iiinova — where your digital self keeps evolving

Reposted by Rohan Alexander

Emily Riederer @emilyriederer.bsky.social · 02/10/2026

Blows my mind that in 2026 Parquet is not yet the absolute default for data sharing A few years back there was a lovely talk at @rohanalexander.bsky.social 's Toronto Data Workshop by some students who articulated one of my soapboxes far better than I ever have: www.youtube.com/watch?v=Evcn...

youtube.com

Exploring Alternatives to REST for Accessing Public Data Sets

YouTube video by Rohan Alexander

3273

Reposted by Rohan Alexander

Laura K. Nelson @lauraknelson.bsky.social · 14/06/2026

I'm really vibing with this collection of essays on the Future of Quantitative Social Science. It flat out has me buzzing. As Kevin Munger says in Chapter 14: "But AI will soon make it impossible for us to ignore the absurdity of the status quo..." rohanalexander.github.io/notes_on_the...

rohanalexander.github.io

Notes on the Future of Quantitative Social Science

1399

Rohan Alexander @rohanalexander.bsky.social · 26/04/2026

Wow sorry to hear about that. Glad you're out.

020

Rohan Alexander @rohanalexander.bsky.social · 22/04/2026

Everyone else in this hackathon is like 20 and me, I'm like, "hey let's please have our theme inspired by 'I want my hat back' by Jon Klassen."

040

Rohan Alexander @rohanalexander.bsky.social · 22/04/2026

All welcome at the workshop in 10 days. rohanalexander.com/tdw.html Also, if you read Ryan's piece and thought "hey, I have thoughts on AI and quant social science as well", please do get in touch.

071

Rohan Alexander @rohanalexander.bsky.social · 21/04/2026

Alex you're in the bad books - didn't even respond to my email asking to contribute to this ;)

120

Reposted by Rohan Alexander

Ryan Briggs @ryancbriggs.net · 21/04/2026

I'll be at a workshop organized by @rohanalexander.bsky.social next week on how AI will change quantitative social science. This is my short paper, and this is the key argument. ryancbriggs.net/blog/as-ai-l...

Information manipulation is going to get cheaper much faster than interacting with the world will. I expect that this will lead to the bottleneck in research shifting from production to consumption. We should be preparing for this shift now by building infrastructure that allows us to adjudicate claims and filter work in a world where production is cheap. The current journal system is not well positioned to do this, and so we should be experimenting with new institutions that can.

5449

Rohan Alexander @rohanalexander.bsky.social · 21/04/2026

That's wonderful news. Congratulations!

110

Rohan Alexander @rohanalexander.bsky.social · 14/04/2026

Congratulations to you and @rocher.lc on the pub - really interesting; I had no idea.

120

Rohan Alexander @rohanalexander.bsky.social · 13/04/2026

You might enjoy this book! en.wikipedia.org/wiki/A_Milli... (And apologies if you linked to it and I missed the link)

en.wikipedia.org

A Million Random Digits with 100,000 Normal Deviates - Wikipedia

140

Rohan Alexander @rohanalexander.bsky.social · 13/04/2026

Please join us in Toronto on the 29th!

052

Rohan Alexander @rohanalexander.bsky.social · 10/04/2026

Thank you very much for taking the time to reply.

010

Rohan Alexander @rohanalexander.bsky.social · 10/04/2026

I just ask because my reading of the Mythos system card e.g. 42-43 is they mostly found gains from human research.

040

Rohan Alexander @rohanalexander.bsky.social · 10/04/2026

I'm interested in the "recursive self-improvement" comment. Is there some documentation behind that one please?

290

Reposted by Rohan Alexander

Crystal Lewis @cghlewis.bsky.social · 09/04/2026

9 months in and RDM Weekly has hit 700 subscribers! 🎉 I would love to hit 1,000 by July for the 1 year anniversary. rdmweekly.substack.com

2296

Rohan Alexander @rohanalexander.bsky.social · 09/04/2026

OMG someone needs to bring back the Dragon Naturally Speaking aesthetic. Ahead of its time!

010

Rohan Alexander @rohanalexander.bsky.social · 08/04/2026

The Mythos system card is worth reading. Nuanced and interesting throughout. www-cdn.anthropic.com/53566bf5440a...

www-cdn.anthropic.com

010

Rohan Alexander @rohanalexander.bsky.social · 06/04/2026

Thanks for sharing your writeup. Super interesting.

000

Rohan Alexander @rohanalexander.bsky.social · 06/04/2026

That's so crazy. Can't believe I not only get modern Berry back in my life, but also seemingly his relatives. Thanks for sharing @emilyriederer.bsky.social

120

Reposted by Rohan Alexander

Tiffany Timbers @tiffanytimbers.bsky.social · 31/03/2026

Our survey closes in just 2 weeks!!! 🏁 Calling folks working in a #DataScience (or related) role 📣 – we want you to fill out our survey aiming to understand if and how practicing data scientists use and integrate #LLMs 🤖 into their data science tasks. www.linkedin.com/posts/tiffan...

022

Rohan Alexander @rohanalexander.bsky.social · 31/03/2026

This was really interesting. Thanks for writing!

010

Rohan Alexander @rohanalexander.bsky.social · 30/03/2026

This Wednesday 1 April 2026, noon (EDT) on zoom - utoronto.zoom.us/j/4784708970 - we host Yiqing Xu, Stanford University on "Scaling Reproducibility: An AI-Assisted Workflow for Large-Scale Reanalysis". All welcome!

