Shedding light on the social dimension of consciousness
In one article, Dumas and his doctoral students Anne Monnier and Léna Adel explored “generative neurophenomenology.” Pioneered in the 1990s by Chilean neurobiologist Francisco Varela, neurophenomenology bridges two traditionally separate perspectives: classical neuroscience, which observes the brain from the outside, in the third person, and subjective, first-person lived experience.
“Generative” neurophenomenology adds an understanding of our lived experience as constructed through development, relationships, culture and history.
“Generative neurophenomenology attempts to connect three levels: first-person experience, third-person objective data and ‘second-person’ experience, meaning one's relationship to the other,” Dumas explained. “It is an approach that has run through my research since my doctoral thesis nearly 15 years ago. This article is a turning point that applies this vision in a modern approach.”
Monnier and Adel were the first members of the team to explore this complex subject.
Varela died an untimely death in 2001 at age 54, in Paris, and his approach, long neglected, has only recently been rediscovered. When combined with mathematical tools, computer models and advanced brain imaging, it opens up new possibilities in psychiatry, according to Dumas.
For example, generative neurophenomenology provides a fresh framework for understanding the therapeutic alliance, taking into account the patient’s experience, the therapist’s empathy and the dynamic between them, he said.
Autism seen through interaction
By highlighting relational atypicalities, generative neurophenomenology also offers a more precise and constructive approach to autism spectrum condition (ASC).
In a second study, Dumas’ team—led by Ghazaleh Ranjbaran, then a master’s student—used hyperscanning (simultaneous recording of the brain activity of two interacting people) and AI to predict whether the interaction involved individuals on the spectrum or neurotypical individuals.
This was challenging because AI models typically require vast datasets, and multi-brain data is scarce. The research team’s solution was transfer learning. They began by training their model on single-brain data to learn general brain signal patterns. Once this learning was complete, the model was duplicated to process two brains simultaneously. Finally, an additional layer learned to combine this information based on the interaction data.
“I’ve been working with hyperscanning for nearly 15 years,” Dumas said. “Since I came to Montreal and joined Mila five years ago, my goal has been to build a concrete bridge between artificial intelligence and multi-brain neuroscience.
"This study is one of the first to provide tangible proof that this synergy is possible when AI is applied to multi-brain neuroscience. This achievement was made possible by the interdisciplinary nature of my team—particularly Ghazaleh Ranjbaran."
Dumas believes this method could open the door to new possibilities and support autism diagnosis, potentially enabling early detection in infants through non-invasive analysis of brain signals.
Towards group neuroscience
Dumas’ research revolves around interpersonal psychiatry, moving beyond individual-focused models to consider relational dynamics. His interdisciplinary team—spanning neuroscience, computer science and psychiatry—embodies this approach.
When it comes to ASC, he calls for a paradigm shift to view social challenges as arising from the interaction between different neurotypes, not just from the individual on the spectrum. This aligns with the “double empathy problem” introduced by British researcher Damian Milton.
Beyond ASC, multi-brain neuroscience could illuminate group cohesion and support creative collaboration, team performance in sports and even the coordination of astronaut crews. Dumas believes.
“By combining generative neurophenomenology and AI, we are sketching the contours of a neuroscience of relationships—a science that explores not just isolated brains, but what emerges when minds connect,” he said.
