Social media feeds today are optimized for engagement, often leading to misalignment between users' intentions and technology use. In a new paper, we introduce Bonsai, a tool to create feeds based on stated preferences, rather than predicted engagement. arxiv.org/abs/2509.10776

I'm thinking a lot about the debate around whether LMs can "understand." Enter a new thought experiment I call "the lean Chinese room." IMO, it shows how the pressure of compression and speed blurs the line between symbol-shuffling and understanding. doomscrollingbabel.manoel.xyz/p/the-lean-c...

Whether or not AIs “truly” understand is less important than how they reshape jobs, politics, and culture. My new post: why the “Human of the Gaps” fallacy distracts us from governing real-world AI impacts. doomscrollingbabel.manoel.xyz/p/the-retrea...

Creators pour years into building a following, but in a growing underground market, you can simply buy accounts and inherit their audience. In our new pre-print, we find this practice of repurposing accounts to be prevalent and consequential on YouTube! arxiv.org/abs/2507.16045

🚨 Study alert! We've developed a prototype that enables you to create personalized social media feeds using natural language rules and import them into Bluesky. Join our study to try out the tool, give us your feedback, and earn a $50 gift card! Sign up: forms.gle/MkSGKzxDfBEc... Please share :)

In a new blogpost, I argue GenAI as is won't "eat social media," because of: - authenticity - user preferences - platforms being able to handle content at scale doomscrollingbabel.manoel.xyz/publish/post...

In a new blog post, I... - show that AI-generated videos are quite expensive to create - argue that even if they weren't, we still would not drown in it. doomscrollingbabel.manoel.xyz/p/ai-generat...

A good chunk of the AI-generated content we see right now is paid for by VC money. E.g., Leonardo.ai, a popular image/video creator, offers Veo3 **cheaper than Google**! It is like early-day Uber.

I’m stoked to have received two honorable mention awards at @icwsm.bsky.social ! With work led by Peppe, thorsten and Maciej! Sad I couldn’t attend this year but looking very much forward to catching up with this community next year!!!

Excellent talk by Hayoung in the Political Information and Opinion Dynamics panel at #ICWSM

This @icwsm.bsky.social paper fine-tunes Temporal Adapters for Llama3 8B on timelines from a panel of British Twitter users, finding that the fine-tuned LLMs retrieve collective emotions! Cool extension of simulation work to the longitudinal scenario: ojs.aaai.org/index.php/IC... cc @dgarcia.eu

In a new (very short) post, I wonder about whether the inability to forget harms large reasoning models doomscrollingbabel.manoel.xyz/p/chatgpt-th...

Laypeople often learn about science from expert explanations & those explanations often contain JARGON. Does jargon make explanations better or worse? In a paper out today in Nature Human Behaviour, @cruzf.bsky.social and I find that jargon can support illusions of understanding...

This paper characterizes (self-described) TikTok addiction with surveys and (n=107) data donations. One interesting finding is that watching positive videos predicts addiction, even controlling for session number and session time. ojs.aaai.org/index.php/IC... cc: @ingmarweber.de

This @chi.acm.org paper compares 1-hour, 1-day, and 3-day suspensions with two massive experiments on Roblox (N=700k+). TL;DR Longer suspensions work better. I hope this gets the recognition it deserves. We need more 'basic' studies on content moderation. dl.acm.org/doi/10.1145/...

This @chi.acm.org paper surveys 319 knowledge workers. The authors find that GenAI reduces cog load for some tasks (info search) but creates new tasks (verification). This made me wonder to what extent automating search will create "rich-get-richer" effects. dl.acm.org/doi/10.1145/...

Before that, I'll spend 2025-2026 as a postdoc at Princeton's CITP @princetoncitp.bsky.social. I'm excited to conduct research about informedness and mental health on social media with @manoelhortaribeiro.bsky.social and @andyguess.com.

It's finally public! 🎉 Excited to announce I'll be joining UIUC's iSchool as an Assistant Professor in Fall 2026. My lab will focus on AI information ecosystems, computational social science, and social computing. I will start recruiting PhD students this cycle, so please reach out if interested.

This @chi.acm.org paper finds that AI assistance homogenized writing in a controlled experiment with American and Indian participants (n=118). With a classifier, it was harder to predict writer origin without AI (84% vs 91%). dl.acm.org/doi/abs/10.1... @informor.bsky.social @imadityav.bsky.social

Thrilled to share that I’ve successfully defended my PhD dissertation and I will be joining MIT as an Assistant Professor starting Fall 2026, with a shared appointment between Sloan and EECS! I will be recruiting 1-2 PhD students this upcoming cycle. Consider applying to MIT EECS!

This @chi.acm.org paper finds non-intentional social media usage / algorithmically recommended content particularly regrettable. The paper's methodology is very neat: the authors triangulate surveys with screenshot analysis (with VLMs). dl.acm.org/doi/10.1145/... cc: @alexishiniker.bsky.social

I'm excited to announce that I’ll be joining the Computer Science department at Johns Hopkins as an Assistant Professor this Fall! I’ll be working on large language models, computational social science, and AI & society—and will be recruiting PhD students. Apply to work with me!

