Lightnews — Scholar-powered news

Hernán Anlló @anllohernan.bsky.social · 11h

This analytic-then-generative approach reveals that effective teaching is not an art beyond measurement, but a discoverable compositional semantic structure with predictable effects!

Hernán Anlló @anllohernan.bsky.social · 11h

We demonstrate causality by synthesizing lessons that weight these dimensions optimally: synthetic lessons matched the behavioral benefits of the best human teachers, while control lessons weighted randomly produced no advantage.

Hernán Anlló @anllohernan.bsky.social · 11h

By developing a novel computational pipeline (LLM-DISC) that decomposes natural teaching language into interpretable semantic dimensions, that independently predict both expert ratings and learning outcomes.

Hernán Anlló @anllohernan.bsky.social · 11h

Teaching through language is uniquely human, yet we cannot predict which explanations will help learners succeed. This gap has profound implications: educational interventions remain trial-and-error, tutoring systems rely on heuristics rather than principles.

2

Hernán Anlló @anllohernan.bsky.social · 11h

I haven't given any news a while, I've been nose deep into this novel preprint with my excellent collaborators @stepalminteri.bsky.social, @urihertz.bsky.social and Bahador Bahrami: "Uncovering the semantics of teaching in
experiential learning with Large Language Models".
doi.org/10.31234/osf...

OSF

doi.org

4 3 8

Hernán Anlló @anllohernan.bsky.social · Mar 12

Very happy to be a part of this! More coming soon!

Stefano Palminteri @stepalminteri.bsky.social · Mar 12

Very happy to see our paper (@nicolasyax.bsky.social and @anllohernan.bsky.social) in this list. We made the risky choice to invent new tests, validate in humans, run control experiments, etc, at the risk of not being 'fast enough', but apparently it paid off 🙂

doi.org/10.1038/s44271-024-00091-8

4

Hernán Anlló @anllohernan.bsky.social · Feb 5

I have the immense pleasure of of announcing that @psl-univ.bsky.social and @cognitionens.bsky.social have granted me an HDR diploma (i.e. the highest qualification in French academia). This wouldn't have been possible without @stepalminteri.bsky.social and the INCREDIBLE team at HRL/LNC2. Thank you

5

Hernán Anlló @anllohernan.bsky.social · Feb 1

Oh heck yes! Wonder if I'll finally be able to drop Matlab for the remaining capricious RL models I have and go full R once and for all...

Ben Bolker @bbolker.bsky.social · Feb 1

🧵 (1/?) Posting technical Qs to distract me from :

thinking about optimizers/MCMC for ugly surfaces (multimodal/rough/weakly identifiable), e.g. differential evolution github.com/ArdiaD/DEoptim, doi.org//10.18637/js... or dx.doi.org/10.1086/670067 affine-invariant ensemble sampler

GitHub - ArdiaD/DEoptim: Differential evolution in R

Differential evolution in R. Contribute to ArdiaD/DEoptim development by creating an account on GitHub.

github.com

1

Hernán Anlló @anllohernan.bsky.social · Feb 1

Amazing, dare I say mandatory, new resource

Vincent Arel-Bundock @vincentab.bsky.social · Feb 1

I just posted a new notebook:

Causal Inference with the 𝚖𝚊𝚛𝚐𝚒𝚗𝚊𝚕𝚎𝚏𝚏𝚎𝚌𝚝𝚜 📦 for #Rstats. It introduces the idea of G-Computation and tries to demystify the differences between ATE, ATT, and ATU.

This is a first draft and I would looove any feedback!

marginaleffects.com/chapters/gco...

8 Causal inference with G-computation – Model to Meaning

marginaleffects.com

3

Hernán Anlló @anllohernan.bsky.social · Feb 1

Like if they lead nation-wide committees that could end your career forever if they blacklisted you, so they get away with discrimination? Asking for a friend...

Marieke van Vugt @mvugt.bsky.social · Feb 1

"It’s not easy to point out errors, especially when they’re attached to the name of someone who holds great power over you." www.science.org/content/arti...

Investigating scientific misconduct is hard—especially when your supervisor is an author

“The paper was riddled with serious problems,” this Ph.D. student writes

www.science.org

1

Hernán Anlló @anllohernan.bsky.social · Jan 29

Most intriguing!!

