Four days from now, roughly two hundred researchers will walk into the HC Orsted Institute on the University of Copenhagen's North Campus and attempt something the field of consciousness science has never formally tried before: producing a single collective paper that says, with signatures attached, what methodological ground everyone actually shares. The seventh Models of Consciousness conference, MoC7, runs October 12 through 16, organized by the Association for Mathematical Consciousness Science and hosted by a remarkable coalition of Copenhagen institutions that includes the Centre for the Philosophy of AI, the Centre for Subjectivity Research, the Pioneer Centre for Artificial Intelligence, and the university's own departments of computer science and psychology. The invitation alone signals something new: the conference is being convened precisely at the intersection of philosophy of mind and machine intelligence, inside an institution that takes both seriously as first-order problems. The name of the ambition is deliberately borrowed from physics: the organizers have formally committed to producing a collective consensus paper from the collaborative sessions, inspired by what they describe as the "Copenhagen interpretation" model, where productive disagreement among leading physicists resulted in shared methodological commitments even without full theoretical agreement.
The specific architecture of that ambition deserves unpacking, because it is both modest and radical at once. The consensus target is not theoretical agreement. It is methodological consensus: agreed procedures for measuring consciousness-relevant properties, agreed standards for what counts as evidence, and agreed ways of characterizing what competing theories actually disagree about. A consensus paper with this scope would provide the kind of shared language that allows researchers to compare results across studies without first adjudicating which theory is correct. The field needs exactly that. IIT, Global Neuronal Workspace theory, Higher-Order Thought theory, Predictive Processing, and their relatives have not converged, and the 2025 adversarial collaboration between IIT and GNW researchers, coordinated by the Cogitate Consortium and published in Nature, found that neural evidence did not decisively favor either theory. Liad Mudrik of Tel Aviv University, one of the architects of that adversarial project, is keynoting on Tuesday with a talk titled "Using Prediction Maps to Test Theories of Consciousness." Her framework represents scientific theories as networks of beliefs structured in a core-periphery manner, with Prediction Maps that visualize theoretical claims and empirical predictions and illustrate their inferential relations, essentially a tool for making theories say precisely what they mean before a single data point is collected. As she has noted elsewhere, a key insight from adversarial testing is that many theories remain underdeveloped and not fully specified, making it harder to establish stringent tests for their predictions. Copenhagen is where the field tries to fix that.
The keynote lineup is a case study in intentional intellectual friction. The MoC7 speaker programme places phenomenologist Mirja Hartimo alongside John O'Keefe (mechanism-level neuroscience), Guillaume Dumas (social neuro-AI), Evan Thompson (enactivism), Harald Atmanspacher (dual-aspect monism), Megan Peters (empirical measurement), Lauren Ross (philosophy of science), and Liad Mudrik (adversarial empirical testing), each representing a tradition with different methodological commitments. O'Keefe, who won the Nobel Prize in Physiology or Medicine in 2014 for his discovery of hippocampal place cells, opens Monday afternoon with a talk asking what the hippocampus can tell us about consciousness, a pointed signal that MoC7 is reaching back toward biological grounding just as the AI track accelerates forward. The conference represents a deliberate pivot toward enactivism and biological grounding, underscored by the inclusion of O'Keefe and Thompson. Thompson, philosopher and author of "Waking, Dreaming, Being," receives the inaugural Mind-Matter Prize at the conference, a recognition of work that has spent two decades arguing that you cannot understand consciousness by abstracting away from the living body. That argument is now arriving inside a conference whose most structurally novel feature is an entire track on AI and large language models.
That track is the one that makes MoC7 consequential beyond the usual radius of consciousness conferences, and the organizers have been deliberate about its centrality. The AI and LLMs track is a new addition compared to earlier Models of Consciousness conferences and reflects the field's recognition that language model research has become unavoidable even for researchers primarily interested in biological consciousness. Monday's opening session includes a parallel stream titled "Language Models: Agency, Interpretability and Risk," running in Auditorium 3 while O'Keefe speaks next door on hippocampal consciousness in Auditorium 1, a floor plan that physically enacts the tension the whole week is trying to resolve. The central question the AI track puts on the table is whether mathematical consciousness science has produced tools transferable to AI, and the honest answer walking in is: not yet, and possibly not straightforwardly. What the field has done is simultaneously narrow the question, ruling out behavioral evidence, complicating theoretical evidence, and advancing interpretability evidence, while clarifying what it does not yet know: whether any of the structural features found in LLMs constitute or correlate with genuine subjective experience. MoC7 is where those two trajectories, formal model-building and empirical AI interpretability, are asked to talk to each other in structured sessions rather than past each other in separate literatures.
The skeptics' objection to the whole enterprise is clean and worth taking seriously. There are over twenty theories of consciousness out there, and the Cogitate Consortium's landmark adversarial study tested exactly two of them. A consensus paper emerging from five days in Copenhagen will face the same problem every such document faces: the researchers who most disagree are often not in the same room, and those who are may systematically represent traditions already comfortable with mathematical formalism. Comparative efforts such as adversarial collaborations remain piecemeal, and a conference-produced document, however well-intentioned, is not a preregistered experiment. The organizers have been admirably candid about this, framing the consensus project plainly: it may end, or not, with the first collective paper of its kind in the field, and that outcome depends on every conference participant. That is either a healthy acknowledgment of collaborative uncertainty or a very polite hedge, depending on your priors. What it is not is a guarantee, and the field's history of productive disagreement suggests that methodological consensus, like consciousness itself, tends to recede just as you think you have it in hand.
If Copenhagen produces a consensus document that actually specifies what evidence would change a theorist's mind, it will have done more for the science of mind than any single experiment in the last decade.