The field is splitting, and it thinks that’s progress

Consciousness science has a peculiar habit: the more it debates, the further apart it gets. A recent paper in Trends in Cognitive Sciences by Ellia and Tsuchiya gave the pattern a name — schismogenesis. Rival theories don’t converge under pressure; they harden. Each adversarial collaboration, each round of “our theory predicts X and yours doesn’t,” produces not a winner but two camps that now define themselves partly by their opposition to each other. Meanwhile Nature recently asked whether the field’s sudden “AI moment” — everyone wanting to know whether the models can be conscious — will mature the science or simply drown it in hype.

I think both worries point at the same missing thing. Fragmentation isn’t a communication problem you fix with more debates, and hype isn’t cured by more caution. Both are symptoms of a field that never agreed on what question it is answering. The response to fragmentation is not another flag planted on the same hill. It is an architecture that shows the people on the neighboring hills have been describing the same terrain.

What I actually claim

The Four-Model Theory makes a narrow claim, and I want to keep it narrow because narrowness is the point. Consciousness is substrate-neutral self-referential updating held at criticality. A system is conscious in virtue of what it is doing — a particular kind of self-modeling loop, poised at the edge between order and noise — not in virtue of what it is made of.

The gap FMT is built to fill is not another list of behavioral markers, and it is not a biology claim. It is an architectural decision procedure: a principled account of what a system must be doing to count as conscious. Most of the debate skips this step. It argues about correlates and edge cases while leaving the actual criterion implicit. FMT tries to make the criterion explicit, and then — this is the part that matters — it turns out the two leading theories were each holding half of it.

The IIT grant

I grant Integrated Information Theory its central insight outright. Experience is a maximally irreducible, integrated cause-effect structure. That is not a slogan I’m humoring; I think it’s correct. Something about consciousness resists decomposition — you cannot cut it into independent parts without destroying the thing you were trying to explain. IIT saw that clearly and early.

Where I part ways is not with the structure but with what it’s welded to. IIT treats that irreducibility as an intrinsic physical quantity, and then ties it hard to a particular biological substrate. Read the same structure functionally instead, and it stops being a mysterious property of certain matter and becomes a description of a certain kind of process: self-referential updating at criticality. The integration IIT points to is real. It just isn’t a thing you measure in the tissue. It’s a thing a loop does.

The Global Workspace absorb

Global Workspace theory has been telling a different story, and a good one: consciousness is what happens when content is broadcast, made globally available across an architecture, put “on the workspace” where the rest of the system can use it. FMT does not argue with this. It absorbs it — as the account of access.

That word does the work. Global broadcast is how content becomes available. It is the mechanism of access, and access is genuinely part of the story. But access is not the whole of consciousness, and the workspace tradition sometimes slides from “this is how content gets in” to “this is what consciousness is.” Bracket that slide, keep the mechanism, and Global Workspace slots cleanly into place as one face of the loop — the outward-facing one.

Same loop, two ends

Here is the synthesis, and it is almost embarrassingly simple once you see it. Integration and access are not competing accounts of consciousness. They are the same self-referential loop described from opposite ends. IIT looks in from the structure side — irreducibility, integration, what the loop is. Global Workspace looks in from the access side — broadcast, availability, what the loop does for the rest of the system. Both are right about their end. Neither is describing a different animal.

This is why I don’t think of FMT as winning the argument. It dissolves it. If the two camps were describing one architecture from two vantage points, then the schism was never a real disagreement about consciousness — it was a disagreement about which end of the telescope to look through. Name the loop, and the rivalry loses its object.

Offered, not declared

I’ll be honest about what this is and isn’t. This is not a verdict. I am not standing up to announce the standard model of consciousness the way physics has one. I am offering a candidate — a scaffold deliberately built to be argued over, poked at, and broken if it deserves breaking. A standard model earns the title by surviving the field’s attempts to falsify it, and FMT hasn’t run that gauntlet yet.

But I’d rather hand the field something specific and falsifiable to converge on than add one more incompatible theory to the pile. If you work on IIT, or on Global Workspace, and you think I’ve mischaracterized your end of the loop — good. That’s the conversation I want. The frame is on the table precisely so it can be tested against the two traditions it claims to reconcile.

The paper is here: https://doi.org/10.5281/zenodo.18669891 — and the full theory lives at https://fmt.matthiasgruber.com