Lightnews — Scholar-powered news

Matt Groh @mattgroh.bsky.social · 4d

Happening today!

NICO @nicoatnu.bsky.social · 4d

People handle the everyday physical world with remarkable ease, but how do we do it? This Wednesday, NICO is thrilled to host @tomerullman.bsky.social, to discuss: "Good Enough: Approximations in Mental Simulation and Intuitive Physics".

🗓️ Wed 10/8 at 12pm US Central
🔗 bit.ly/WedatNICO

An event flyer for the Wednesdays@NICO Seminar Series, on October 8, 2025, featuring Tomer Ullman of Harvard University.

1

Matt Groh @mattgroh.bsky.social · Jul 21

On my way to @ic2s2.bsky.social in Norrköping!! Super excited to share this year’s projects in the HAIC lab revealing how (M)LLMs can offer insights into human behavior & cognition

More at human-ai-collaboration-lab.kellogg.northwestern.edu/ic2s2

See you there!

#IC2S2

4

Matt Groh @mattgroh.bsky.social · Jun 18

Thanks! I imagine we'd see similar results in the Novelty Challenge that when experts are reliable we can fine-tune LLMs to be reliable but experts may only be reliable in some disciplines/settings and less reliable in others.

Very cool challenge!!

Matt Groh @mattgroh.bsky.social · Jun 17

When are LLMs-as-judge reliable?

That's a big question for frontier labs and it's a big question for computational social science.

Excited to share our findings (led by @aakriti1kumar.bsky.social!) on how to address this question for any subjective task & specifically for empathic communications

aakriti1kumar.bsky.social @aakriti1kumar.bsky.social · Jun 17

How do we reliably judge if AI companions are performing well on subjective, context-dependent, and deeply human tasks? 🤖

Excited to share the first paper from my postdoc (!!) investigating when LLMs are reliable judges - with empathic communication as a case study 🧐

🧵👇

1 1 7

Matt Groh @mattgroh.bsky.social · May 29

Thank you for sharing your brilliance, quirks, and wisdom. I started reading your work after coming across your Aeon article on Awe many years ago, and I feel inspired everytime I read what you write.

5

Matt Groh @mattgroh.bsky.social · May 1

Awesome write up in Kellogg Insight on our paper published at #CHI2025 this week!

insight.kellogg.northwestern.edu/article/are-...

When Put to the Test, Are We Any Good at Spotting AI Fakes?

For the most part, yes! And the more we look, the better we get.

insight.kellogg.northwestern.edu

4 8

Matt Groh @mattgroh.bsky.social · Apr 25

And follow negrain.bsky.social who just joined Blue Sky today!

negrain.bsky.social

1 1

Matt Groh @mattgroh.bsky.social · Apr 25

If you're curious about learning more, say hi to
Negar Kamali at #CHI2025 and see video

Awesome collaboration with
Karyn, @aakriti1kumar.bsky.social, Angelos, @jessicahullman.bsky.social

Video: youtu.be/PL_ggNzMd-o?...
Preprint: arxiv.org/pdf/2502.11989
CHI: dl.acm.org/doi/10.1145/...

CHI 2025 - Characterizing Photorealism and Artifacts in Diffusion Model-Generated Images

YouTube video by Negar Kamali

youtu.be

1 1 3

Matt Groh @mattgroh.bsky.social · Apr 25

This taxonomy offers a shared language (and see our how to guide on arXiv for many examples) to help people better communicate what looks or feels off.

It's also a framework that can generalize to multimedia.

Consider this, what do you notice at the 16s mark about her legs?

1 2

Matt Groh @mattgroh.bsky.social · Apr 25

Based on generating thousands of images, reading the AI-generated images and digital forensics literatures (and social media and journalistic commentary), analyzing 30k+ participant comments, we propose a taxonomy for characterizing diffusion model artifacts in images

1 1

Matt Groh @mattgroh.bsky.social · Apr 25

Scene complexity, artifact types, display time, and human curation of AI-generated images all play significant roles in how accurately people distinguish real and AI-generated images.

