Michal Monselise
michalmonselise.bsky.social
Michal Monselise
@michalmonselise.bsky.social
Data Science PhD from Drexel CCI
Reposted by Michal Monselise
🚨 new working paper with Aakaash Rao (his JMP!)

Why has American politics become engulfed by a "culture war" in recent decades? We trace the culture war back to changes in the media environment in the 1980s/1990s and to the distinctive economic incentives of viewership-maximizing news outlets
November 1, 2025 at 10:35 PM
I guess duolingo works because I understand like 80% of this (I did study French in high school and college though).
i seem to have watched enough french-language instagram videos that québecois influencers are suggested to me on the site

(this one is about french-from-france folks being snooty about QC-specific words and accents)
October 19, 2025 at 1:50 AM
Reposted by Michal Monselise
CITP is now accepting applications for the 2026–27 Fellows Program. We're looking for the following:

➡️ Postdoctoral Research Associate
➡️ Visiting Research Scholar (Visiting Professional)
➡️ Microsoft Visiting Research Scholar (Visiting Professor)

Apply online: citp.princeton.edu/news/2025/no...
October 6, 2025 at 4:06 PM
Reposted by Michal Monselise
In the first seven days since I launched this new game, almost ten thousand people from 82 countries have played it. (Top cities so far: New York City, Zurich, Toronto, Chicago, Berlin, Vienna, Stockholm)

What topics would you like me to add next?

dataguessr.com
Dataguessr
Update your knowledge of the world. One quiz at a time.
dataguessr.com
September 19, 2025 at 3:17 PM
Reposted by Michal Monselise
Over 150 active academic #jobs (postdoc and TT) across Computer Science, Information Systems, Social Sciences, Management, Finance, Psychology, and related fields, spanning the US, Europe, the Middle East, Canada, and a few in New Zealand/Australia.

Most deadlines in September and October.
September 11, 2025 at 3:37 AM
Reposted by Michal Monselise
The new GESIS Panel.dbd Digital Behavioral Data Sample offers for the second time the opportunity to submit studies: www.gesis.org/en/gesis-pan...

Accepted proposals will be fielded in 2026
August 26, 2025 at 1:11 PM
Reposted by Michal Monselise
There are a bunch of surveys of evaluation and leaderboards, but the recent ones I've found are all focused on post-trained models doing specific downstream tasks (like question-answering or coding).
August 16, 2025 at 8:02 AM
Reposted by Michal Monselise
If anyone is curious about the current list of search terms being used by right wing activist organizations to harass university professors, here's a public records request I received today from Mike Howell of the Heritage Foundation. Not intimidating or meant to curtail by my speech at all, right?
August 6, 2025 at 10:03 PM
Reposted by Michal Monselise
My team has updated our free database of POSTDOC fellowships.

This database contains 286 entries. For each entry, we provide a link, short description, deadline, amount, and eligibility criteria.

Download this updated and expanded database here: research.jhu.edu/rdt/funding-...
August 10, 2025 at 7:50 PM
Reposted by Michal Monselise
🚀 We've just open-sourced Embedding Atlas – a tool for exploring large embedding spaces through rich, interactive visualizations 📊.
August 1, 2025 at 8:24 AM
Reposted by Michal Monselise
Just finished reviewing the final proofs of my forthcoming book on #techworkers. It sheds light on the high-paid professionals in the tech industry. The book is coming out open-access with @mitpress.bsky.social this December. Very excited about it!
July 31, 2025 at 11:35 AM
Reposted by Michal Monselise
The Public Voices in AI project proposes a special issue in the journal Big Data and Society, aiming to advance scholarship on the state-of-the-art and future prospects of including public voices in AI.

Find out more here: digitalgood.net/call-for-abs...
July 31, 2025 at 10:00 AM
Mamdani is making a huge mistake here: comparing democratic primary voters with general election voters. One glaring example is Crown Heights. My intuition is that one of the groups (hasidic jews) does not register as Democrat at the same rate as others in the neighborhood.
This @zohrankmamdani.bsky.social video is absolutely INCREDIBLE. The numbers he cites, the strategy and results they produced — it’s the future. When have you heard *any* campaign talk about this, ever? (Video split in 2 to fit in Bluesky’s limits)
July 2, 2025 at 4:27 AM
Reposted by Michal Monselise
Denied a loan, an interview, or an insurance claim by machine learning models? You may be entitled to a list of reasons.

In our latest w @anniewernerfelt.bsky.social @berkustun.bsky.social @friedler.net, we show how existing explanation frameworks fail and present an alternative for recourse
April 24, 2025 at 6:19 AM
Reposted by Michal Monselise
thinking again about this 2018 cover
June 18, 2025 at 5:40 PM
What draws me to a company is not "culture" but a belief in the company's mission, respect from management, and the opportunity to solve hard & interesting problems. The thing about "culture" is that we are encouraged to act like friends but when someone leaves, we pretend like they never existed.
Realization #2: by default, remote-only teams tend to have a weaker...culture, for lack of a better term.

There's much less of a shared sense of what it feels like to work at the given company, and it varies massively from team to team.
May 28, 2025 at 3:54 AM
Reposted by Michal Monselise
Want to spend funded, remote summer time making something cool with data+art/making/craft for display in @scholarslab.bsky.social's public, library-based lab space? Proposals due this Fri 5/23, but they're short+project can be small, open to all: scholarslab.lib.virginia.edu/blog/scholar...
May 19, 2025 at 9:13 AM
I'm siding with the students on this one. On a side note, I was once too positive in my comments to students and someone gave me a bad rating and said I used chatgpt to grade. I decided to reduce the cheerfulness in my tone and never got that comment again.
Professors should only use ChatGPT if using the same responsible practices I assume they'd require of students--e.g. disclosure and CHECKING LITERALLY EVERYTHING. A particularly offensive part of the business prof's behavior is that it seems he didn't read the output. www.nytimes.com/2025/05/14/t...
The Professors Are Using ChatGPT, and Some Students Aren’t Happy About It
www.nytimes.com
May 17, 2025 at 12:53 AM
Reposted by Michal Monselise
This is an awesome awesome-llm-unlearning resource, maintained by Chris Liu : github.com/chrisliu298/...
GitHub - chrisliu298/awesome-llm-unlearning: A resource repository for machine unlearning in large language models
A resource repository for machine unlearning in large language models - chrisliu298/awesome-llm-unlearning
github.com
May 13, 2025 at 2:53 PM
Reposted by Michal Monselise
PhD Timeline xkcd.com/3081
April 25, 2025 at 3:32 PM
Reposted by Michal Monselise
First draft online version of The RLHF Book is DONE. Recently I've been creating the advanced discussion chapters on everything from Constitutional AI to evaluation and character training, but I also sneak in consistent improvements to the RL specific chapter.

rlhfbook.com
April 16, 2025 at 7:01 PM
Reposted by Michal Monselise
NYC undergrads: still one more day to apply for this year's Microsoft Research Data Science Summer School! Apply here by April 15th: bit.ly/3pCQENh
April 14, 2025 at 6:48 PM
Reposted by Michal Monselise
🚨 First preprint from the lab! 🚨 Josh Rozner (w/@weissweiler.bsky.social and @kmahowald.bsky.social) uses counterfactual experiments on LMs to show that word distributions can provide a learning signal for diverse syntactic constructions, including some hard cases.
Constructions are Revealed in Word Distributions
Construction grammar posits that constructions (form-meaning pairings) are acquired through experience with language (the distributional learning hypothesis). But how much information about…
arxiv.org
March 11, 2025 at 6:06 PM