🧠 How do transformers learn relational reasoning? We trained small transformers on transitive inference (if A>B and B>C, then A>C) and discovered striking differences between learning paradigms. Our latest work reveals when and why AI systems generalize beyond training data 🤖

🚨 Job adverts LIVE! 🚨 We want to understand how the brain supports abstract cognition in psychosis, using cutting-edge neuroscience. Non-Clinical RA: lnkd.in/eva-Zzfj Clinical RA: lnkd.in/eSgdad_F Postdoc: lnkd.in/eid8gtbR Oxford University Due June 23

“The gender-equality paradox entails that countries with more equality have larger gender differences in behavior. Positive associations between equality & gender differences disappear or reverse when accounting for cultural clusters of countries. The paradox appears to be a methodological artifact”

🚨 Job adverts LIVE! 🚨 We want to understand how the brain supports abstract cognition in psychosis, using cutting-edge neuroscience. Non-Clinical RA: lnkd.in/eva-Zzfj Clinical RA: lnkd.in/eSgdad_F Postdoc: lnkd.in/eid8gtbR Oxford University Due June 23

How do cognitive maps fail? And how can this help us understand/treat psychosis? My lab at @oxexppsy.bsky.social is hiring a Postdoc tinyurl.com/2p935hhz and RA tinyurl.com/3myfpb78 to answer these questions in mouse models. Here's why you might want to join: 🧵

2-The people: you’ll be part of an ambitious and highly collaborative team studying psychosis in humans and neural networks led by @mattnour.bsky.social at @oxfordpsychiatry.bsky.social @drrickadams.bsky.social at @ucl.ac.uk and @mariaeckstein.bsky.social at DeepMind

New paper from our lab now in @nature.com . We show that our brains use basis functions or "building blocks" to navigate social interactions. Rather than tracking every individual separately, social information is compressed into patterns that can be flexibly combined. www.nature.com/articles/s41...

📢 Register now: 🚨 The registration for the Computational Psychiatry Conference 2025 in Tübingen is now open - register and get the early bird discounts: www.conftool.com/cpconf2025/i... #CPConf2025

Why academia is sleepwalking into self-destruction. My editorial @brain1878.bsky.social If you agree with the sentiments please repost. It's important for all our sakes to stop the madness academic.oup.com/brain/articl...

❗RESEARCH OPPORTUNITIES❗ We're hiring for an ambitious @wellcometrust.bsky.social project exploring brain mechanisms of planning/inference in psychosis 🧠 Tackled at multiple scales: from single neurons to human behaviour/fMRI and neural networks. Opportunities for 1 post-doc 1 clinical fellow 👀

Hope you'll forgive a quick thread as I take over as the new EiC of @elife.bsky.social

Comment on our recent paper challenging the link between MMN dysfunction and positive symptoms: www.sciencedirect.com/science/arti... Thanks @mattnour.bsky.social I may have missed it but cognitive map dysfunction is also unrelated to positive symptoms so far?

Our symposium on structure knowledge biases in psychopathology was accepted at #CPConf2025! with @mattnour.bsky.social @mgarvert.bsky.social @carocharp.bsky.social Talks will span social learning, inductive inference in planning, language representation & map-based generalization. Can't wait!

Curious to see how they plan to anonymise our NHS data for training ML models. I personally wouldn’t know how to start doing that www.theguardian.com/politics/202...

Very excited to lead this work at @ox.ac.uk. We're developing computational tools for neuroimaging data and naturalistic behaviour, some based on LLMs. We hope these will yield new insights into the mechanisms of CNS drugs, neurobiology of psychosis, and human cognition more broadly.

Excellent review paper (by @rob-mccutcheon.bsky.social @mattnour.bsky.social & others) Clearly lays out the historical challenges and future frameworks to guide drug development in psychiatry. www.biologicalpsychiatryjournal.com/article/S000...

In principle, ~10^15 bits needed to specify all connections in brain assuming any wiring diagram is possible In practice, much less needed. Indeed innate behaviors must fit in the genome (~10^9 bits in humans) So: GPT4 has enough parameters. Just not the right ones www.pnas.org/doi/10.1073/...

We're going all out to try and understand mechanisms of cognitive symptoms in psychosis - including human neuroimaging with @mattnour.bsky.social, computational work with @mariaeckstein.bsky.social and @drrickadams.bsky.social and mouse electrophys/optogenetics with me ... watch this space!

💥 Some POST DOC JOB news 👀... I'm very happy to say that @melgaby.bsky.social, @mattnour.bsky.social, @mariaeckstein.bsky.social and I have been awarded funding from @wellcometrust.bsky.social to look at how planning 📝and mental models of the world 🌍 are affected by psychosis...

There's a movement in neuroscience suggesting we should be pursuing bigger bets with larger teams. I think there's a case for doing a bit of this, but I think it's a bad idea to prioritise it for two reasons, and a good case for saying we should be moving in the exact opposite direction. 🧠🧪

We have a starter pack that we will continue to expand that represents the #MentalHealth and #Neuroscience research community. Please let us know if we missed you! #Psychology #AcademicSky #Research go.bsky.app/6LKA9F4

OK If we are moving to Bluesky I am rescuing my favourite ever twitter thread (Jan 2019). The renamed: Bluesky-sized history of neuroscience (biased by my interests)

I'm happy to share our Review article, "Cognitive Maps and Schizophrenia", synthesising evidence for representation-level abnormalities in the condition, and proposing directions for future work. Out now in Trends in Cognitive Science. www.sciencedirect.com/science/arti...