1 1

Matt Groh @mattgroh.bsky.social · Apr 25

We examine photorealism in generative AI by measuring people's accuracy at distinguishing 450 AI-generated and 150 real images

Photorealism varies from image to image and person to person

83% of AI-generated images are identified as AI better than random chance would predict

1 1

Matt Groh @mattgroh.bsky.social · Apr 25

💡New paper at #CHI2025 💡

Large scale experiment with 750k obs addressing

(1) How photorealistic are today's AI-generated images?

(2) What features of images influence people's ability to distinguish real/fake?

(3) How should we categorize artifacts?

2 4 16

Matt Groh @mattgroh.bsky.social · Apr 3

Agreed with your observation that disciplinary perspectives can be too narrow minded on this problem and forget the big picture of both sides

www.nature.com/articles/s41... does a really nice job systematically reviewing the Human-AI collaboration literature across a bunch of different domains

When combinations of humans and AI are useful: A systematic review and meta-analysis - Nature Human Behaviour

Vaccaro et al. present a systematic review and meta-analysis of the performance of human–AI combinations, finding that on average, human–AI combinations performed significantly worse than the best of ...

www.nature.com

3 9

Matt Groh @mattgroh.bsky.social · Apr 3

At a high level, it depends on:

- human expertise
- human understanding for what the AI system is capable of
- quality of AI explanations
- task-specific potential for cognitive biases and satisficing constraints to influence humans
- instance-specific potential for OOD data to influence AI

1 4

Matt Groh @mattgroh.bsky.social · Apr 3

NICO Intake form: kellogg.qualtrics.com/jfe/form/SV_...

Qualtrics Survey | Qualtrics Experience Management

Survey Software, Enterprise Survey software for enterprise feedback management and CRM solutions. Enables high-quality data collection, panel management and results analysis. Perfect for market resear...

kellogg.qualtrics.com

Matt Groh @mattgroh.bsky.social · Apr 2

📣 📣 Postdoc Opportunity at Northwestern

Dashun Wang and I are seeking a creative, technical, interdisciplinary researcher for a joint postdoc fellowship between our labs.

If you're passionate about Human-AI Collaboration and Science of Science, this may be for you! 🚀

Please share widely!

1 4 5

Matt Groh @mattgroh.bsky.social · Apr 2

You're welcome!! Def makes makers who move between both worlds feel very seen

1

Matt Groh @mattgroh.bsky.social · Mar 29

Impressive on the 20 minute bits approach!

I definitely need 4 hour windows for productive, creative work.

Paul Graham's essay on the Maker/Manager schedule (paulgraham.com/makersschedu...) offers some tips for how to create schedules that address roles where one is both a Maker and Manager

Maker's Schedule, Manager's Schedule

paulgraham.com

1 2

Matt Groh @mattgroh.bsky.social · Mar 20

V2 of the Human and Machine Intelligence 😊🤖🧠 is in the books!

So many fantastic discussions as we witnessed the frontier of AI shift even further into hyperdrive✨

Props to students for all the hard work and big thanks to teaching assistants and guest speakers 🙏

3

Matt Groh @mattgroh.bsky.social · Mar 7

and present evidence that perception is more than simply transforming light into representations of objects and their features, perception also automatically extracts relations between objects!

3

Matt Groh @mattgroh.bsky.social · Mar 7

What is perception? What do we really see when we look at the world?

And, why does the amodal completion illusion lead us to see a super long reindeer in the image on the right?

This week @chazfirestone.bsky.social joined the NU CogSci seminar series to address these fundamental questions

1 1 13

Matt Groh @mattgroh.bsky.social · Jan 8

well said!

6

Matt Groh @mattgroh.bsky.social · Jan 1

PDF with links here: mattgroh.com/pdfs/annual_...

mattgroh.com

7

Matt Groh @mattgroh.bsky.social · Dec 31

2024 marks the official launch of the Human-AI Collaboration Lab, so I wrote a one page letter to introduce the lab, share highlights, and begin a lab tradition of reflecting on the year and sharing what we're working on in an easy to digest annual letter to share with friends and colleagues.

1 2 13