In August, internal research at Facebook showed that ~70% of the platform's most active civic Groups were considered too toxic to recommend to users
The giant struggled to balance Mark Zuckerberg's free-expression mantra against findings that rabid partisanship had overrun a feature central to its future
Wall Street JournalJeff Horwitz
Context & Ripple Effects
Facebook had already shelved internal research on its polarizing effects, while earlier enforcement efforts were reported to push toxic activity into private Groups and messaging services. The new finding matters because it identifies the problem inside a feature Facebook treated as central to its future, sharpening the conflict between growth through recommendations and Zuckerberg's free-expression posture.
First-order effects
Facebook's Groups and recommendation teams must contend with internal evidence that a large share of the most active civic Groups fails the platform's own threshold for recommendation.
Administrators of highly active civic Groups face a direct distribution risk: recommendation eligibility becomes distinct from their ability to remain on Facebook.
Second-order effects
Restricting recommendations without removing Groups can shift civic engagement toward private or less visible channels, repeating the monitoring challenge described in Facebook's earlier cleanup efforts.
Facebook's content-policy staff must make recommendation decisions that are more politically fraught than simple removal decisions, because they determine amplification rather than speech access.
Third-order effects
The case points toward interface-neutral platform governance: platforms increasingly have to govern what their ranking and recommendation systems amplify, not only what users are allowed to post.
If internal harm findings repeatedly conflict with engagement-oriented product priorities, the durable issue becomes whether platform governance can override incentives built into distribution systems.
The trend: Social platforms are moving from content-removal debates toward scrutiny of recommendation systems as the mechanism that scales civic toxicity.
New: FB researchers warned execs late last summer that 70% of the platform's most active civic Groups were toxic. Facebook quietly began reining Groups in - then scrambled as the measures weren't enough. https://www.wsj.com/...
This time last year Facebook was running a massive ad campaign to get users to join groups. By Aug, its own employees knew how problematic groups were. Pretty solid evidence that Zuckerberg and FB's leaders have no idea what happens on their platform. https://www.wsj.com/...
I've been writing ab Facebook Groups for 6 yrs: how amoral recommendations of certain groups + desire to create entirely new interest-based social graph for users ($) ultimately led to brigading and harassment, created QAnon, led to echo chambers. Now: https://www.wsj.com/...
Facebook was more ready to listen than it had been about past earnings on Groups. Disabling recommendation algorithms was a start. But then #StoptheSteal Groups blew up thanks to other growth-friendly product design features. The problem wasn't just “bad algorithms!”
@wiczipedia, @CCDHate and @Mozilla all raised prescient concerns about FB Groups - but the smarts/depth of FB's own research is breathtaking. Here's hoping it gets shared with outside academics one day, so talk of FB's societal impact doesn't depend on what falls off the truck.
“The researchers told executives that ‘enthusiastic calls for violence every day’ filled one 58,000-member Group, according to an internal presentation.” $FB https://twitter.com/...
Letting members invite thousands of users to new groups per day. Injecting “previews” of Group content in newsfeed, then auto-joining users who engage. Allowing Groups to get big before integrity systems kicked in. These were choices — that are now being rethought by FB.
Also have to wonder if Zuckerberg knew @JeffHorwitz's reporting was coming down the pipe when he announced on last week's earnings call that FB would permanently suspend political group recommendations. Certainly appears that way...
“70% of the top 100 most active US Civic Groups are considered non-recommendable for issues such as hate, misinfo, bullying & harassment,” $FB's staffers said in Aug. “Our existing integrity systems aren't addressing these issues.” https://www.wsj.com/... A+ story by @JeffHorwitz
The situation was really bad. “Financially motivated Albanians” repeatedly building up Groups with more followers than all but a handful of top-tier American newspapes have subscribers, for example. The most successful groups were toxic.
“Facebook's own research found that American Facebook Groups became a vector for the rabid partisanship and even calls for violence that inflamed the country after the election.” (surprising exactly no external observers) https://www.wsj.com/...
.@Facebook absolutely knew for years that Groups had a huge problem with extremism and calls for violence. We, and many other orgs and researchers called this out repeatedly with no action. Facebook only acts after the damage is done. https://www.wsj.com/...
Great reporting on Facebook Groups, Mark Zuckerberg's most blatant effort at social engineering which has been supercharging extremism and hate for years https://www.wsj.com/...
“Facebook executives were aware for years that tools fueling Groups' rapid growth presented an obstacle to their effort to build healthy online communities...” https://www.wsj.com/...
Yet another article about the stunning amorality of Zuckerberg, Sandberg, et al. Facebook has repeatedly & willfully harmed society without ever having to deal with real consequences. All senior leadership must be fired and the company needs multi-billion-dollar fines. https://tw…
Facebook's Groups were the “heart of the app,” CEO Mark Zuckerberg said in 2019. Then they became a vector for rabid partisanship and even calls to violence. Now Facebook is clamping down. https://www.wsj.com/...
This is a conclusion that @alexstamos said was needed in an interview with me last year: “We need to do something to stop these conversations from happening.” Facebook's data scientists warned that extremists were gathering in its Groups. https://www.wsj.com/...
Facebook keeps pivoting and keeps ending up in the same place. “We need to do something to stop these conversations from happening.” https://www.wsj.com/...
Antitrust seems the only way to fix this. Apparently, through WhatsApp's money transaction business & Jio, FB seeks to mine data in India. Even if you anonymize data (which lets face it it hardly ever happens except for tokenism) it constitutes serious threat to people's privacy.…
In 2019, Mark Zuckerberg called FB Groups a feature “central to the future of the app”. Now, it is clamping down on them after its own research found these groups to be a vector for “the rabid partisanship” and “even calls for violence after the election” https://www.wsj.com/...
.@accountabletech & @Mozilla led a coalition urging Facebook to #StopGroupRecs in September. we warned: “Facebook Groups pose a singular threat to this election season.” https://accountabletech.org/ ... @JeffHorwitz reveals FB's own researchers were warning the same. Mark didn't …
For some reason everything they focus on ends up amplifying the worst in humanity. First Newsfeed, then Groups. What should we focus on, and probably ruin, next? VR? https://www.cnbc.com/... https://twitter.com/...
Of course it did. 1) profits 2) government relations 3) public relations 4) audience integrity. Always seems to be that order. Always. https://twitter.com/... https://twitter.com/...