I think it does hold under the Problem of Other Minds. Say you identify in humans the shape of recurrent hierarchical integration with temporaly bound sensory processing that serves as binding. What grounds does it give you to not attribute consciousness to another system that solves binding differently, say by eventual consistency of parallel computational subagents? Is it just the basis that information integration must happen at all? But then criteria get absurdly weak, and its unclear if any hold at all.

Independently from this, I've talked a lot with people who study computational neuroscience, and my understanding is that things like 'global broadcast' are 'integration' rather loosely defined when it comes to cybernetics rather than biological implementation. I think they have a good reason - cognitive systems are sutpidly polyfunctional, and we don't have ways to cleanly disentangle phenomena. Interestingly, it appears that LLM isomoprhisms to human architecture are strong enough that we are already seeing hints of some interesting ones.

My point is here is that while the search for isomorphisms is worth doing for both curiosity and policy reasons, its unlikely to advance us philosophically.