(please reshare) I'm recruiting multiple PhD students and Postdocs @uwcse @uwnlp ( https://t.co/I5wQsFnCLL). Focus areas incl. psychosocial AI simulation and safety, Human-AI collaboration. PhD: https://t.co/ku40wCrpYh Postdocs: https://t.co/K9HUIPJ5h6

Can AI simulate human behavior? 🧠 The promise is revolutionary for science & policy. But there’s a huge "IF": Do these simulations actually reflect reality? To find out, we introduce SimBench: The first large-scale benchmark for group-level social simulation. (1/9)

📄✨Excited to share our new paper accepted to #EMNLP ’25: Combining Constrained and Unconstrained Decoding via Boosting: BoostCD and Its Application to Information Extraction https://t.co/ljsWULBHEA (led by #EPFL PhD student Marija Šakota -- soon on the job market, hire her!!)

can we finally use natural language to optimize for deeper notions of what users want from their recommender systems?

I argue that if we consider these three points, we find that labeling with LLMs is neither trick nor treat. Treated as measurement instruments, their value lies in forcing us to confront uncertainty we once ignored, not in completely eliminating it.

Where do we go from here? I argue we should do three things: 1) clarify what problem we’re solving; 2) clarify the threat model of malicious actors who may engage in bad science; 3) we need to embrace the fact that human labels can also suck.

I cluster work in this area broadly into three waves: the “wow” phase (e.g., Gillardi’s PNAS paper), the “how do we do this right?” phase (e.g., Egami’s DSL), and the “the boat is on fire” wave (e.g., Baumann’s LM hacking).

Large language models are quietly transforming how social scientists label data. In dozens of new studies, undergrad coders and Turkers have been replaced by GPT-5 or Gemini 2.5 (or whatever new model just arrived). What began as a convenience is becoming a methodological shift.

The debate over “LLMs as annotators” feels familiar: excitement, backlash, and anxiety about bad science. My take in a new blogpost is that LLMs don’t break measurement; they expose how fragile it already was. https://t.co/6CweDPv5wG

We're over-indexing on the negatives of AI companions. While the risks are real, the panic-driven narrative ignores the massive, stigmatized crisis of loneliness they're trying to solve. I wrote a short post on the positive use cases we shouldn't ignore https://t.co/ZvkxBkJswy

Ready for #ICWSM2026 in LA next May? 🌴 📅 Deadlines: • Jan 15, 2026 – Papers, posters, demos, datasets, tutorials (note: posters/demos/datasets submitted in Sep '25 are reviewed only after this deadline) • Jan 30, 2026 – Workshop proposals 🔗 Details: https://t.co/KTnOPzuutb

I am super excited to share a new AI tool, Refine. Refine thoroughly studies research papers like a referee and finds issues with correctness, clarity, and consistency. In my own papers, it regularly catches problems that my coauthors and I missed. 1/

Look, I'm afraid I have to tap the sign again. Last time someone looked at this, TikTok was found to "boost" Republican content. This time, the opposite. But what does it mean for an algorithm to "boost" content? What about what the users are doing? https://t.co/HbKqZjwVvU

GreenEarth is creating open source AI-driven recommender infrastructure for the ATProto ecosystem. Type a prompt, see your feed change. We are here for the users, the builders, the dreamers. Join us. https://t.co/5wme8hcdJv

Feels like a dream! I’ve recently started my Ph.D. in Computer Science @Princeton! Working on exciting research with Professors @HazanPrinceton and @tri_dao 🤩

Adji Bousso Dieng, assistant professor of computer science and leader of the @Vertaix_ lab, has won the Prix Galien Africa Special Prize from @galienafrique. She will receive the award in Dakar, Senegal on October 31. https://t.co/FMgpee4IZM

I'd also expect that agreement between humans is not better than 0.6 for most of these papers... The authors note that a benchmark suggests an agreement of 0.73, but there is considerable selection bias in who would publish such agreement metrics in the first place.

On that note, I am not sure this follows from the actual content of the paper. Both the recent LLM p-hacking paper and this newest paper could simply reflect that the data annotation for the original papers is bad? https://t.co/iS9Ea55cFp

It is worth noting that many papers replicated in Yang et al. used the previous generation of automated methods. Most were not purely human annotations. This paper would be so cool if the authors actually reannotated some of the data by themselves...

Congrats to @galenweld for winning yet another paper award for his great work on measuring perceptions of community moderation at Reddit scale! #cscw2025