Jason Hartline

AI's gains in efficiencies will result in many kinds of jobs being eliminated. It has always been the responsibility of governments to control the economy to ensure that gains in efficiency benefit all people, rather than the ultra rich. AI gains without economic adjustments are the looming danger.

📢 Postdoc position 📢

I’m recruiting a postdoc for my lab at NYU! Topics include LM reasoning, creativity, limitations of scaling, AI for science, & more! Apply by Feb 1.

(Different from NYU Faculty Fellows, which are also great but less connected to my lab.)

Link in 🧵

Great culture can save lives. Literally.

Amazing letter in today’s @thetimes.com about Tom Stoppard

Depressing late Friday holiday news dump: Northwestern agrees to pay $75 million to the federal government.

The faculty vote against doing this, @heidikitrosser.bsky.social points out, was 595-4.

That's the kind of margin where going the other way reflects a major failure of university governance.

For example, many years ago SODA had a two page paper option. This option was terminated, as I understand it, because page limits are an upper bound, and a significant work that could be written concisely didn't need a separate track. Thus, sufficient length as a criteria excludes real papers.

For many conferences, not counting short papers is an ok policy as these are not considered published and, e.g., can be submitted as full papers to same conference. I believe whether or not the paper can be subsequently submitted to another conference is the right criterion for counting, not length.

Including Extended Abstracts from ACM EC · Issue #9506 · emeryberger/CSrankings Currently, only full papers at EC (the ACM Conference on Economics and Computation) are indexed by CSRankings. This issue is to request that EC extended abstracts be counted and indexed as publicat...

Csrankings.org does not count papers that are under six pages, so most ACM EC papers are not counted (most authors choose the extended abstract format). But @acmsigecom.bsky.social's policy is that full and extended abstracts are the same!

See discussion: github.com/emeryberger/...

The 2025 Chicago Junior Theorists Workshop is December 8-9, hosted jointly by Northwestern and TTIC. Monday Dec 8 will be at TTIC and Tuesday Dec 9 will be at Northwestern (in the SkAI/NITMB in the Hancock tower in downtown Chicago). Register in advance.

theory.cs.northwestern.edu/2025/11/14/j...

Every year I track the number of CS faculty jobs ads in the November issues of CRA News. It's not a perfect measure but it is forward looking to the upcoming job season.

This year at 27 pages is down 69% from last year, 83% from the peak in 2022 and the lowest since 2011.

🚨 EC'26 website is now live! 🚨

Deadline for abstracts 2 Feb & for papers 9 Feb ⏱️ 1st round decisions 26 March 😥 Rebuttal period 21-25 April 🙅 Decisions by 18 May 🎯 Conference 6-10 July in Rome 🍕🇮🇹🍝.

Instead of tracks, there are 13 amazing track chairs!
ec26.sigecom.org/index.html

The 2026 SIGecom job market candidate profiles are now available online!

Share the following article with anyone who may be interested in hiring at the intersection of computer science, operations research, and economics:
www.sigecom.org/exchanges/vo...

Best resource every year, come look who's on the market!

Thanks for editing, @gkatzelis.bsky.social and @jasonhartline.bsky.social!

📣 Postdocs at Yale FDS! 📣 Tremendous freedom to work on data science problems with faculty across campus, multi-year, great salary. Deadline 12/15. Spread the word! Application: academicjobsonline.org/ajo/jobs/31114 More about Yale FDS: fds.yale.edu

A picture of the audience at BEACH Day

Yiling Chen giving a talk on Scoring Rules

Yifan Wu giving a bootcamp on Calibration

BEACH Day poster

A truly incredible BEACH Day today on Predictions for Decision Making, full of amazing talks.

Hope everyone enjoyed the view from @bucds.bsky.social!

