predict your ’16 and ’20 Election preferences with 90%+ accuracy. Sure, that first post won’t accurately predict your birth year, but that’s the point. Look at the comments (running into hundreds of thousands) and you’ll see people simply posting that data, plus colorful details. Thanks to advances in sentiment analysis and the declining cost of computation power, I can get remarkable insights from masses of text-based data.

What data do these posts produce? For each account that interacts with these posts, page administrators can use Facebook Insights to query aggregate information on post reach and demographics, but that’s for rookies. With a little python expertise responses on my page can be scraped directly from Facebook and stored for analysis. There are legitimate paid services that will do this, but this would likely be an in-house capability for social media affiliate networks. Janie in Memphis with her mommy blog or Steve in Phoenix with his RV musings might enjoy a little extra money each month from their social media engagement. Meanwhile the larger network with administrator access to their accounts is collecting the real gold in the form of individual user data.

Without more research, or perhaps even a subpoena, it’s impossible to nail down exactly what’s happening here, but there are some clear precedents. If these affiliate networks aren’t being used to improve on CA’s voter profiling project it would be a surprise.

Don’t take candy from strangers and don’t feed your personal information to bots. If that mommy blog is jammed with ads and promotions, leave immediately. For context, compare the layout and content on an authentic blog like Scary Mommy with the spam, ads, and brain-numbing botspeak you’ll encounter on a MediaVine blog like the A Typical Mom. The difference is easy to spot.

Watch this space. There’s a pretty good chance that the big data scandal of the 2024 Election is unfolding on your Facebook feed right now.

15 Comments

  1. Other weird things happening with Facebook include a product manager leaking internal documents to the Wall Street Journal and now being down for about 4 hours internationally today.

    The only reason I bother to comment about this is that I remember the heady early days of “the information superhighway” where the theory was that the network distribution made it impossible to pull things down. It’s interesting to see downages among such big, powerful, and highly distributed companies, but thems the times.

    1. Scary Mommy has blog entries obviously created by hand, by a person, with the intent to communicate some sort of insight. Meanwhile, Typical Mom, like all MediaVine affiliates, is a junkyard of cribbed content, dressed up here and there with a few lines of third grade prose. It’s brain-hammer. Sample: “It’s September, which means all things pumpkin in our house! Ok, who’s lying..I make things like our 2 ingredient pumpkin muffins and these year round.”

  2. Generally speaking, people should become more and more wary of memes.

    Anything that requires interaction and response is data-mining, sure, but anything that’s just for reaction and sharing is compression at best ( https://www.epsilontheory.com/the-elton-hootie-line/ ) and direct misinformation campaign at worst.

    I would like to say people should avoid memes entirely, but the problem is they are built into the environment of social media and basically any social media interaction you make has the capability of becoming a meme. Or another way of saying that is that all social media interaction is meme-making of different total reach.

    A non-intellectual way to say this is that telling people not to interact with memes is like telling people not to put bumper stickers on their car or wear branded tee-shirts.

    ————————-

    Aside: when I was a teenager first logging onto the Internet I was told under no uncertain terms never to give personal information away, even to the point of hobbies and interests. Social media switched online ‘trust’ to those who have Real Names (c)(tm) and it’s always been uncomfortable for me. Normally I’d posit a theory that maybe we would switch to go the other direction again due to these scaled privacy violations, but the fact is almost all large institutions benefit from having individual citizen/consumer/members identity-tagged as a systems and logistics process of their own, so all larger ‘privacy’ rights are no longer about individual choices but about institutional restraints.

    In other words, laws and regulations.

  3. It is not much of a stretch these days to conspiracy theory merging with fact.

    So who would benefit from gaining so much information on Facebook participants, has already been documented with attacking the u.s. via the election process, and has the resources to manage such a endeavour?

    The russians we know can and will exploit such tools, and I can’t imagine the chinese are not equally capable. Imagine the opportunities globally for those enemies as the u.s. self-immolates. No, self-immolation is perhaps not accurate. How much of what is happening in democracies all over the planet is due to external attacks?

    The world has witnessed the damage the russian’s useful idiot did to the u.s. and the world in his 4 plus years or tyranny and disfunction. Why would any enemy nation stop now, when the spineless and incompetent dems, who supposedly control the reins in Washington at the moment, will not make a peep about it?

    Imagine Biden making an address on national television, saying “Our nation, and democracies all over the world, are under attack, not just from overt cyber-attacks, but from insidious psy ops using the internet. As of today, we are putting all enemies of the u.s. on notice. We too have resources. We will track down the sources of these attacks, which we consider acts of war. If these are independent actors, we will treat them as terrorists. If these actors are backed by states, said states will suffer the consequences the same as if they bombed Washington.”

    As I said, imagine Biden making that statement. Given the utter lack of resolve demonstrated by the current regime in all areas defending the ramparts of democracy, I can’t.

  4. They are using Facebook to grow right-wing conspiracy theorists who don’t believe in “government” and only believe whomever is the current Republican cult leader.

    Find ’em, feed ’em bullshit to destroy their critical thinking abilities, and then train them to vote Republican and buy guns.

    It’s not a real mystery of where this (has been) heading.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.