Post by Ignacio Saez, PhD

How does the human brain filter out the noise to learn and act efficiently in a chaotic world? We live in the loudest era in human history, inundated with high-dimensional data. To learn effectively, the brain must compress this overwhelming environment, focusing strictly on what is task-relevant while ignoring the rest. In our newest preprint, we use intracranial recordings and computational modeling to explore how selective attention shapes learning and decision-making in the human brain. 🔗 Read the full preprint on bioRxiv:  https://lnkd.in/gkYdwW5c We worked with patients undergoing intracranial monitoring who performed a multidimensional reinforcement learning task. We combined three powerful approaches: -Computational cognitive modeling -Local field potential (LFP) recordings -Single-unit recordings Here is what we discovered about how the brain compresses complex environments for efficient learning: Attention drives performance: Participants with stronger selective attention to the relevant feature dimension maintained more compressed task-state representations, performed better, and responded faster. Prefrontal value encoding: Both the orbitofrontal cortex (OFC) and lateral prefrontal cortex (LPFC) encoded expected value in high-frequency activity (60–200 Hz) and theta-band activity (3–8 Hz). However, their temporal dynamics differed: - OFC value signals were transient and centered around decisions. - LPFC value signals were more sustained across the decision epoch, and critically, were strongly modulated by attentional strategy — with stronger attention linked to earlier and stronger LPFC value encoding. Attention-dependent circuit coordination: Value-related LPFC-OFC theta-band coordination emerged before a choice was made in high-attention individuals, but only after in low-attention individuals. During the pre-choice period, local LPFC high-frequency value signals predicted this theta-band circuit coordination, and this relationship was itself shaped by the participant's attentional strategy. Bridging the scales: At the cellular level, single neurons in the OFC encoded both expected value and the relevant feature dimension. Crucially, the directionality of this single-unit coding mirrored population-level HFA signals, bridging the gap from local neurons to mesoscale networks. The key takeaway: These findings provide a mechanistic framework for how the brain dynamically coordinates attention and valuation systems across prefrontal circuits, allowing us to build compressed, efficient representations of the world during learning. Congratulations to lead author Christina Maher, PhD, Salman Qasim, Gelana Tostaeva, Lizbeth Nuñez Martinez, Saad Nadeem, fedor panov, Angela Radulescu, and the incredible patients who volunteered their time to take part in this research.