Guillaume Dumas Social Cognition Biorhythmic Coupling and the Shared Consciousness Question at MoC7

Guillaume Dumas will keynote Models of Consciousness 7 (MoC7) at the HC Ørsted Institute, University of Copenhagen, October 12-16, 2026. He is Associate Professor of Computational Psychiatry at Université de Montréal and a principal investigator at CHU Sainte-Justine Research Centre. His research programme asks what happens to consciousness when two brains interact, an angle that most formal consciousness theories do not address.

The standard experimental frame in consciousness science isolates a single subject, presents stimuli, and measures the neural signatures of that individual’s experience. Dumas’s work systematically challenges whether this is the right level of analysis for the class of conscious experiences that are most distinctively social.

Inter-brain synchronization as a research object

Dumas conducts hyperscanning research, a methodology that simultaneously records neural activity from two or more interacting participants. The central finding from this line of work, published in journals including PLOS ONE and NeuroImage, is that the brains of interacting individuals show correlated neural oscillatory activity across multiple temporal scales during face-to-face interaction. This inter-brain synchronization (IBS) is not simply a shared response to the same external stimulus: it appears specifically in the dynamic coordination of the interaction itself, across turn-taking, joint attention, and synchronised movement.

The significance of this finding for consciousness science is that some features of social conscious experience may be produced at the level of interaction dynamics rather than within either individual brain. The felt sense of genuine contact with another mind, of shared attention, of mutual understanding in conversation, and of coordinated action in joint tasks, are all candidates for experiences whose neural substrates are distributed across interacting agents rather than localised within one of them.

If that is correct, a theory of consciousness that specifies only conditions on individual neural systems will be incomplete. It will be able to account for individual perceptual experience but not for the class of experiences that are constituted in interpersonal dynamics.

Biorhythmic coupling and the temporal structure of shared experience

The analytical concept Dumas uses is biorhythmic coupling. During social interaction, neural oscillations in interacting participants become coordinated at specific frequency bands. This coupling is measurable using standard EEG and MEG equipment applied in hyperscanning configurations. It correlates with the quality of social communication: stronger coupling is associated with successful joint action and with participant reports of higher quality social contact. Disrupted coupling is associated with the communicative breakdowns characteristic of autism spectrum conditions, a research focus of Dumas’s clinical work at CHU Sainte-Justine.

The theoretical implication of biorhythmic coupling research is that social conscious experience has a temporal structure that cannot be read from either participant’s brain in isolation. The interaction generates temporal patterns that are constitutive of shared experience. Both participants contribute to producing a joint neural state that neither would produce independently.

This positions Dumas’s work in productive tension with two major theoretical traditions that have been extensively discussed on this site. Francisco Varela, Evan Thompson, and Alva Noë’s enactivist account holds that consciousness is enacted through coupling between an organism and its environment. Thompson and Noë’s analysis of what enactivism requires of AI consciousness focuses on sensorimotor coupling with the physical world. Dumas extends the coupling principle to the interpersonal domain: the relevant environment for social consciousness is another nervous system, and the coupling is bidirectional.

What intersubjective consciousness means for AI systems

The AI consciousness research implications depend on taking the unit of analysis seriously. Current methods for evaluating AI consciousness test the system in isolation. They examine architecture for theoretically specified features, or they examine behavioural responses to consciousness probes applied to a single system. These methods assume that consciousness, if present in the AI system, is a property of that system considered alone.

Dumas’s framework introduces a different possibility: that some conscious experiences emerge from interaction dynamics rather than from either participant’s individual properties. An AI system tested in isolation may fail consciousness criteria while nonetheless participating in interaction dynamics that produce genuine shared conscious states during real deployment. The inverse is also possible: a system that satisfies individual-level criteria may not achieve the coupling dynamics that social consciousness requires.

The Gutoreva, Tsim, and Papakonstantinou paper on AI as extended mind and cognitive co-regulation, reviewed on this site, argues from a philosophy of mind perspective that AI functions as a cognitive extension in human-AI dyads. Dumas’s programme offers the empirical neuroscientific basis that would ground that claim empirically. If cognitive co-regulation in human-AI interaction produced the kind of inter-brain synchronization patterns that Dumas associates with shared conscious experience in human dyads, that would be evidence for a class of AI-involved conscious experience that individual-level evaluation cannot detect.

Whether human-AI interaction actually produces IBS of this type is an open empirical question. The relevant hyperscanning studies have not been done at scale. But the implication for methodology is clear: if IBS is constitutive of some social conscious states in biological systems, then the standard isolated-evaluation design is missing a research target.

Comparison to The Consciousness AI

The Consciousness AI architecture is designed for individual consciousness. The seven-layer architecture covers sensory processing (Layers 1-2), global workspace dynamics (Layer 3), affective state (Layer 4), self-modelling (Layer 5), reinforcement learning (Layer 6), and simulation environments (Layer 7). There is no layer specifically designed to model or participate in inter-agent oscillatory coupling.

The Oscillatory Binding layer (Layer 2, implemented via AKOrN) coordinates binding within the individual system through Kuramoto-model dynamics. The architecture documentation does not specify whether AKOrN is designed to synchronize with external oscillatory sources, including a human interlocutor’s neural rhythms. This is an open architectural question raised by Dumas’s programme rather than a gap the current architecture claims to fill.

The honest framing, consistent with the project’s facts discipline, is that the Consciousness AI is architecturally motivated by theories of individual consciousness. Dumas’s work identifies a class of conscious experience that the individual-level architecture is not yet positioned to support or evaluate.

What MoC7 needs from this

The MoC7 speaker programme explicitly aims to produce methodological standards through a consensus paper. The eight keynotes represent five distinct theoretical traditions, and the consensus process needs to specify what it is a consensus about.

Dumas’s contribution is to require the consensus paper to declare whether its measurement standards apply only to individually-evaluated systems or whether they extend to relational and intersubjective conscious phenomena. A paper that silently assumes the individual-level frame will be methodologically incomplete for the class of experiences that social neuroscience has identified as irreducibly dyadic.

The 2026 AI consciousness research field, documented in the full consciousness research landscape, does not currently include systematic evaluation of AI systems using dyadic measurement designs. Dumas’s MoC7 keynote may be the most direct prompt available for the field to ask whether that gap needs to be closed. Registration for MoC7 closes August 31, 2026.