120

Rohan Alexander @rohanalexander.bsky.social · 27/03/2026

Nice. Appendix D is wonderful! We could build out a full course with each week just going through one of the classes and how to avoid it.!

100

Rohan Alexander @rohanalexander.bsky.social · 23/03/2026

This Wednesday 25 March 2026, 1:30pm (EDT) on Zoom - utoronto.zoom.us/j/4784708970 - we host James Tebrake, International Monetary Fund on "StatGPT: AI for Official Statistics". All welcome.

000

Rohan Alexander @rohanalexander.bsky.social · 22/03/2026

Thank you very much for posting these.

040

Rohan Alexander @rohanalexander.bsky.social · 18/03/2026

Many analysts evaluate many AI agents doing data analysis - arxiv.org/abs/2602.14349 - arxiv.org/pdf/2602.16666 - papers.ssrn.com/sol3/papers.... - arxiv.org/abs/2602.18710 - arxiv.org/pdf/2602.20440

091

Rohan Alexander @rohanalexander.bsky.social · 16/03/2026

This Wed, noon (EDT), on Zoom - utoronto.zoom.us/j/4784708970 - we host Zora (Zhiruo) Wang, Carnegie Mellon University on "AI Agents at Work - But Whose Work?" All welcome! Details: rohanalexander.com/tdw.html

000

Reposted by Rohan Alexander

Inessa De Angelis @inessadeangelis.bsky.social · 16/03/2026

New article w/ Lindsay Katz & @rohanalexander.bsky.social in @ausjpolsci.bsky.social! Analyzing Australian House of Reps debates, we find that as women’s representation increases, interruptions of women MPs & interruptions overall decline Read more: doi.org/10.1080/1036... #polsky #gendersky

Screenshot of a journal article published at the Australian Journal of Political Science titled "Do women politicians face more interruptions? An analysis of interjections in the Australian parliamentary debates (1998-
2025)." Authored by Lindsay Katz, Inessa De Angelis, and Rohan Alexander.

086

Rohan Alexander @rohanalexander.bsky.social · 13/03/2026

Remarkable! Congratulations on having the comment published.

020

Rohan Alexander @rohanalexander.bsky.social · 10/03/2026

This Wednesday 11 March 2026, noon (EDT) on Zoom - utoronto.zoom.us/j/4784708970 - we host Morris Greenberg, University of Toronto on “The Past, Present, and Future of Scrabble Engines”. Should be a fun one, all welcome! Upcoming/details: rohanalexander.com/tdw.html

010

Rohan Alexander @rohanalexander.bsky.social · 06/03/2026

Right. I've re-run this to get both 5.2 and 5.4 again, and it's basically the same difference. Maybe I've made some persistent mistake; will be interesting to see if others also find differences, in different settings, between these models.

020

Reposted by Rohan Alexander

Emma Zang @dremmazang.bsky.social · 06/03/2026

Call for Submissions: AI for Social Science Methodology (Yale) • Keynote: @nachristakis.bsky.social • Panel with editors of leading journals on publishing AI research • Mentoring roundtables for early-career scholars • Generous travel support Discussion-driven, high-quality research.

yalefds.swoogo.com

Home

11411

Rohan Alexander @rohanalexander.bsky.social · 06/03/2026

Yeah it's really interesting how different it is (also nice that it's easy to systematically test over time).

000

Rohan Alexander @rohanalexander.bsky.social · 06/03/2026

OpenAI API. So it's possible I've done something wrong, but it's just swapping a parameter, so I'd hope not. IDK. As you can imagine, trying to get to the bottom of it today.

120

Rohan Alexander @rohanalexander.bsky.social · 06/03/2026

I didn't believe it and ran it again last night and got the same. I'm doing another pass this morning. Some of the more critical papers mention the models thing, but I was really surprised at the extent of the difference especially because it's the same provider, and very closely related models.

031

Rohan Alexander @rohanalexander.bsky.social · 06/03/2026

I broadly agree, but I wanted to take it seriously and to try it myself in a setting I knew. And I'm glad I did it because I learnt about this huge model effect! I'd like to think I'm now a better-informed sceptic of the approach. But reasonably people can disagree ;)

130

Rohan Alexander @rohanalexander.bsky.social · 06/03/2026

I was really surprised at how different the silicon sample results were between GPT-5.2 and 5.4. Light red is 5.2 and dark red is 5.4 and it's an Australian-setting so we're seeing big differences for right-wing parties.

Screenshot showing silicon sample responses to a question about increased gun laws in Australia. There are three question variants, and four parties - the main result is that 5.2 tends to have higher levels of support for restrictions than 5.4, especially for right wing parties.

5263

Reposted by Rohan Alexander

Rohan Alexander @rohanalexander.bsky.social · 17/02/2026

Two recent pre-prints: 1) "Same Prompt, Different Outcomes: Evaluating the Reproducibility of Data Analysis by LLMs" with Jiaxin (Allyson) Cui, arxiv.org/abs/2602.14349 2) "Benchmarking AI Performance on End-to-End Data Science Projects" with Evelyn Hughes, arxiv.org/abs/2602.14284 Comments welcome.

062

Rohan Alexander @rohanalexander.bsky.social · 25/02/2026

Congratulations. Wonderful to see this released!

020