This paper finds LLMs' ability to understand that others have different beliefs (Theory of Mind) comes from 0.001% of their parameters. Break those specific weights & the model loses both its ability to track what others know AND language comprehension

Interesting implications for models (& minds?)

I'm so excited to help bring this Computational Neuroscience workshop to life! Join a fantastic group of researchers in Denmark this August (11-14).
We've designed it for PhDs & postdocs to connect with experts and build new skills. Hope to see you there! Please do share with your colleagues!

Ready to shape the future of #ComputationalScience?
We’re inviting researchers like you to mentor student-led projects this July. Just 2-3 hours of commitment and no prep needed!

This is your chance to give back and grow!
➡️ Learn more and sign up to #volunteer neuromatch.io/mentoring/

✨Excited to share that our new paper is now out in iScience!✨

🧠 We show that people can coordinate surprisingly well in novel interactions by violating others' expectations - without requiring deep, recursive reasoning about others’ beliefs.

📄 Read the full paper here: www.cell.com/iscience/ful...

What about the social cerebellum during development we wondered? Here you find some cool insights on this awesome part of the brain! BIG thanks for the great work Katerina Manoli, and support from Charlotte Grosse Wiesmann and Frank van Overwalle 🧠🦋 rdcu.be/epKmZ

Ten simple rules for queer data collection and analysis dx.plos.org/10.1371/jour...

Funded PhD opportunity with @sjblakemore.bsky.social and I, at Cambridge. We are looking for someone interested in developmental science, to start in the coming academic year. Please share it with anyone you think might be interested (see details attached 😁).
drive.google.com/file/d/1RIvg...

Dorsomedial and ventromedial prefrontal cortex lesions differentially impact social influence and temporal discounting journals.plos.org/plosbiology/... new lab paper from the fantastic @zhilinsu.bsky.social with wonderful collaborators 🌟

Deadline extended to 28th April!! Summer School in Computational Social Cognition with fantastic keynotes by Matthew Rushworth, @dianatamir.bsky.social and @davidamodio.bsky.social. Come & learn more about computational modelling, make new friends, & explore the UK's 2nd largest city 🙂👇. Please RT!

Comment: "Can a single brain cell be surprised?"

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

There is again a model-based neuroscience summer school from Aug 4-8th in Amsterdam! modelbasedneurosci.com

getting this paper published was a bit painful but I am proud of it: www.nature.com/articles/s41.... We use deep RL to find mechanisms that help (real) people sustain the commons. Well done to Raphael Koster and Miruna Pislar (not on BlueSky). non-paywalled version on arxiv.

New paper “Basis functions for complex social decisions in dorsomedial prefrontal cortex” in @nature.com led by @mkwittmann.bsky.social with many others. We show basis functions code relations between people, similar to their role in other perceptual and motor domains www.nature.com/articles/s41...

Our paper in Nature (@mkwittmann.bsky.social et al.): the brain does not only process the *identity* of a person but primarily our *relationship* to them. Even on a neural level, who someone is *in relation to others* is key. www.nature.com/articles/s41...
#PsychSciSky #socialpsyc #neuroskyence

Think about conference location to avoid visa discrimination www.nature.com/articles/d41...

creative ways that students use chatGPT www.nature.com/articles/d41...

6️⃣Our results show that surprise is used to communicate in novel interaction, reflected in behavior, physiology, and neural activity. This suggests a cognitive mechanism shaping how we develop new communication systems and has implications for human communication, AI, and human-machine interaction.

5️⃣We thought that If Receivers recognize goals through surprise, we should see neural markers of expectation violations. Indeed, EEG data shows surprise is encoded in fronto-central brain regions, linking it to cognitive processes of detecting unexpected events.

4. To provide the physiological evidence of the model, we conducted model-based analysis of receive’s PDR data. Receivers’ PDR strongly correlates with model-derived surprise—suggesting that unexpected movements grab attention and influence cognitive processing.

3️⃣The Surprise Model accurately predicts Sender’s message design and participant behavior across different samples. It outperforms models using only movement or state priors, showing that all components are essential to capture human communication!

2️⃣TCG is a two-player game where the Sender moves on a grid-creating a message, while the Receiver must infer their hidden goal based on the observed message. After analyzing hundreds of messages, we identified three distinct message types senders used to communicate.

1️⃣To test this, we built a Surprise Model with two key components. The movement component assumes motion continues in a straight line, so deviations create surprise. The state prior encodes Sender’s goal information. We tested this model in the Tacit Communication Game (TCG).

Here we proposed that without a common language, people rely on universal physical principles to build shared understanding. Once a system is established, Senders strategically defy expectations to create meaning—making surprise a powerful communicative tool.

✨Excited to share that our paper on how expectation violations shape novel human communication has just been published in Nature Communications! ✨

📖 read the full paper: www.nature.com/articles/s41....

🧵 Detailed breakdown bellow! 👇