Photos of Yiling Cheng and @yifanwu.bsky.social

@bostoncollege.bsky.social is launching a PhD program in Computer Science! Help spreading the word to interested undergrads would be appreciated; our department has faculty research clustered in ML and CS theory, and Boston's a nice place to be :) www.bc.edu/content/bc-w...

The monthly CSLaw Workshop is today! Virtual!

I am recruiting PhD students at NYU Courant to conduct research in learning theory, algorithmic statistics, and trustworthy machine learning, starting Fall 2026. Please share widely! Deadline to apply is December 12, 2025.

Reminder, the CSLaw Workshop meets Friday 1pm-3pm Central [Nov 14, 2025]. This month we have

- Grimmelmann (@jtlg.bsky.social), Sobel, and Stein on "Generative Misinterpretation" and
- Aileen Nielsen on "Law is vulnerable to AI influence; interface design can help"

Details: www.cslawworkshop.org

Loss Minimization Yields Multicalibration for Large Neural Networks Multicalibration is a notion of fairness for predictors that requires them to provide calibrated predictions across a large set of protected groups. Multicalibration is known to be a distinct goal tha...

This is a really lovely result. Again, it's in this paper:

"Loss Minimization Yields Multicalibration for Large Neural Networks"
arxiv.org/abs/2304.09424

It seems the paper is not very well known and, hence, worthwhile to post about.

3. so most sizes n do not improve accuracy much by increasing to n+k+2 so they must have low bias (by 1).

4. minimizing loss with regularizer that penalizes for large sizes n will avoid the ones where there is significant improvement by making the network a little bigger (i.e., to n+k+2).

2. consider the sequence of optimal losses as a function of network size. there can't be too many sizes n where the networks of size n+k+2 are significantly better. this is because the cumulative difference in losses of any sequence of networks is at most the maximum loss of 1.

Proof outline:
1. neural networks are rich enough so that if there is expected multi-calibration bias for a size n network for points identified by a size k network the two networks can be combined with size n+k+2 to reduce squared loss by the squared bias.

Back to "Loss Minimization Yields Multicalibration for Large Neural Networks". Main Theorem:

The minimizer of "squared loss minus a regularizer term that penalizes for neural network models by their size" is automatically multi-calibrated with respect to small neural networks.

The calibration module was heavily based on the paper reading course co-taught with @yifanwu.bsky.social last year. Details of both courses are here:

- paper reading on calibration: sites.northwestern.edu/hartline/cs-...
- lecture based on data economics: sites.northwestern.edu/hartline/cs-...

Loss Minimization Yields Multicalibration for Large Neural Networks Multicalibration is a notion of fairness for predictors that requires them to provide calibrated predictions across a large set of protected groups. Multicalibration is known to be a distinct goal tha...

Just wrapped up teaching a module on Calibration in my Foundations of Data Economics class (PhD level), that I am cocreating with Bo Waggoner and Nicole Immorlica.

The final lecture which was on "Loss minimization yields multicalibration for large neural networks." arxiv.org/abs/2304.09424

FOCS 2025: Job openings for postdocs and researchers Please provide details about any theory job openings in your group (e.g., postdoc openings, research-related jobs or academic positions). This information will be made available as a Google Spreadshee...

If your institution wants to advertise Theory jobs to the FOCS 2025 attendees, a form has been set up on our website: focs.computer.org/2025/activit...

The spreadsheet collating job openings in TCS will be made available to attendees (and online), to help graduating students and junior researchers.

Econ, artificial intelligence and machine learning …
#EconConf #econsky
marketdesigner.blogspot.com/2025/11/econ...

If you work at the intersection of CS and economics (or think your work is of interest to those who do!) consider submitting to the ESIF Economics and AI+ML meeting this summer at Cornell: www.econometricsociety.org/regional-act...

Computational Complexity Conference launches its first test of time award. Nominations due by March 2.

computationalcomplex...

On the other hand, some esteemed scholars are finding the AI slop to be pretty good. bsky.app/profile/lanc...