Oops — here’s the right link to the article:
www.biorxiv.org/content/10.1...

22.04.2025 07:59 — 👍 1    🔁 1    💬 0    📌 0

Dissociating the functional role of the para-hippocampal and the parietal cortex in human multi-step reinforcement learning Many real-life decisions involve a tension between short-term and long-term outcomes, which requires forward-looking abilities. In reinforcement learning, this tension arises at the initial stage of m...

Hello, and thanks for your interest in our work!

I accidentally shared the wrong link earlier — here is the correct one to read the article:
www.biorxiv.org/content/10.1...

22.04.2025 07:55 — 👍 3    🔁 1    💬 0    📌 0

bioRxiv Manuscript Processing System Manuscript Processing System for bioRxiv.

🙏 Thanks to the amazing collaborators

Alexandre Salvador, Sabrine Hamroun, @mael-lebreton.bsky.social and @stepalminteri.bsky.social

📄 Want all the details? Read the full preprint here: submit.biorxiv.org/submission/p...

16.04.2025 10:03 — 👍 2    🔁 0    💬 0    📌 0

So, what did we learn?

🧩 The parahippocampal cortex supports forward planning

🧩 The parietal cortex helps build an internal model of the task’s structure

A dissociation of roles in multi-step reinforcement learning!

16.04.2025 10:03 — 👍 0    🔁 0    💬 1    📌 0

🧠 At outcome time, brain activity highlighted a different network — involving the bilateral parietal cortex.

→ Its activity correlated with the structure learning signal.

16.04.2025 10:03 — 👍 0    🔁 0    💬 1    📌 0

🧠 Results showed that first-stage decisions — the forward-looking ones — recruited a network centered on the parahippocampal cortex.

→ This region appears crucial for model-based planning.

16.04.2025 10:03 — 👍 0    🔁 0    💬 1    📌 0

🎯 Participants (N=28) completed this task inside an fMRI scanner.
We analyzed both their behavior and brain activity to understand the neural basis of:
– Forward planning
– Structure learning (i.e., learning transition probabilities)

16.04.2025 10:03 — 👍 0    🔁 0    💬 1    📌 0

We used a two-step reinforcement learning task where participants chose between abstract symbols that probabilistically led to different states. Crucially, first-step options with low immediate rewards led more often to high-reward second states, encouraging forward planning. Participants had to learn both state-transition structure and reward values simultaneously, without instructions — capturing how people balance short-term and long-term outcomes in real life.

We used a two-step reinforcement learning task

At each first-stage choice, participants had to think ahead:
– Should I go for the immediate reward?
– Or choose an option that brings me to a better future state?

16.04.2025 10:03 — 👍 1    🔁 0    💬 1    📌 0

Real-life decisions often involve a trade-off between short- and long-term outcomes.

Choosing a small reward now vs. a bigger one later? That’s forward thinking.

But how does the brain manage that?

16.04.2025 10:03 — 👍 0    🔁 0    💬 1    📌 0

bioRxiv Manuscript Processing System Manuscript Processing System for bioRxiv.

🚨 New preprint on bioRxiv!

We investigated how the brain supports forward planning & structure learning during multi-step decision-making using fMRI 🧠

With A. Salvador, S. Hamroun, @mael-lebreton.bsky.social & @stepalminteri.bsky.social

📄 Preprint: submit.biorxiv.org/submission/p...

16.04.2025 10:03 — 👍 23    🔁 9    💬 2    📌 3