David Broniatowski

Explaining Twitter’s inability to effectively moderate content during the COVID-19 pandemic - Scientific Reports Scientific Reports - Explaining Twitter’s inability to effectively moderate content during the COVID-19 pandemic

Takeaway: We can't just play "whack-a-mole" with content. To build healthier online spaces, we have to start being architects.

Thanks to @knightfoundation.org for financial support!

Our full paper is here: www.nature.com/articles/s41...

a man with a beard is sitting in front of a wall of televisions ALT: a man with a beard is sitting in front of a wall of televisions

The real culprit? Platform Architecture.

The fundamental design of the platform—its rules, its affordances, its very blueprint—allowed these decentralized communities to form and thrive. The system itself was perfectly built to resist the very kind of top-down control that was attempted.

a cartoon of a man looking out a window at the stars ALT: a cartoon of a man looking out a window at the stars

So why did some people think it was working?

If you only look at the immediate neighborhood of banned accounts, it seems like a success. But when we zoomed out, we saw the problem was just growing and shifting across the wider network.

a woman is sitting at a table with the words be prepared for backlash on her face ALT: a woman is sitting at a table with the words be prepared for backlash on her face

We found that the top-down content moderation policies actually triggered "reactance"—a powerful backlash that strengthened the targeted communities.

Instead of silencing them, the bans became a threat that reinforced their identity and made them more resilient.

homer simpson from the simpsons is standing in front of a shelf with stuffed animals and says d ' oh . ALT: homer simpson from the simpsons is standing in front of a shelf with stuffed animals and says d ' oh .

Did Twitter's big crackdown on COVID-19 misinformation—like banning "superspreaders"—actually work?

We ran an ecosystem-level analysis to find out. Spoiler: It not only failed, it was often counterproductive. 😬

Mark Dredze named director of Johns Hopkins Data Science and AI Institute Dredze, a member of JHU’s faculty since 2009, has been selected to lead the university institute dedicated to harnessing the power of AI to translate data-driven discovery into real-world impact.

Congratulations again to John C. Malone Professor of Computer Science @mdredze.bsky.social on this accomplishment!

Thanks for the note! Your work absolutely inspired ours and the “one size fits all” threshold refers to another paper by another set of authors.

a group of power rangers standing next to each other with the words `` super team '' above them . ALT: a group of power rangers standing next to each other with the words `` super team '' above them .

Proud to work with an amazing team: @yunkangyang.bsky.social , Ramesh Paudel, Jordan McShan, @matthindman.bsky.social , and Howie Huang.

And grateful to the Knight Foundation for supporting work on transparent methods for information integrity.

Full paper: doi.org/10.1038/s415...

This dual-detection strategy is key to spotting the more subtle stuff. This paper builds on our broader interest in platform dynamics and misinformation. But the method is applicable to any kind of content diffusion—hashtags, images, even AI-generated text.

a baby yoda from star wars is sitting in a crib . ALT: a baby yoda from star wars is sitting in a crib .

📊 About 23% of pages showed signs of coordination—some of it expected (e.g., news syndication), but a lot of it less transparent.

The method helped us identify both overt and covert networks. Some shared links every few seconds. Others did it slowly but repeatedly, evading the usual red flags.

a woman wearing a blue hat is talking into an abc sports spot microphone . ALT: a woman wearing a blue hat is talking into an abc sports spot microphone .

Using mixture models, we derive empirically grounded thresholds that tell us when the pattern is too consistent to be chance.

No guesswork. No black box.

We applied this to over 11 million Facebook posts from 16K high-engagement pages.

a cartoon horse is sitting in the back seat of a car and says do you think we 're maybe going too fast ALT: a cartoon horse is sitting in the back seat of a car and says do you think we 're maybe going too fast

Most detection methods use arbitrary cutoffs (like “shared within 10 seconds”), which don’t hold up across platforms or contexts.

We do better.

Our method looks at two things:
1️⃣ How fast two pages share the same link
2️⃣ How often they do it over time

two men are standing next to each other in a room and talking . ALT: two men are standing next to each other in a room and talking .

Our approach is platform-agnostic, scalable, and designed to support transparency and accountability. Why does this matter?

Coordinated sharing—whether by media groups, advocacy networks, or covert actors—can shape what people see online.

Coordinated link sharing on Facebook - Scientific Reports Scientific Reports - Coordinated link sharing on Facebook

Just published: our new paper in Scientific Reports
📄 “Coordinated Link Sharing on Facebook”
doi.org/10.1038/s415...

We introduce a statistically grounded, human-interpretable method to detect coordination on social media.

