Algorithmic representations of sequential behaviour in the human brain - 2-071. Friday.
@skuechenhoff.bsky.social, Habiba Azab
ABCD task in humans, with single cells and fMRI. Multiple different sequence representations, and a crazy anatomical organisation of the future in PFC.
personal note: There aren't words to adequately describe Nill's fortitude and perseverance in continuing to do great science during this horrendous time for her and her family in Tehran. She is an inspiration!
We're currently digging into the brain data from the DRE patients and excited to discover the neural underpinnings of the context neglect that leads people to choose maladaptively when decks are shuffled.
There's a lot more in the preprint, here:
www.biorxiv.org/content/10.6...
We've also been running the task with drug-resistant epilepsy patients and modeling their behavior. Interestingly, every aspect of the patients' behavior was indistinguishable from that of non-epileptic participants, except their choice times (DRE pts were slower to choose):
The risk-sensitive models also best fit most participants' behavior, and critically, the latent risk-sensitive model variables (Q-values) best predicted participants' choices and response times:
To do this, we first had several basic reinforcement learning models do the task, showing that asymmetric (AKA, risk-sensitive, 2-process distributional) reinforcement learning models (i.e., those learning from both gains and losses) performed best:
This behavior, that reflects the disCARDing ;) of a context that participants clearly learned, is similar to a well-known cognitive fallacy called base-rate neglect.
We then sought to understand whether there were some latent computational principles underlying this process.
However, despite clearly learning how to behave with different card decks when they were separated, upon shuffling the card decks together, participants reverted to behavior reflecting the deck with the uniform distribution:
participants learned a latent choice boundary for each card deck, reflected in multiple behavioral measurements:
In a novel gambling task, played with three card decks, each with a different distribution of cards 1-9, humans were instructed to choose whether their card was higher or lower than a virtual opponents card; first with each deck separately, then with all decks mixed.
New (purely behavioral, computational psychiatry) preprint out from the neurosmiths, led by the brilliant @niloufarshahdoust.bsky.social
We hope this work will help us learn about the computational bases of impulsive choices underlying numerous mental health disorders.
skeeprint below:
Compositional representation of self, others, and gaze direction in *human* hippocampus - super cool.
arxiv.org/abs/2603.04747
Yay! Congrats, @hazimi.bsky.social!
Tragic. Please pass this along to US Chairs and Deans, and consider it when you sit on promotion committees. This, combined with the review backlog, is dire - especially for our early career researchers.
$1B per day is about $42 million per hour. About 30 min worth of that spending can fund my lab for my ENTIRE career (~20 3-year NSFs or ~10 5-year R01s). Please wrap your head around that.
Incredible news!
Shervin Rahimpour, MD, has joined the Neurona trial for the implantation of stem cells to treat drug-resistant epilepsy.
Although still experimental, this transformative trial represents a huge step forward in care without removing or ablating part of the brain.
At Your Neighborhood Scientist, we believe public understanding starts with listening. What science question do you want answered? Maybe it's something you’ve seen online and aren’t sure about?
Submit it here and we'll answer it: forms.gle/wPyPUVcZ2LbG...
#SciComm #STEM #YourNeighborhoodScientist
Where you look next isn’t arbitrary.
In our new paper, we model human eye movements in immersive visual search as reinforcement learning under cognitive constraints. 🧵
Everybody still on there convinced that they are a special flower unaffected by propaganda.
3) social media is not only for advocacy, but for interaction. Even the bigger accounts that are ostensibly doing science advocacy over there are still looking at X, which at this point is an informational dumpster fire. Maybe just my feed, but I doubt it.
2) empirically, the biggest accounts are staying on twitter. Is it because they have a deep commitment to advocacy in adversarial spaces, or they have a lingering attachment to their following? Are marginalized perspectives promoted in this case? Is that good for sci comms?
1) the cost function of social media is to increase engagement by promoting the most inflammatory posts. Not ideal for science comms.
So you all are still looking at X? I totally agree that it’s super important to advocate for science everywhere, but I think it’s a little more complicated than that, based on how social media works. 3 quick points:
My friend & colleague, @bkundu.bsky.social, is hiring for a super interesting & well-paid position:
neurojobs.sfn.org/job/39531/le...
#neurojobs
@summerfieldlab.bsky.social and I are very happy to share this paper! Building on work by @scychan.bsky.social, we show that how people learn depends on the distribution of examples they see, and changes in a way that’s very similar to transformer models.
I am looking to hire 2-3 post-docs over the course of the next few months to work on questions related to cognitive control in humans, broadly construed. EEG, TMS, DBS, sEEG, fMRI or related methodological experience preferred.
Apply here:
jobs.uiowa.edu/jobSearch/po...
Lab website: wessellab.org
New paper alert 📣 #Neuroskyence
"Rhythmic sampling of multiple decision alternatives in the human brain" @natcomms.nature.com
together with @ycaoneuro.bsky.social @maryamtohidi.bsky.social @donnerlab.bsky.social @ktsetsos.bsky.social
www.nature.com/articles/s41...
🧵1/
If you work at the intersection of computational neuroscience and machine learning, consider applying for this postdoc position (January 2027 start date):
academicpositions.harvard.edu/postings/15868
An opportunity to work with a great group of people across Harvard, MIT, and UC Berkeley.
Our paper is out in @natneuro.nature.com!
www.nature.com/articles/s41...
We develop a geometric theory of how neural populations support generalization across many tasks.
@zuckermanbrain.bsky.social
@flatironinstitute.org
@kempnerinstitute.bsky.social
1/14
