Robert Hawkins

Excited that this is now out in @nathumbehav.nature.com 🎉

David Rose (davdrose.github.io) led this project on how children's understanding of causal language develops.

📃 (preprint): osf.io/preprints/ps...
📎: github.com/davdrose/cau...

🧩Natural language isn’t all you need.

We’re great at evaluating text-based reasoning (MATH, AIME…) but what about long-horizon visual reasoning?

Enter 𝗞𝗻𝗼𝘁𝗚𝘆𝗺: a minimalistic testbed for evaluating agents on spatial reasoning along a difficulty ladder

A figure demonstrating the different aspects of the corpus described in the tweet. There is a main isomorphic 3D view of a level in the Portal 2 co-op game, with some portals, lasers, and the blue and orange players. Inset, there are first-person captures of the blue and orange player views. There is also a box containing the transcribed dialogue with timestamps and labels for the discursive acts. Finally, there is a box containing a task and a list of subtasks. Some subtasks are already crossed out, with the time that they have been completed. The last subtask ("Player 2 places portal 4 on wall 4") is marked incomplete. The dialogue is as follows: Blue: Can you put your other portal up here? (tagged as directive) Orange: Where? (tagged as request for clarification) Blue: On uh, on this wall. (tagged as directive) Blue: So that it uh points at the circle. (tagged as directive) Orange: Okay. (tagged as commit) The full list of subtasks is: Task: Redirect lasers Subtask: Player 1 places portal 1 on wall 1. (completed) Subtask: Player 1 polaces portal 2 on wall 2 or 3. (completed) Subtask: Player 2 places portal 3 opposite of portal 2. (completed) Subtask: Player 2 places portal 4 on wall 4. (incomplete)

A couple years (!) in the making: we’re releasing a new corpus of embodied, collaborative problem solving dialogues. We paid 36 people to play Portal 2’s co-op mode and collected their speech + game recordings.

Paper: arxiv.org/abs/2512.03381
Website: berkeley-nlp.github.io/portal-dialo...

1/n

🚨 New in Nature+Science!🚨
AI chatbots can shift voter attitudes on candidates & policies, often by 10+pp
🔹Exps in US Canada Poland & UK
🔹More “facts”→more persuasion (not psych tricks)
🔹Increasing persuasiveness reduces "fact" accuracy
🔹Right-leaning bots=more inaccurate

How Conflict Aversion Can Enable Authoritarianism:An Evolutionary Dynamics ApproachChad M. TopazDecember 6, 2025AbstractWe use evolutionary game theory to examine how conflict-averse centrism can unintentionallyfacilitate authoritarian success in polarized political conflicts. Many such conflicts are asym-metric: authoritarian actors can employ norm-breaking or coercive tactics, while democraticresistance faces stronger constraints on what counts as normatively acceptable behavior. Yetformal models typically treat opposing sides symmetrically and rarely examine conflict-aversebehavior. Drawing on empirical research on protest backlash, civility norms, and authoritarianresilience, we model these dynamics as a three-strategy evolutionary game in which resistance,authoritarianism, and conflict-averse centrism interact under replicator dynamics. This frame-work yields two distinct outcomes—cyclic resurgence of authoritarian strength through a hete-roclinic cycle and a stable centrist–authoritarian coalition that excludes resistance—dependingon how actors respond to confrontation. The analysis shows how payoff differences can reor-ganize long-run dynamics in asymmetric conflicts. Our contribution is to demonstrate how anestablished dynamical framework, combined with empirically grounded behavioral assumptions,clarifies the strategic conditions under which conflict aversion can diminish the effectiveness ofdemocratic resistance.

Ok activists and nerds... I promised that I was SO MAD ABOUT HOW CONFLICT-AVOIDANT "CENTRISTS" UNDERMINE RESISTANCE TO FASCISM that I would resort to math to make this point and that I would write it up in a paper. So, here ya go. Share, enjoy, and get up on that soapbox. osf.io/preprints/so...