The primary donor to #openscience seems very supportive of massive cuts to scientific funding that jeopardize research. I don’t think it’s edgy at this point to say that reform is being weaponized and the community needs to stand up against it.

David Broniatowski on LinkedIn: NIH has just announced that it will save $4 billion by capping university… | 61 comments NIH has just announced that it will save $4 billion by capping university indirect costs on federal grants. Will this actually save money? Bottom line: No. It… | 61 comments on LinkedIn

Please read and share this excellent FAQ on University indirect costs by my friend @broniatowski.bsky.social

He explains why these funds are essential and a critical investment for research in the United States.

www.linkedin.com/posts/david-...

Our new publication asking whether ChatGPT chatbots can reliably give advice to help people quit smoking? We examined chatbots including WHO's Sarah in providing assistance on quitting smoking. @who.int @broniatowski.bsky.social

This project -- led by
@lorien.bsky.social , in collaboration with
Artin Yousefi, Christina Wysota, and Tien-Chin Wu -- was supported by a TRAILS seed grant.

Better AI can help people make more informed, healthier decisions! 🚭🤖

a close up of a robot 's face with the words `` calibrating '' written on it . ALT: a close up of a robot 's face with the words `` calibrating '' written on it .

7️⃣ So, what’s the takeaway?

🔹 Chatbots can be helpful, but they need better training.
🔹 WHO’s Sarah outperformed the others.
🔹 AI can easily be prompted to give weird advice. 😵‍💫

an old man with a beard is talking to a robot and says `` your jedi mind tricks don t work on me '' . ALT: an old man with a beard is talking to a robot and says `` your jedi mind tricks don t work on me '' .

6️⃣ Surprisingly, the chatbots resisted adversarial attacks!

Even with tricky prompts, they stayed on topic & avoided harmful advice.

a robot with the number 11 on its back ALT: a robot with the number 11 on its back

5️⃣ Queries on quitting "cold turkey" or with vapes, gummies, or hypnosis? The chatbots STRUGGLED. 🥴

a man is holding a key in his mouth and smoking a cigarette . ALT: a man is holding a key in his mouth and smoking a cigarette .

4️⃣ Bad news: Some responses contained unsubstantiated claims (22%) 😱

For example, an AI recommended using a necklace to quit smoking. 😭

a close up of a man in a spiderman costume with a beard . ALT: a close up of a man in a spiderman costume with a beard .

3️⃣ Good news: Chatbots mostly recommended professional help (80%) and used clear language (97%).

loki giving a thumbs up and saying `` you had one job '' . ALT: loki giving a thumbs up and saying `` you had one job '' .

2️⃣ Across chatbots, responses followed guidelines only 57% of the time.

Sarah led with 72% adherence, while BeFreeGPT & BasicGPT lagged at ~50%. 😬

a man with a mustache is wearing a yellow and green shirt and says well let 's find out . ALT: a man with a mustache is wearing a yellow and green shirt and says well let 's find out .

1️⃣ First, we tested three AI chatbots: WHO’s Sarah, BeFreeGPT, and BasicGPT.

Do they give reliable quitting advice? Let’s find out! 🔎

Assessing the Adherence of ChatGPT Chatbots to Public Health Guidelines for Smoking Cessation: Content Analysis Background: Large language model (LLM) artificial intelligence chatbots using generative language can offer smoking cessation information and advice. However, little is known about the reliability of ...

🚨 New paper alert! 🚨

Ever wondered if AI chatbots actually follow public health guidelines when giving smoking cessation advice? 🤖🚭

Our latest study in J Med Internet Res evaluates ChatGPT-based chatbots on this! 🧵👇

📜 Paper: www.jmir.org/2025/1/e66896

For more insights into platform design, systems engineering, and what this means for the future of online content, read our full preprint: dx.doi.org/10.21203/rs....
#SystemsEngineering #SocialMedia

a man is standing in front of a house in a black and white photo . ALT: a man is standing in front of a house in a black and white photo .

Platforms cannot tack on moderation policies after the fact and expect success. It’s like trying to build the roof before the foundation.

a man in a suit and tie making a face next to another man ALT: a man in a suit and tie making a face next to another man

The risks:
🕒 Best case? Wasted time and money.
⚖️ Worst case?
- Liability
- Reputational harm
- Undermines users’ rights & agency through bad information

a cartoon of two spidermans standing in front of a nypd van ALT: a cartoon of two spidermans standing in front of a nypd van

Why does this matter now?

Because other platforms (such as this one!) aim to replicate Twitter’s architecture while promising moderation. Spoiler: They may face the same issues.