Dyadic neural dynamics: extending representation learning to social neuroscience
Dyadic neural dynamics: extending representation learning to social neuroscience
Social communication fundamentally involves at least two interacting brains, creating a unique modeling problem. We present the first application of Contrastive Embedding for Behavioral and Neural Analysis (CEBRA) to dyadic EEG hyperscanning data, extending modeling paradigms to interpersonal neural dynamics. Using structured social interactions between participants, we demonstrate that CEBRA can learn meaningful representations of joint neural activity that captures individual roles (speaker-listener) and other behavioral metrics (autism quotient/AQ-10, etc.). Our approach to characterizing interactions, as opposed to individual neural responses to stimuli, addresses the key principles of foundational model development — scalability and cross-subject generalization —while opening new directions for representation learning in social neuroscience and clinical applications.
Glushanina, Mariya
992d97aa-b6e2-4799-a793-532762715e06
Huang, Jeffrey
42a7e64e-3034-4434-9d6b-b2266219fcf7
Mcleod, Michelle
6103710b-3805-4a96-a93c-3d5c4847ca77
Ames, Brendan
8ca36119-6cf2-495b-9cf0-983c976e12f7
Malaia, Evie A.
7d1b42ff-6599-4b36-9c96-ed66d4490e77
12 March 2026
Glushanina, Mariya
992d97aa-b6e2-4799-a793-532762715e06
Huang, Jeffrey
42a7e64e-3034-4434-9d6b-b2266219fcf7
Mcleod, Michelle
6103710b-3805-4a96-a93c-3d5c4847ca77
Ames, Brendan
8ca36119-6cf2-495b-9cf0-983c976e12f7
Malaia, Evie A.
7d1b42ff-6599-4b36-9c96-ed66d4490e77
Glushanina, Mariya, Huang, Jeffrey, Mcleod, Michelle, Ames, Brendan and Malaia, Evie A.
(2026)
Dyadic neural dynamics: extending representation learning to social neuroscience.
Computational and Systems Neuroscience (COSYNE) 2026, Lisbon and Cascais, Portugal, Lisbon and Cascais, Portugal.
12 - 17 Mar 2026.
10 pp
.
(doi:10.48550/arXiv.2509.23479).
Record type:
Conference or Workshop Item
(Poster)
Abstract
Social communication fundamentally involves at least two interacting brains, creating a unique modeling problem. We present the first application of Contrastive Embedding for Behavioral and Neural Analysis (CEBRA) to dyadic EEG hyperscanning data, extending modeling paradigms to interpersonal neural dynamics. Using structured social interactions between participants, we demonstrate that CEBRA can learn meaningful representations of joint neural activity that captures individual roles (speaker-listener) and other behavioral metrics (autism quotient/AQ-10, etc.). Our approach to characterizing interactions, as opposed to individual neural responses to stimuli, addresses the key principles of foundational model development — scalability and cross-subject generalization —while opening new directions for representation learning in social neuroscience and clinical applications.
Text
2509.23479v1
- Author's Original
More information
Accepted/In Press date: 22 December 2025
Published date: 12 March 2026
Additional Information:
Presenting author Mariia Glushanina will present this work as a poster presentation at COSYNE 2026.
We submitted this paper to the NeurIPS 2025 Workshop BrainBodyFM on 08/22/2025, which was unfortunately not accepted. A working draft of this manuscript is shared on arxiv; we plan to expand this draft and submit for publication in 2026.
Venue - Dates:
Computational and Systems Neuroscience (COSYNE) 2026, Lisbon and Cascais, Portugal, Lisbon and Cascais, Portugal, 2026-03-12 - 2026-03-17
Identifiers
Local EPrints ID: 509265
URI: http://eprints.soton.ac.uk/id/eprint/509265
PURE UUID: eccc654d-fb79-498b-b7d0-8fcaea936835
Catalogue record
Date deposited: 16 Feb 2026 17:59
Last modified: 17 Feb 2026 03:09
Export record
Altmetrics
Contributors
Author:
Mariya Glushanina
Author:
Jeffrey Huang
Author:
Michelle Mcleod
Author:
Brendan Ames
Author:
Evie A. Malaia
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics