Don't miss out on new replications with the Zotero Replication Checker!

This is the current set of >1.6K studies in the FORRT Library of Replication Attempts (#FLoRA). Many of them belong to large-scale projects and are not even cited in the final report, so how to keep track of them?

Emotion concepts and their function in a large language model Interpretability research from Anthropic on emotion concepts

OK folks this reads to me like a stacked pile of nothing put in a box labeled 'wow, so interesting!' and if there is something anyone thinks I am missing (I'm talking to you, bluesky-doesn't-get-AI guys) I would like to know what it is. Here's how I read it: 1/n

www.anthropic.com/research/emo...

New preprint out today (osf.io/preprints/ps...). We tested whether AI agents are actually infiltrating online surveys.

Spoiler alert: they aren't

Thread 🧵

[1/9]

Reliable research in the social and behavioural and sciences Sweeping new investigations probe the replication, robustness and reproducibility of results across the behavioural and social sciences.

The SCORE investigation of repeatability and credibility is a lot. There are a few ways to get your head around it.

1: The Nature collection includes 3 papers from SCORE, an amazing paper from @i4replication.bsky.social and several commentaries about the work.

www.nature.com/collections/...

1/

SCORE | Center for Open Science SCORE shows that there is no shortcut to producing credible research findings, and there is no single indicator of trustworthiness. Research progress depends on transparency, rigor, and establishing r...

SCORE, a collaboration of 865 researchers, is now released as three papers in Nature, six preprints, and a lot of data (cos.io/score/). SCORE examined repeatability of findings from the social-behavioral sciences and tested whether human and automated methods could predict replicability.

1/🧵 A major update to our paper: "Scaling Reproducibility" w/ Leo Yang [Cross-posted from X]

We move beyond reanalyzing a single design to (almost) full-paper replication!

Paper: bit.ly/repro-ai

I describe it as such as a historian of disability because I want to remind scientists and the world that the history of science IS the history of eugenics, scientific racism and medical racism, and more.
I’m not comforting them with the idea that it was a ‘pseudoscience.’
This is your history.

Three years ago I was wondering whether LLM could act in “social groups” of any sort, and how such networks would resemble real ones.

A brilliant student, Nicola Zomer, came to me to ask for a thesis on exactly this question.

Today, that work is published in NPJ AI!

Paper: rdcu.be/e9gRH

1/🧵🧪

With Eugene Koonin, we propose a concept of “the selfish ribosome”, under which evolution of life is viewed as a ribosomal takeover, where the ribosome evolved to consume most of the cell’s resources, while other cellular componentry ensures the propagation of the ribosome. arxiv.org/abs/2602.23268

Bots have made their way to Prolific experiments. Our lab has stopped online testing of adults entirely now for this reason - we want to know if what we study is real. Probably data collected 2-3 years ago are ok, but moving forward we just can't know. www.pnas.org/doi/10.1073/...

How horrible to be a CS grad student under pressure to submit multiple first author papers to every conference deadline, whether they feel ready or not. This serves no one’s best interests in long run (science included). But lots of students appear to being getting advice it’s necessary to compete

Within the topic of AI alignment, there are a million tinier, but consequential, alignment choices.

This paper looks at the willingness of AI to engage in scientific misconduct (p-hacking). The most recent AIs resist instructions to p-hack and do good analysis, but the guardrails can be breached.

New preprint: “A dynamical perspective on biological reproduction”

hal.science/hal-05491732

The prevailing view is that an organism reproduces by building a new organism from its genomic representation, like von Neumann’s self-reproducing machine. But...
1/11

📖 ≠ 🧪 The Story is Not the Science.
Code is submitted but rarely executed during peer review—an issue likely to worsen with research agents. 🧑‍🔬
We introduce 𝐌𝐞𝐜𝐡𝐄𝐯𝐚𝐥𝐀𝐠𝐞𝐧𝐭, an execution-grounded evaluation of narrative + execution. 𝐕𝐞𝐫𝐢𝐟𝐲 𝐭𝐡𝐞 𝐬𝐜𝐢𝐞𝐧𝐜𝐞, 𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐭𝐡𝐞 𝐬𝐭𝐨𝐫𝐲.
1/n

I have plenty of reservations about the reliability of funnel-plots and z-curves and such, as well as their interpretation.....