Receiving positive feedback on Reddit leads to slightly more frequent & much higher-quality contributions (for some days). Sustained positive feedback can complement content moderation. (Cool to see obs studies at @chi.acm.org!) dl.acm.org/doi/full/10.... @kous2v.bsky.social @ceshwar.bsky.social

Stay hungry, stay foolish

This paper increased my hype over Visual Language Models for CSS. Masood et al. mixed a user study, VLMs, and data donation to create a descriptive analysis of TikTok's algorithm. arxiv.org/pdf/2503.20030

This is a poor argument against LLM simulations. It suggests that LLM-based behavioral models should capture nuances that no model ever has (do agent-based models go to the toilet?). Ofc, ABM is meant to be interpretable, but social science also needs predictive power (where LLM simulations excel)

Throwback to the LIS Symposium at @rutgerscomminfo.bsky.social. Key takeaway from our panel: Researchers must get “scrappy & scrape-y” as platforms increasingly restrict data access @informor.bsky.social @haimson.bsky.social @manoelhortaribeiro.bsky.social & Shagun Jhaver

Shower thought: As developers rely more on AI to code, we might see fewer new languages/frameworks. AI excels with what already exists... Innovation outside its training data is harder, which could subtly discourage reinventing (slightly better) wheels.

Cannoli: The coolest dog in New Jersey! (And yeah, a typo in Italian)

📣 Calling all #CHI2025 attendees who work with human participants: Join our panel discussion on #LLM, #simulation, #syntheticdata, and the future of human subjects research on Apr 30 (Wed), 2:10 - 3:40 PM (JP Time) Post your questions for panelists here: forms.gle/m2mXY3xFafAX...

Ever notice how something that makes your blood boil barely registers with your friend? Our emotional reactions aren't universal at all—they're deeply personal. And AI needs to understand that. Excited to share our new paper: "iNews" 🧵 (1/8) arxiv.org/abs/2503.03335

📣Introducing our "Meet the Researcher" series where undergrads Jason Persaud '27 + Tsion Kergo '26 interview the faces of #CITP to discuss their work. First up: new Assistant Professor of Computer Science Manoel Horta Ribeiro! @manoelhortaribeiro.bsky.social blog.citp.princeton.edu/2025/03/06/m...

I think that the truth is somewhere in the middle. Societies are kind of like weather systems - we can make some near-term predictions, and those can get better and better with good models, but we will never be able to make long-term predictions. There is too much inherent randomness.

In a new blog post, I argue that we need further research on the connection between AI models' capabilities and their social and economic impact. doomscrollingbabel.manoel.xyz/p/ai-progres...

I'm super happy with this video from Communications of the ACM on our paper about LLM use in crowd work. Our work will be the cover paper of CACM in March. cacm.acm.org/research/pre... Cc: @veniamin.bsky.social @andrewgordon.bsky.social

In a new blog post, I argue we must be more critical in evaluating research on “algorithmic amplification.” Otherwise, we risk rediscovering that algorithms do not exist in a vacuum. We're better off doing other things! doomscrollingbabel.manoel.xyz/p/does-tikto...

#TechTakes - Brand new CITP Blog post just dropped 👀 "It Wasn't Me: #DeepSeek denies they stole anything, but is their model still a cost revolution?" 📢 This is part of a new series called #TechTakes, where CITP researchers comment on tech news & tech policy. freedom-to-tinker.com/2025/02/07/i...

Many companies use internal social-media-like platforms. However, we know little about their impact on companies' communication networks! In this upcoming WebSci paper, we study this empirically! arxiv.org/abs/2502.01787

Most tests for LLM biases use questionnaires, asking the model to generate a stance towards a given topic. Sadly, biases can re-emerge when the model is used in the application context. We show that apparently unbiased LLMs exhibit strong biases in conversations. Preprint: arxiv.org/abs/2501.14844

Lots of confusion around DeepSeek having "stolen" ChatGPT data... Some comments / personal takes in the thread below. 🧵

In a new blog post, I argue that "we must go beyond applying social science instruments to LLMs. We must use our collective experience to create and validate AI-specific instruments that can predict and explain machine behavior." doomscrollingbabel.manoel.xyz/p/robopsycho...

It is crazy how releasing an open-source model can make half a trillion dollars vanish into thin air

Neat short on rules in online communities: 1: studies how Reddit rules evolve, finding that diff kind of rules appear at different times. @acm-cscw.bsky.social 2: characterizes rules on Mastodon, finding emphasis on harassment/hate. @chi.acm.org 1 doi.org/10.1145/3584... 2 doi.org/10.1145/3584...

This @ACM_CHI paper conducts a RCT on the impact of a Language Model interaction on mental health-related outcomes. Very cool intervention based on existing practices, large scale-deployment, careful analysis of effect heterogeneity, interesting results. dl.acm.org/doi/10.1145/...

I'm teaching my first course! A seminar on "Machine Behavior." Readings are a mix of NLP, CSS-y, and ML work on how machines (focus LLMs) "behave" within sociotechnical systems and on how they can be used to study human behavior. Syllabus: manoelhortaribeiro.github.io/teaching/spr...

Feels fitting to have our very first-ever post be to celebrate two #CITP Associated Faculty members, Professor Aleksandra Korolova & Professor Olga Russakovsky, who were both awarded the Presidential Early Career Award for Scientists & Engineers from the White House this week! 🎉 💯 📖 #womeninSTEM

In a new blog post, I discuss the emerging literature on simulating human behavior with LLMs. doomscrollingbabel.manoel.xyz/p/simulating...

I minor comment, I fully agree with, and really like, both your post and @iyadrahwan.bsky.social & al.'s paper. But the Simon reference is a bit out of context. His "artificial" means something designed, manufactured, and fully understood/controlled by humans. Not what it means in today's "AI."

Happy holidays :-), In a new blog post, I argue for the "Machine Behavior" buzzword (created by @iyadrahwan.bsky.social et al., AFAIK) doomscrollingbabel.manoel.xyz/p/why-study-...

In my work on local subreddits, I find these subreddits to be highly accessible, which can sometimes lead newcomers to accidentally post repetitive content. A feature like this can help new users learn community rules faster, and ultimately promote interesting & original content