Another fun project from @yangxiang.bsky.social. She asks the question: do people assign responsibility to personality traits in the same way that they assign reponsibility to people? The answer: sort of!

osf.io/preprints/ps...

Well this is exciting!

The Department of Psychological & Brain Sciences at Johns Hopkins University (@jhu.edu) invites applications for a full-time tenured or tenure-track faculty member in Cognitive Psychology, in any area and at any rank!

Application + more info: apply.interfolio.com/178146

The U.S. Is Funding Fewer Grants in Every Area of Science and Medicine A quiet policy change means the government is making fewer bets on long-term science.

“Thousands of very competitive projects in areas like cancer, diabetes, aging, neurological disorders and public health improvements most like went unfunded in 2025.
Similarly, at the National Science Foundation, the roughly 3,000 fewer new grants encompassed reductions to every area of science”

Clayton Littlejohn, Disinformation is for Degrading the Value of Information, not Confirming Falsehoods - PhilPapers According to a recent account of disinformation, disinformation is content that “generates ignorance” (Simion 2024a; 2024b). The view improves upon previous accounts that focused upon the potential fo...

I have news. Wonderful news.

New paper on #disinformation. The title is the thesis. Think of disinformation as content that functions to optimise the decisions of the audience for the disinformant. This doesn't require deception or content that induces false belief.

philpapers.org/rec/LITDIF

On major questions of university governance, a 99.3% majority vote of the faculty means nothing to the people who are actually in charge. This is a fact all of us need to take much more seriously.

My alma mater, Northwestern, has decided to grovel to a desperate loser, at vulnerable people’s expense

Monkeys have rhythm Synchronizing movements to music is a hallmark of human culture, but its evolutionary and neurobiological origins remain unknown. This ability requires (i) extracting a steady rhythmic pulse, or beat,...

New in Science, Macaques tap to the beat.

Very cool study for its main result and its null one: consistent with nearly every other comparative study of music, monkeys don't differentiate beats by their relative strength—which even young children do innately. Monkeys have rhythm but not meter!

Almost off to @euripsconf.bsky.social in Copenhagen 🇩🇰 🇪🇺! I'll present 3 posters:

🧠 Neurosymbolic Diffusion Models: Thursday's poster session.

Going to NeurIPS? @edoardo-ponti.bsky.social and @nolovedeeplearning.bsky.social will present the paper in San Diego Thu 13:00
arxiv.org/abs/2505.13138

"There is a power law of joint communicative effort and it reflects communicative work"

📢New paper from: Sara Bögels (sarabogels.bsky.social), Tianyi Li, Marlou Rasenberg, Lotte Eijk, Ivan Toni, & Wim Pouw (wimpouw.bsky.social)

www.sciencedirect.com/science/arti...

Assessing Others’ Knowledge Through Their Speech Disfluencies and Gestures As part of the multimodal language system, gestures play a vital role for listeners, by capturing attention and providing information. Similarly, disfluencies serve as a cue for the listeners about o...

Our new paper in Cognitive Science, led by Can! @canavci.bsky.social

"Assessing Others’ Knowledge Through Their Speech Disfluencies and Gestures"

Cognitive Science - Wiley Online Library onlinelibrary.wiley.com/doi/10.1111/...

@kudilvebilis.bsky.social

Building compositional tasks with shared neural subspaces - Nature The brain can flexibly perform multiple tasks by compositionally combining task-relevant neural representations.

Task learning is compositional, so would behavior be!
www.nature.com/articles/s41...

What does it mean to understand language? Language understanding entails not just extracting the surface-level meaning of the linguistic input, but constructing rich mental models of the situation it describes. Here we propose that because pr...

What does it mean to understand language? We argue that the brain’s core language system is limited, and that *deeply* understanding language requires EXPORTING info to other brain regions.
w/ @neuranna.bsky.social @evfedorenko.bsky.social @nancykanwisher.bsky.social
arxiv.org/abs/2511.19757
1/n🧵👇

In our new tutorial, @meanwhileina.bsky.social and I walk you through meaningful hypothesis testing using a less known form of calculating Bayes Factor.

osf.io/preprints/ps...