Maxime Derex @maximederex.bsky.social · Jan 29

New paper in #ProcB with @alexmesoudi.com, @jfbonnefon.bsky.social, @rmcelreath.bsky.social and Rob Boyd.

We investigate how social learning shapes the way we explore and show that it can even preserve useless theories! 🧵👇

1

Hernán Anlló @anllohernan.bsky.social · Jan 28

I'm sure you saw that all data is available for our 11-country study www.nature.com/articles/s41...

Comparing experience- and description-based economic preferences across 11 countries - Nature Human Behaviour

Anlló et al. show that context effects on reward value encoding are found across 11 diverse countries, suggesting that this may be a basic feature of human decision-making.

www.nature.com

1

Hernán Anlló @anllohernan.bsky.social · Jan 26

Hmmmm. I get what's going on here, but indeed the only way to conclusively debunk this IMO is to replicate in rural areas and non-urban ethnic groups. As I see it now (pun intended) the effect is clearly low level in origin, but culture could defo determine its instantiation.

Hernán Anlló @anllohernan.bsky.social · Jan 26

Hmmmm. I get what's going on here, but indeed the only way to conclusively debunk this IMO is to replicate in rural areas and non-urban ethnic groups. As I see it now (pun intended) the effect is clearly low level in origin, but culture could defo determine its instantiation.

2

Hernán Anlló @anllohernan.bsky.social · Jan 26

End of an era? I also cited this study :o

Dorsa Amir @dorsaamir.bsky.social · Jan 25

Does the culture you grow up in shape the way you see the world? In a new Psych Review paper, @chazfirestone.bsky.social & I tackle this centuries-old question using the Müller-Lyer illusion as a case study. Come think through one of history's mysteries with us🧵(1/13):

1 2

Hernán Anlló @anllohernan.bsky.social · Jan 25

We're getting there people!!

Laurent Bègue Shankland @lbegueshankland.bsky.social · Jan 25

Yesterday in Nature. A new place for science 👌

4

Hernán Anlló @anllohernan.bsky.social · Jan 24

Fascinating work!

David Lindner @davidlindner.bsky.social · Jan 23

New Google DeepMind safety paper! LLM agents are coming – how do we stop them finding complex plans to hack the reward?

Our method, MONA, prevents many such hacks, *even if* humans are unable to detect them!

Inspired by myopic optimization but better performance – details in🧵

Reposted by Hernán Anlló

Stefano Palminteri @stepalminteri.bsky.social · Jan 23

Epistemic biases in human reinforcement learning: behavioral evidence, computational characterization, normative status and possible applications.

A quite self-centered review, but with a broad introduction and conclusions and very cool figures.

Few main takes will follow

osf.io/preprints/ps...

1 19 38

Reposted by Hernán Anlló

Melanie Mitchell @melaniemitchell.bsky.social · Jan 23

Worth also pointing out that there are many "tests so easy no AI system can pass them".

Moravec's paradox remains.

E.g., arxiv.org/abs/2404.12390

7 35 120

Hernán Anlló @anllohernan.bsky.social · Jan 12

Do include me please!

1

Hernán Anlló @anllohernan.bsky.social · Jan 11

Lovely, thanks for sharing. Open for a chat indeed!

Hernán Anlló @anllohernan.bsky.social · Jan 9

What a cautionary tale waiting to happen this is!!

Neuroskeptic @neuroskeptic.bsky.social · Nov 22

You know the "🔹AI Overview" you get on Google Search?

I discovered today that it's repeating as fact something I made up 7 years ago as a joke.

"Kyloren syndrome" is a fictional disease I invented as part of a sting operation to prove that you can publish any nonsense in predatory journals...

3 6

Hernán Anlló @anllohernan.bsky.social · Dec 30

This!!!!!! A plague in EEG and fMRI

PessoaBrain @pessoabrain.bsky.social · Dec 28

In task based analysis we use multiple repetitions of a trial/block, in fact a lot across "runs". One cannot average those *before* statistical analysis because it completely misestimates variability, mishandles degrees of freedom, does not account for autocorrelation, etc.

2

Hernán Anlló @anllohernan.bsky.social · Dec 29

It will do, it will do *faints into oblivion*

Hernán Anlló @anllohernan.bsky.social · Dec 29

One of my personal stats heroes E J Wagenmakers has just liked one of my goofy comments. I can't begin to explain my excitement. Now I'm just missing @bbolker.bsky.social, and I can retire in peace.

1 4