But holy shit look at that.

From death comes diversity - Nature Ecology & Evolution A high-throughput laboratory experiment tracking the assembly of soil-derived communities shows that species-rich bacterial necromass supports increasingly diverse communities, with each additional de...

From death comes diversity! 🦠🧪🌐🌏

www.nature.com/articles/s41...

Advancing regulatory variant effect prediction with AlphaGenome - Nature AlphaGenome, a deep learning model that inputs 1-Mb DNA sequence to predict functional genomic tracks at single-base resolution across diverse modalities, outperforms existing models in variant effect...

I had intended to post something about this new Google DeepMind paper that appeared yesterday in Nature, but the press coverage has added to what there is to say. So this is a long 🧵
www.nature.com/articles/s41...

If you tell an AI to convince someone of a true vs. false claim, does truth win? In our *new* working paper, we find...

‘LLMs can effectively convince people to believe conspiracies’

But telling the AI not to lie might help.

Details in thread

An attempt to express how I principally use LLMs.

Rotating the Space: On LLMs as a Medium for Thought
sbgeoaiphd.github.io/rotating_the...

How complex should network models be?

🚨 In our latest paper we quantify (if and) when higher-order interactions are informative versus reducible to pairwise structure without losing functional signal (e.g., diffusion behavior).

👉 www.nature.com/articles/s41...

1/

What if animals emerged by installing a new biological operating system that repurposed what already existed, much like the rise of the smartphone? Here's our new paper in @embojournal.org @ibe-barcelona.bsky.social @melisupf.bsky.social @sfiscience.bsky.social link.springer.com/article/10.1...

We keep saying: "AI will handle the boring stuff, and humans will supervise." But the problem is--as AI reliability improves, it becomes really hard to motivate a human to conscientiously monitor it.

In a new WP with Gerard Cachon, we describe the "human-AI contracting paradox."

Measuring Intrinsic Dimension of Token Embeddings In this study, we measure the Intrinsic Dimension (ID) of token embedding to estimate the intrinsic dimensions of the manifolds spanned by the representations, so as to evaluate their redundancy quant...

📄 Measuring Intrinsic Dimension of Token Embeddings (2025) arxiv.org/abs/2503.02142
📄 Do We Really Need All Those Dimensions?
An Intrinsic Evaluation Framework for Compressed Embeddings (EMNLP 2025) aclanthology.org/anthology-fi...

Evolution by natural induction Abstract. It is conventionally assumed that all evolutionary adaptation is produced, and could only possibly be produced, by natural selection. Natural ind

A new, long paper on evolution - natural induction - split into 2:

royalsocietypublishing.org/rsfs/article...

royalsocietypublishing.org/rsfs/article...

@RichardWatson90 and Tim Lewens

I know some of you have strong views on LLMs and might not agree with me on this, but if you genuinely value diversity in academia, e.g., in welcoming neurodivergent researchers and non-native speakers, then I think you should acknowledge the positive influence AI can have in fostering inclusivity.

Closing the loop: how semantic closure enables open-ended evolution? Abstract. This study explores the evolutionary emergence of semantic closure—the self-referential mechanism through which symbols actively construct and in

Happy to share my new paper w/ @cgershen.bsky.social, just published at @royalsocietypublishing.org Interface!

Open Access🔓: royalsocietypublishing.org/rsif/article...

Instead of proposing a new theory, we offer a synthesis in theoretical biology. Want to know more? Read the full thread./1 👇🧵

Benchmarks from historians show that AI transcription from handwriting is now better than human, and a very cheap model is as good as people.

There are now massive troves of documents that could be made available for research that would have been impossible or prohibitive to transcribe before.

🚨 What if evolution is the ”law”… and networks are the machines that do the work?

In this paper (just published) I try to formalize how living systems are non-equilibrium, information-processing, adaptive matter. With a great biological flavor! 🧪🌐🌍🧬🦠

👉 iopscience.iop.org/article/10.1...

🧵 1/

Centuries of ontological dualisms (even recently permeating the literature) have muddied goal-directedness as something mystical. 🔮

It’s time to naturalize this concept and unpack its relationship to agency!/1

#complexitycat 😼

www.complexitycat.org/posts/goal-d...