People transitioning to Bayes often wonder how to decide if an effect is meaningful or not. Here you go... 1/🧵

We recently released OlmoEarth—an open spatio-temporal foundation model for planetary intelligence, trained on multimodal Earth observation, OpenStreetMap, & other open data. It sets a new accuracy/efficiency Pareto frontier and powers our OlmoEarth Platform. Technical report now on arXiv. 👇

I am very excited to share our new preprint, spearheaded by the brilliant @lunahuestegge.bsky.social, w/ @peterkok.bsky.social and others: ‘An attempt to push mental imagery over the reality threshold using non-invasive brain stimulation’

doi.org/10.31234/osf...

Assistant Professor, Associate Professor or Full Professor of Cognitive Science - Vacancy at Aarhus University Vacancy at School of Communication and Culture - Linguistics, Cognitive Science and Semiotics, Dept. of, Aarhus University

CogSci in Aarhus is hiring, open rank (assi, asso or full prof).
We want somebody working on and teaching computational modelling of cognitive processes and/or social processes. Students are amazing, work/life balance very satisfactory, and colleagues are nice!

international.au.dk/about/profil...

New paper from the lab, "Perceiving Event Structure in Brief Actions," now out in Cognitive Psychology :)

Led by the inimitable Zekun Sun

This was my lab's first foray into event cognition

gift link: sciencedirect.com/science/arti...

How physical information is used to make sense of the psychological world - Nature Reviews Psychology Reasoning about minds and reasoning about physical objects are governed by two distinct systems. In this Perspective, Liu et al. review research from developmental psychology and cognitive neuroscienc...

New perspective paper (w/ @sedaakbiyik.bsky.social, Joseph Outa, & @minjaek.bsky.social ) in @natrevpsychol.nature.com ⚽💭🧠👶 : www.nature.com/articles/s44...

I'm excited to announce that I have been asked to join the Mamdani Transition Team to work on community safety issues. A New Era for NYC.

We have a new paper in Science Advances proposing a simple test for bias:

Is the same person treated differently when their race is perceived differently?

Specifically, we study: is the same driver likelier to be searched by police when they are perceived as Hispanic rather than white?

1/

📍Excited to share that our paper was selected as a Spotlight at #NeurIPS2025!

arxiv.org/pdf/2410.03972

It started from a question I kept running into:

When do RNNs trained on the same task converge/diverge in their solutions?
🧵⬇️

Disentangling the Roles of Distinct Cell Classes with Cell-Type Dynamical Systems Latent dynamical systems have been widely used to characterize the dynamics of neural population activity in the brain. However, these models typically ignore the fact that the brain contains multiple...

My collaborator @aditijh.bsky.social has some cool work on cell-type LDS :)

www.biorxiv.org/content/10.1...

Screenshot of paper title: Tokenisation over Bounded Alphabets is Hard.

Tokenisers are a vital part of LLMs, but how hard is it to find an optimal one? 🤔 Considering arbitrarily large alphabets, prior work showed this is NP-hard. But what if we use bytes instead? Or unary strings like a, aa, aaa, ...? In our new paper, we show this is still hard, NP-hard!

we released Olmo 3! lot of exciting stuff but wanna focus on:

🐟Olmo 3 32B Base, the best fully-open base model to-date, near Qwen 2.5 & Gemma 3 on diverse evals
🐠Olmo 3 32B Think, first fully-open reasoning model approaching Qwen 3 levels
🐡12 training datasets corresp to different staged training

proud to share this work, led by the brilliant @ilinabg.bsky.social, now out in Nature! Ilina finds that speech-sound neural processing is VERY similar in a language you know and one you don't. differences only emerge at the level of word boundaries and learnt statistical structure 🧠✨